url
stringlengths 33
62
| title
stringlengths 10
115
| author
stringclasses 2
values | content
stringlengths 61
41k
| comments_count
int64 0
1.68k
| scraped_at
float64 1.76B
1.76B
|
|---|---|---|---|---|---|
https://www.phoronix.com/news/AMD-CES-2020-Announcements
|
AMD Announcements From CES 2020: Ryzen 4000 Mobile Series, Radeon RX 5600 XT
|
Michael Larabel
|
Kicking off the Consumer Electronics Show 2020 in Las Vegas, AMD CEO Lisa Su has begun with some interesting product announcements...
- The AMD Ryzen 4000 series mobile processors are coming this year. The Ryzen 4000 series mobile parts are based on Zen 2, 7nm (obviously), 20% lower SoC power, 2x performance-per-Watt is coming this year. The SoC features up to eight cores / sixteen threads and will be used in varying form factors and devices. The Ryzen 7 4800U will feature 8 Radeon cores, 8c/16t, 4.2GHz boost / 1.8GHz base, and a 15 Watt TDP. The Radeon cores are based on Vega but with better performance. The first laptops (reportedly 12+ designs) will come in Q1 while over the course of the calendar year is said to be 100+ systems with Ryzen 4000 series.
- The Radeon RX 5600 XT was officially announced. As expected, this Navi card "ultimate 1080p gaming" card has 36 compute units, 1375MHz game clock, 1560MHz boost clock, 6GB GDDR6 video memory. AMD is positioning the RX 5600 XT to beat out the GeForce GTX 1660 Ti. The AMD Radeon RX 5600 XT will be available 21 January at $279 USD. In H1'2020, a Radeon RX 5700M notebook GPU will also be released.
- AMD Ryzen 7 4800H with a 45 Watt TDP features 8 cores / 16 threads, 4.2GHz boost, 2.9GHz base for offering "desktop class performance" in mobile form factors. The first systems with the Ryzen 7 4800H coming out in February.
- Ryzen Threadripper 3990X announced as 64 cores / 128 threads, up to 4.3GHz boost, 288MB cache. The Ryzen Threadripper 3990X will retail for $3990 USD. The Threadripper 3990X will come 7 February.
That's a wrap. No word on Arcturus nor any bigger Navi cards for now.
| 44
| 1,760,719,518.603937
|
https://www.phoronix.com/news/AMD-TEE-Driver-For-Linux-5.6
|
AMD's Trusted Execution Environment Is Coming With Linux 5.6
|
Michael Larabel
|
Last week I wrote about the AMD Secure Processor support for Linux 5.6 being queued as part of the cryptography subsystyem work with supporting the PSP / Secure Processor of Raven Ridge APUs. That AMD Secure Processor support is now rounded out with the Trusted Execution Environment (TEE) driver being queued for wiring into that subsystem.
Added this Saturday morning to the crypto queue for Linux 5.6 was the AMD-TEE driver, complementing last week's queuing of the CCP support and PSP related bits for Raven Ridge APUs. The Trusted Execution Environment subsystem has been around in the mainline kernel for several years now for Arm SoCs and this new driver extends the support to the AMD processors with their Secure Processor built off Arm TrustZone.
The AMD-TEE driver is focused on AMD APU support, supports sharing memory between the normal and secure worlds, and acts as a "Rich Execution Environment" communication agent for submitting commands to execute on the secure processor. Running on the AMD Secure Processor meanwhile is the "AMD-TEE Trusted OS" that is loaded via firmware.
The AMD-TEE driver supports the GlobalPlatform TEE Client API in user-space as the generic means of interacting with the AMD support. The driver allows loading Trusted Application binaries into the environment, mapping/unmapping shared memory, and session handling with the loaded Trusted Application. The TEE is designed for areas like content protection / digital rights management, identification / authentication purposes, and other access rights handling.
As covered in previous articles, this AMD Trusted Execution support for their recent APUs appear to be part of AMD's Chromebook play in meeting the requirements set forth by Google, similar to AMD recently adding HDCP (High-bandwidth Digital Content Protection) support for Raven to their Linux driver. For those frightened by Trusted Execution / Secure Processor talk, on Linux 5.6+ you can always build your kernel with AMDTEE disabled so the AMD-TEE support will not be built.
| 38
| 1,760,719,519.549464
|
https://www.phoronix.com/news/AMDGPU-APU-noretry
|
A Possible Workaround For AMD APUs With Stability Issues On Recent Kernels
|
Michael Larabel
|
While we have found recent Linux kernels paired with latest motherboard BIOS releases to work out generally well for recent AMD APUs, not everyone has been having a trouble-free experience on recent kernels. But an affected user has discovered a possible workaround if hitting stability issues.
There has been a bug report since October stemming from graphics ring timeouts since Linux 5.2 and 5.3 kernels with multiple users affected. Those ring timeouts ultimately lead to graphics no longer functioning until rebooting the system.
The workaround discovered by a user is booting with amdgpu.noretry=0 as a kernel module parameter for AMDGPU. This comes after AMD set no-retry as the default in an effort to enhance performance by disabling retry faults.
Judging from this Gitlab issue, the amdgpu.noretry=0 and an affected user who alerted me to that ticket, it seems to be helping AMD APU owners on Linux ~5.4 that had been hitting stability issues. So should you be affected, give that option a whirl and chime in if it helped your situation.
| 17
| 1,760,719,520.187734
|
https://www.phoronix.com/news/Linux-5.4-4.19-4.14-MCE-Fix-TR
|
Linux 5.4.7 / 4.19.92 / 4.14.161 Bringing The AMD MCE Fix For New Threadripper CPUs
|
Michael Larabel
|
With the recently launched Threadripper 3960X / 3970X processors there was a workaround needed to boot them on Linux until an AMD MCE driver issue was resolved. That patch was upstreamed last week into the Linux 5.5 development kernel while now is getting ready to make its debut into supported Linux stable release branches.
Getting back onto his stable kernel maintenance duties now that Christmas week has passed, Greg Kroah-Hartman sent out his hundreds of patches back-ported for the Linux 5.4 / 4.19 LTS / 4.14 LTS stable series he oversees. Besides the kernel mailing list, the hundreds of patches queued for the next point releases in those branches can be seen via linux-stable-rc.git. Catching my eye from those patches is that the AMD MCE fix for the newer processors is indeed part of the queues for all three stable series.
So within a few days he'll have released them as Linux 5.4.7, Linux 4.19.92, and Linux 4.14.161 and with that the AMD MCE fix is ready to allow the AMD Ryzen Threadripper 3900 series to boot without the need for any workaround (booting with MCE disabled via mce=off) or having to use the Linux 5.5 Git kernel.
It's a pity that it took just over one month past the actual hardware shipping for this issue to be resolved with stable kernel series, but better late than never (in having to otherwise wait for Linux 5.5). We're crossing our fingers that AMD will be able to deliver more punctual Linux support/fixes with their hardware launches of 2020 and beyond.
| 4
| 1,760,719,521.426732
|
https://www.phoronix.com/news/Wraith-Prism-CM-RGB-Linux
|
Controlling AMD Wraith Prism RGB Heatsinks On Linux Is Easy Now With CM-RGB
|
Michael Larabel
|
With the Wraith Prism heatsink fan included with many modern AMD Ryzen processors there is configurable RGB lighting, which unfortunately AMD hadn't publicly documented or offered a Linux utility for manipulating the RGBs under Linux. Fortunately, there is now a straight-forward solution for dealing with those Wraith Prism RGB LEDs thanks to the open-source and independent CM-RGB project.
Just like AMD doesn't offer any CPU overclocking client from the Linux desktop, they don't offer any RGB control software for Linux. But CM-RGB is a Python-written independent utility that is command-line based and allows easily controlling the heatsink's lighting under Linux. The program allows setting the lighting mode, color based upon hex code, brightness, and other factors.
Those wanting to try out CM-RGB with a modern Wraith Prism heatsink can do so via GitHub and Python's PIP.
With AMD's continually improving finances thanks to their increasing successful product launches after laying off their Operating System Research Center years ago, hopefully in 2020 we see more support from them in the Linux space in areas like this for non-critical areas but "nice to have" features.
| 21
| 1,760,719,521.76034
|
https://www.phoronix.com/news/Linux-5.6-AMD-Secure-TEE
|
Linux 5.6 Adds TEE For AMD's Secure Processor To Run "Trusted Applications" On Raven APUs
|
Michael Larabel
|
Last month I wrote about AMD working on TEE driver support to load "trusted applications" onto the AMD Secure Processor under Linux. That work is now queued for introduction with Linux 5.6 and wired through for Raven Ridge APUs.
The AMD TEE (Trusted Execution Environment) support is for loading sensitive data and secure/trusted applications for running on the Arm-based AMD Secure Processor / PSP that's found embedded into modern AMD CPUs. While found on embedded through server AMD CPUs, with the Linux 5.6 support it appears all focused on Raven Ridge APUs - possibly due to Google Chromebook requirements, just as we've seen HDCP content protection support for Linux with Raven Ridge in recent kernels. The Chromebook angle would also explain why it's taken years for AMD to enable their Secure Processor usage with TEE under Linux until recently, just like their belated HDCP support and other work that has appeared to be pushed along by Google's interests.
The TEE support for Raven APUs with Linux 5.6 provides a "secure environment" for running Trusted Applications (TAs) and requires using their specific TEE interface for loading the binaries into the trusted environment, mapping shared memory, and related functionality.
Until the Linux 5.6 kernel cycle kicks off in about one month, the AMD TEE / Secure Processor patches are currently queued as part of the cryptodev code.
| 6
| 1,760,719,522.883929
|
https://www.phoronix.com/news/GCC-10-PGO-3960X-Xmas-Eve
|
GCC 10 PGO Benchmarks On AMD Ryzen Threadripper 3960X + Ubuntu 19.10
|
Michael Larabel
|
For those looking for some fresh reference numbers on the impact of using GCC's Profile Guided Optimizations (PGO), here are some benchmark runs looking at the GCC 10 PGO performance on an Ubuntu 19.10 workstation built around the Ryzen Threadripper 3960X.
I also have some fresh GCC 10 LTO optimization benchmarks coming in the next few days but using the PGO PTS module I ran some fresh benchmarks with just "-O3 -march=native" and then again after making use of profile guided optimizations for the benchmark runs.
More GCC 10 benchmarks will be coming as the stable release approaches towards the end of Q1'2020.
Of the rather random assortment of tests run, TSCP saw the biggest help from enabling PGO in 13% better performance for this chess benchmark.
The MT-DGEMM benchmark was also close to 13% faster after being PGO'ed.
The PGO impact for various other tests ran ranged from no measurable change up to just a couple percent better performance. More tests beyond these were ran but they ended up being even less interesting. Stay tuned for more end-of-year GCC and Clang benchmarks soon. If you enjoy our different open-source/Linux benchmarks you can't find readily elsewhere, consider showing your support this holiday season.
| 14
| 1,760,719,523.657321
|
https://www.phoronix.com/news/Linux-5.0-5.4-EPYC-7642
|
Linux 5.0 Through Linux 5.4 Benchmarks On AMD EPYC 7642 "Rome" Server
|
Michael Larabel
|
A month ago I posted benchmarks looking at the performance of Linux 4.16 through Linux 5.4 kernels using an Intel Core i9 workstation. Stemming from that was a request for an AMD EPYC kernel comparison, so I carried out said tests. Due to the Rome support being newer, this round of testing is looking at the EPYC 7642 performance on Linux 5.0 to Linux 5.4.
The tests were done last month but with the results not being too interesting, publishing them escaped my mind until this week firing up some of the Linux 5.5 kernel benchmarks. Those initial Linux 5.5 numbers for AMD EPYC should be out in the days ahead and should be interesting given some prominent changes with Linux 5.5 and other early performance numbers showing some interesting changes.
For this testing Linux 5.0 through Linux 5.4 Git was tested on the same EPYC Rome server.
But without any recent vulnerabilities and mitigations like was the case with Intel as recently as their November disclosures, the Linux 5.0 through Linux 5.4 performance with EPYC Rome wasn't all that interesting and mostly just the usual movement/noise we see in some select tests.
The geometric mean of all the test results basically point to flat performance for Linux 5.0 through 5.4, which is better than the performance hits seen in the recent Intel Linux kernel comparison.
A few real-world tests were trending up with Linux 5.3~5.4.
And some small kernel scheduler hits since Linux 5.3, but with 5.5 and the scheduler changes there the Hackbench numbers should be more interesting.
While not all of the same tests and other system differences, here was the i9 workstation results on Linux 4.16 through 5.4 for reference.
But for the most part of 76 tests ran, there wasn't much in the way of surprising movement for this EPYC testing of 5.0 through 5.4. The rest of the data can be found via this OpenBenchmarking.org result file. The upcoming Linux 5.5 results should be more interesting.
| 0
| 1,760,719,524.538246
|
https://www.phoronix.com/news/Threadripper-3000-MCE-5.5-Fix
|
Linux 5.5 Lands Fix For Booting New AMD Ryzen Threadripper Processors Without MCE Hang
|
Michael Larabel
|
The AMD machine check exception (MCE) code fix for Linux has landed ahead of this weekend's anticipated 5.5-rc3 release. This AMD MCE fix allows for the AMD Ryzen Threadripper 3900 series processors introduced last month to boot the Linux kernel without hangs or other workarounds.
The AMD Ryzen Threadripper 3960X / 3970X on Linux is beautifully fast but sadly out-of-the-box would not boot the Linux kernel due to an MCE error. Fortunately, there is an easy workaround to use for booting the Linux kernel by disabling MCE while now for Linux 5.5 (and to hopefully back-ported stat to current stable series) is a proper fix.
Nearly one month to the day since these high-end desktop/workstation processors launched, the fix was merged today along with some other AMD RAS fixes.
So if using Linux 5.5 Git today or a release of Linux 5.5-rc3 or later, the new Threadripper processors should be running fine out-of-the-box.
| 11
| 1,760,719,526.249395
|
https://www.phoronix.com/news/Zen-Microcode-2019-12-18
|
AMD Sends Updated Zen CPU Microcode To Linux-Firmware Tree
|
Michael Larabel
|
New AMD Family 17h (Zen) CPU family microcode was merged today into linux-firmware.git.
Though as is usually the case with the binary-only microcode/firmware files hitting the tree just not from AMD but from all hardware vendors, change-logs are usually non-existent and that is the case again today.
The update simply notes it's the 2019-12-18 Family 17h CPU microcode revision. The previous update was 2019-10-21.
The microcode is notably smaller in going from 9700 to 6476 bytes, which actually more closely mirrors the size of older Family 17h microcode files with this past October update having been the notably larger version.
I'm trying to find out more if there are any notable changes or just a routine update following the recent AMD CPUs launching and will also be doing some tests locally to see if spotting any performance changes. Stay tuned.
| 42
| 1,760,719,526.259252
|
https://www.phoronix.com/news/AMD-GPUOpen-GPUPerfAPI-3.5
|
AMD's GPU Performance API Library 3.5 Drops ROCm/HSA Support
|
Michael Larabel
|
Released on Friday was a new version of AMD's GPU Performance API "GPUPerfAPI" project under the GPUOpen umbrella. This is the AMD library used by CodeXL, Radeon Compute Profiler, and others for tapping GPU performance counters and to help in analyzing performance/execution characteristics for Radeon hardware. But this new GPUPerfAPI 3.5 release comes with a rather surprising change.
With the GPU Performance API library being used by the Radeon Compute Profiler and other prominent developer tools, we found it surprising that the GPU Performance API has dropped ROCm/HSA support. The release notes leave it simply as, "Remove ROCm/HSA support."Unfortunately no further insight on that decision if it's to be replaced by some better implementation later on or if there are Radeon Open eCosystem / HSA changes forthcoming... But seemingly rather odd as it's the AMD flagship compute stack at the moment and one would imagine developers specifically targeting Radeon GPUs would be very much still interested in profiling/analyzing ROCm execution unless some other changes are coming. At least the OpenCL support itself is still present along with OpenGL / DirectX / Vulkan.
But besides dropping ROCm/HSA support from this library, the GPU Performance API 3.5 release adds support for Radeon 5300/5500 series, per-API static counter generation, Unicode support for GPUPerfAPI on Linux, and many bug fixes for Vulkan and other APIs.
More details on the GPU Performance API 3.5 update via GitHub for this cross-platform library.
| 33
| 1,760,719,527.523774
|
https://www.phoronix.com/news/Threadripper-3-MCE-In-RAS-Core
|
AMD Threadripper 3900 Series MCE Fix Queued In RAS/Core But Not Yet Mainlined
|
Michael Larabel
|
The AMD Ryzen Threadripper 3960X/3970X are incredibly fast and trounce the competition, but as noted on launch-day most (all?) Linux distributions have a boot issue with them over a machine check exception. There is an easy workaround to let these core-happy CPUs boot and run Linux while the proper fix was queued last week in ras/core in what looks like it will wait until Linux 5.6 for merging.
As noted in the launch day article, AMD developers proposed a patch days ahead of launch for addressing that MCE issue with the new Threadripper systems. Though given the Linux 5.4 kernel stable release was just days later immediately followed by the Linux 5.5 merge window, that fix has yet to be merged into the mainline kernel or back-ported to any stable series.
Last week though the patch was picked up into the ras/core tree. Though as of writing isn't in mainline. And it's a bit fun the AMD MCE patches continue to be routed through Borislav Petkov thanks to his currently employment by SUSE (where he continues work on the upstream kernel, including maintainership of various areas of the kernel) where as he previously was an open-source AMD developer years ago before AMD closed their "operating system research center" when letting Linux developers go almost a decade ago... Given their current CPU successes, hopefully in 2020 we'll see them better ramping up their Linux support to avoid launch-day support blunders again. As it stands now, AMD's Linux engineering team is still very lean and tiny compared to Intel's Linux engineering resources.
Anyhow, unfortunately this patch hasn't been sent in as a "fix" for the ongoing Linux 5.5 cycle. Even with the patch fixing a boot issue and addressing "unexpected behavior" out of the system, the patch hasn't made it in yet nor does the commit message CC stable for immediate back-porting.
So with it quite possibly waiting until Linux 5.6, this is just a reminder for anyone that may find themselves having a new Threadripper system this holiday season that there is an easy workaround to boot the system in the interim of just disabling MCE on affected systems. If it doesn't land until Linux 5.6, hopefully distribution vendors like Ubuntu will take care of patching their kernel otherwise (Ubuntu 20.04 LTS will likely be riding Linux 5.5) to improve the out-of-the-box experience.
| 8
| 1,760,719,527.748868
|
https://www.phoronix.com/news/AMD-Ryzen-V1000-R1000
|
AMD Launches Ryzen V1000/R1000 Embedded Processors For Small Form Factor PCs
|
Michael Larabel
|
AMD today announced the Ryzen V1000 and Ryzen R1000 series embedded processors intended for mini PCs and other small form factor PCs.
These embedded processors are based on Zen CPU cores (sadly not Zen 2) and Vega graphics into a SoC solution ranging from 6 to 54 Watts.
More details on these new AMD Ryzen embedded SoCs can be found via AMD.com.
The folks at OnLogic are among the vendors announcing new mini PCs built around the Ryzen Embedded parts. The ML100G-40 and MC510-40 are these new PCs to begin shipping next month. Given our successes with the OnLogic Karbon 300 and Karbon 700, these AMD OnLogic boxes should be quite interesting and well built.
| 13
| 1,760,719,529.211441
|
https://www.phoronix.com/news/AMD-IOMMU-Linux-5.5
|
AMD IOMMU Driver Reworked For Linux 5.5
|
Michael Larabel
|
With the IOMMU updates for the Linux 5.5 kernel there is a major rework to the AMD IOMMU driver to make use of more common DMA IOMMU code for implementing the DMA API but with an admitted risk of potential new regressions.
The AMD IOMMU Linux driver now makes use of the "dma-iommu" kernel code that allows the driver to be lightened up by several hundred lines of code thanks to the code sharing/re-use. Besides the DMA IOMMU changes, the AMD IOMMU driver now also has multiple PCI DMA alias support.
The IOMMU updates acknowledge, "This gets rid of quite some code in the driver itself, but also has some potential for regressions (non are known at the moment)."
The IOMMU changes for Linux 5.5 also include Qualcomm SMMUv2 support with the SDM845 SoC, ARM SMMU driver changes, Renesas IOMMU driver changes working towards future hardware support, and other changes.
| 3
| 1,760,719,529.45412
|
https://www.phoronix.com/news/3970X-Follow-Up-Win-Linux
|
Threadripper 3970X Performing Better On Windows Relative To Linux - Thanks To Microsoft Or Zen 2?
|
Michael Larabel
|
With the AMD Ryzen Threadripper 3970X benchmarks on Windows 10 and Linux, Ubuntu 19.10 and other common distributions were just ~2% faster than the Microsoft OS and Clear Linux was just ~10% faster, based on 80+ benchmarks carried out. Those margins are much closer than we have seen with past iterations of Threadripper, but is that due to the Zen 2 microarchitecture and the improved topology of the new Threadripper CPUs or due to Microsoft's scheduler changes and other software improvements made in Windows 10 November 2019 Update? Here are some benchmarks.
One of the frequent questions that came up following that Windows 10 vs. Linux benchmark article yesterday was whether the relatively close Windows 10 vs. Linux performance was due to AMD Zen 2 / Threadripper 3970X itself or the ongoing software improvements Microsoft has been working on around its scheduler and other areas for dealing with today's HEDT processors.
For answering that question at least in part, I fired up some additional Windows benchmarks on the same system when instead running the Microsoft Windows 10 October 2018 Update as opposed to the latest feature release, Microsoft Windows 10 November 2019 Update. Here's a look at how comparing last year's and the current feature builds panned out:
The demanding SPECjbb 2015 was the biggest benefactor of the new Windows 10 release on Threadripper 3970X with a massive 30% improvement while 7-Zip and Renaissance's Java tests saw 12~18% improvements and other workloads at 8% or less. Shown are just the tests with measurable differences between the Windows 10 releases. With over 80 tests ran, the vast majority of them saw a negligible change in performance. In a few cases, the older Windows 10 build offered slightly better performance.
So while the Windows 10 November 2019 Update did help some workloads on Threadripper 3970X, the vast majority of the tests were within a margin of error to the previous Windows 10 release. Thus for the more competitive Windows 10 performance to Linux on the new Threadripper parts appear to be largely due to hardware improvements as opposed to Microsoft's software optimizations over the past year.
In case you missed it, see the 3970X Windows vs. 5 Linux distribution benchmarks from yesterday. If you enjoy our new and original content 365 days per year and often exclusive Linux hardware testing, consider showing your support this Black Friday / Cyber Monday.
| 11
| 1,760,719,530.690316
|
https://www.phoronix.com/news/AMD-TEE-Linux-Driver-Secure
|
AMD's TEE Driver For Loading "Trusted Applications" On Their Secure Processor Under Linux
|
Michael Larabel
|
A few weeks back AMD published a TEE "Trusted Execution Environment" driver for APUs on Linux for utilizing the controversial AMD Secure Processor.
The AMD Secure Processor / PSP is what's been built into their processors for a half-decade now for providing a secure hardware environment similar to Intel's Management Engine. This ARM-based secure processor is now seeing a Trusted Execution Environment driver for Linux.
The AMD-TEE driver is for volleying sensitive data / secure trusted applications. One of the motivating factors for this AMD APU TEE driver is for writing a trusted application providing Digital Rights Management for multimedia content protection. This goes along with what we've been seeing recently with the AMD graphics driver adding HDCP support and these security / rights management bits seemingly being done for the likes of AMD-powered Google Chromebooks.
It wouldn't be surprising if this TEE-based DRM is also being pursued for the Chromebooks use-case... There isn't much of an AMD APU desktop use-case otherwise for TEE/DRM and likely explains why this driver is only coming now while the AMD Secure Processor has been in hardware for years.
Their TEE driver amounts to just over one thousand lines of new code, building upon the kernel's existing TEE subsystem that to date has been mostly led by the ARM stakeholders. The driver allows for loading trusted application binaries into the trusted environment, starting/closing sessions, mapping shared memory, and related operations.
For those not fond of these trusted/secure features, once this code makes it into the mainline kernel can be disabled with the AMDTEE Kconfig switch.
| 9
| 1,760,719,531.136304
|
https://www.phoronix.com/news/Linux-Boot-Threadripper-Zen2MCE
|
The Workaround To Boot Linux On AMD Threadripper 3960X/3970X Systems
|
Michael Larabel
|
As outlined in our AMD Ryzen Threadripper 3960X / 3970X Linux review, these new Zen 2 Threadripper processors are phenomenal processors that offer significant uplift over earlier Threadripper CPUs and easily dominate over Intel's Core i9 HEDT competition. But there is one big issue right now with the Linux support: on Ubuntu and the like, it doesn't boot without a workaround. Here's that workaround for easy future reference.
See the details in our 3960X/3970X Linux article for more information, this is just a quick public service announcement for those that may have problems booting Threadripper 3900X series processors with TRX40 motherboards and wondering what's going on and looking for a solution...
With the two motherboards AMD supplied as part of the reviewer's kit and the AMD Threadripper 3960X/3970X, all the combinations yielded a system that appears to hang on boot like shown above. But as you can see, if just looking at the tail of the kernel output there isn't any hints. This doesn't seem to be an isolated issue either as other reviewers had asked me if/how to get the new Threadripper processors running on Linux. This is unrelated to the earlier RdRand trouble or other Ryzen/Zen issues with previous launches.
Fortunately, after figuring out it was an MCE issue and knowing about mce=off as a kernel module parameter. That worked for working around the issue.
So if you have boot hangs or issues booting a new Threadripper system, set mce=off as part of your kernel command line parameters. That should workaround the issue.
As pointed out in the formal article, after asking AMD's Linux engineers about the problem, they do have an MCE patch pending. However, as of writing that patch hasn't yet made it into the mainline kernel -- it's not in Linux 5.4 final or the early MCE/RAS changes for Linux 5.5. Hopefully it will make it in as a fix soon to the mainline kernel and get back-ported to stable.
So for now be aware of this one shortcoming in the Linux support but that it can be easily worked around by disabling the machine check exception kernel code. Now if you didn't already, go check out the impressive Threadripper 3960X/3970X Linux benchmarks. More tests are on the way too.
| 10
| 1,760,719,532.086724
|
https://www.phoronix.com/news/AMD-Athlon-3000G-Shipping
|
AMD's Athlon 3000G Processor Begins Shipping At $49 USD
|
Michael Larabel
|
Announced earlier this month, the Athlon 3000G is shipping today a week ahead of the Ryzen Threadripper 3000 series. The Athlon 3000G is AMD's new sub-$50 processor for lightweight desktop purposes.
The Athlon 3000G is a dual-core / four-thread processor with 3.5GHz clock frequency and 35 Watt TDP. The Athlon 3000G features Radeon Vega 3 graphics. The Athlon 3000G is compatible with existing AM4 motherboards and making it more interesting than the earlier Athlon 200GE is this is the first Zen-based Athlon processor to be unlocked for allowing some interesting overclocking possibilities with this low-power part.
The Athlon 3000G will begin appearing at Internet retailers with a suggested retail price of $49 USD. I'll be working on getting my hands on an AMD Athlon 3000G in the days ahead for some Linux benchmarking.
| 34
| 1,760,719,532.680248
|
https://www.phoronix.com/news/AMD-With-Intel-JCC-Assembler
|
Intel's Assembler Changes For JCC Erratum Are Not Hurting AMD
|
Michael Larabel
|
When writing about the Intel Jump Conditional Code (JCC) Erratum and how Intel is working to mitigate the performance hit of the CPU microcode update with patches to the GNU Assembler, there was some concern expressed by readers that it might hurt AMD performance. That does not appear to be the case.
On an AMD Ryzen 7 3700X box I installed Clear Linux and set it up in the same manner I used for this week's Skylake / Cascade Lake testing. In particular, comparing of Clear Linux builds 31470 and 31480 as that was the release where the patched version of the GNU Assembler was introduced and many bundles (packages) rebuilt as explained in the aforelinked article.
When running those tests on Clear Linux 31470/31480, there wasn't any real difference in performance. In just a couple of tests there were slight variations but nothing significant overall and also as pointed out in our earlier article the GNU Assembler behavior was odd in a few of the tests on Intel's own hardware.
So while some users were quick to speculate that it would hurt AMD or be some nefarious play by Intel, from this quick testing in repeating the same Clear Linux software setup as the JCC Erratum article, the updated assembler didn't introduce any real changes for the AMD Zen 2 system. Plus the assembler patches do add switches for controlling the behavior. Those wanting to look at the numbers in full anyhow can do so via this OpenBenchmarking.org result file.
| 28
| 1,760,719,533.561938
|
https://www.phoronix.com/news/AMD-7-November-Details
|
AMD Details 3rd Gen Threadripper, Ryzen 9 3950X + Their New $49 USD CPU
|
Michael Larabel
|
While the processors are not shipping until later this month, AMD is today making known their next batch of CPUs being released.
Pardon this brief article today as somewhat on paternity leave this week, but for the 25 November launch day will be all the interesting Linux-isms to talk about compatibility and performance. Being detailed today is the Ryzen 9 3950X, the first of the 3rd Gen Ryzen Threadripper processors, and even a new budget Athlon desktop processor.
The Ryzen 9 3950X specs are 16 cores / 32 threads, 3.5GHz base, 4.7GHz boost, 72MB L2+L3 cache, and 105 Watt TDP.
The Ryzen 9 3950X will cost $749 USD.
The Ryzen 9 3950X is designed to be much better than the Core i9 9900K and Core i9 9920X but AMD is already being forward in making known proper water cooling is recommended for adequate performance.
At the opposite end of the spectrum, the Athlon 3000G is shipping in mid-November as a $49 USD processor. The Athlon 3000G will offer 2 cores / 4 threads with a 3.5GHz clock frequency and Radeon Vega 3 graphics while combined is just a 35 Watt TDP. The Athlon 3000G is a Zen+ processor and is talked up as a better processor over the budget Athlon 200GE while taking on the likes of the Pentium G5400. One other key difference compared to the earlier Athlons is the 3000G is unlocked for offering up overclocking potential.
But the most exciting of today's disclosures is 3rd Gen Threadripper. The Ryzen Threadripper 3960X will launch at $1399 USD and provide 24 cores / 48 threads with a 3.8GHz base frequency and 4.5GHz boost with an incredible 140MB cache. The Threadripper 3970X meanwhile at $1999 USD is a 32 core / 64 thread part with 3.7GHz base / 4.5GHz boost and 144MB cache. These new Threadripper CPUs require the new sTRX4 socket that is an evolution of the TR3 socket. With that there is the new TRX40 platform with 88 PCI Express 4.0 lanes and other Zen 2 features compared to 2nd Gen Threadripper.
More details once these new processors begin shipping and then being able to deliver the actual performance figures. Pardon the short article today given the circumstances but it's certain to be an exciting remainder of the year with the new AMD parts as well as Intel's new Core i9 processors shipping later this month.
| 56
| 1,760,719,534.316326
|
https://www.phoronix.com/news/GNU-Binutils-Zen-2-Bits
|
GNU Binutils Adds Bits For AMD Zen 2's RDPRU + MCOMMIT Instructions
|
Michael Larabel
|
GNU Binutils with its "Gas" assembler has now added the rest of the instructions supported by the AMD Zen 2 microarchitecture that previously were unsupported by this piece of the GNU toolchain.
RDPRU and MCOMMIT are the two instructions for Zen 2 added to Binutils by SUSE's Jan Beulich. RDPRU has been covered multiple times on Phoronix and is for reading a processor register typically limited to privilege level zero. This allows for registers like MPERF/APERF to be easily read at user-level.
MCOMMIT on the other hand is an instruction set extension for Zen 2 that seems to actually be new to us. AMD hasn't mentioned MCOMMIT on their press slides nor commented on it when asking them previously at the July events about new instructions for Zen 2 or what we've seen in other Zen 2 toolchain patches to date.
AMD's tech doc only added the MCOMMIT instruction in September. MCOMMIT is used for committing stores to memory and provides a fencing / error detection capability for stores to system memory. The tech doc does note MCOMMIT needs to be explicitly enabled by the OS though when looking at the kernel, there hasn't been any MCOMMIT related kernel bits yet, so at least for Linux users is likely not enabled at this time even with compatible processors.
In any case, MCOMMIT and RDPRU are now plumbed into Binutils.
| 0
| 1,760,719,535.066966
|
https://www.phoronix.com/news/Renoir-DMCUB-AMDGPU-Patches
|
AMD Linux Graphics Driver Prepping "DMCUB" Support For Renoir APUs
|
Michael Larabel
|
While we have seen a lot of open-source AMD Linux graphics driver patches for Renoir and that initial support within the 5.4 kernel, support for this 2020 APU platform is still maturing. The newest work on the Linux upbringing for Renoir is enabling the "DMCUB" support.
DMCUB is short for the "Display MicroController Unit B" and is used with this next-gen APU for handling real-time display features like Panel Self Refresh and Adaptive Backlight Management as well as initializing the display hardware. From the software side, the DMCUB support is shortened to being "DMUB."
Bringing up this new display micro-controller within the AMDGPU kernel driver is over three thousand lines of code, but a good portion of that is header files.
Besides the kernel patches, there are also new firmware binaries required by the Display MicroController Unit B.
The patches for now are on amd-gfx while there's still time for potentially seeing this material land in DRM-Next ahead of the Linux 5.5 cycle. Renoir APUs are expected in 2020 with Zen 2 CPU cores while continuing to make use of Vega graphics but with a new display engine and other improvements compared to Raven 2 / Picasso.
| 10
| 1,760,719,535.843437
|
https://www.phoronix.com/news/Linux-5.5-AMDGPU-More-Features
|
AMD Sends In Another Batch Of Feature Changes For Linux 5.5 AMDGPU Driver Code
|
Michael Larabel
|
Complementing the big AMDGPU feature pull request from two weeks ago, on Friday AMD sent out a second batch of features targeting the upcoming Linux 5.5 kernel merge window.
On the feature front, the main change with this pull request for Linux 5.5 is the BACO support for older AMD GPUs. Those Radeon 200/300 series parts are seeing BACO support for possible power-savings ultimately but the short-term focus is using it for GPU reset functionality. BACO is already supported on the newer AMD GPUs.
- BACO (Bus Active, Chip Off) support for Sea Islands and Volcanic Islands GPUs.
- Faster memory training for Navi GPUs.
- MSI-X support.
- RAS fixes.
- Various Renoir APU fixes.
- PowerPlay fixes.
- Other code clean-ups and fixes.
More details via the pull request. The feature window cut-off for new bits to DRM-Next ahead of Linux 5.5 is closing off within the next week or so, we'll see if AMD volleys anymore changes for this next kernel cycle or they begin shifting focus to what features they want to introduce with Linux 5.6.
| 2
| 1,760,719,536.467993
|
https://www.phoronix.com/news/AMD-Joins-Blender-Foundation
|
AMD Joins The Blender Foundation With An Emphasis On Vulkan
|
Michael Larabel
|
Just earlier this month NVIDIA announced their funding of the Blender Foundation at the flagship "patron" level and now AMD has followed them in backing this foundation for assisting the development of this leading 3D creation software.
AMD now joins NVIDIA and Epic Games at the patron level, which means contributing at least €120k per year to the foundation.
The Blender Foundation shared that the money will be used for general development as well as for supporting their migration to supporting Vulkan and supporting other AMD technologies.
Today AMD joined the Blender Foundation Development Fund at Patron level. We will invest it in general development, Vulkan migration and to keep AMD technologies well supported for our users. Thanks a lot! https://t.co/vFdhNcJbOR #b3d— Blender (@blender_org) October 23, 2019 This is great news and will hopefully mean seeing great Blender Vulkan acceleration in the not too distant future. While this is the first time they have supported the foundation at the patron level, this isn't AMD's first rodeo with Blender as in the past they have collaborated over OpenCL improvements as well as the AMD Radeon ProRender for Blender, among other efforts.
More details on the Blender Development Fund at fund.blender.org.
| 12
| 1,760,719,537.428066
|
https://www.phoronix.com/news/AMD-Zen-201910-CPU-Microcode
|
Updated AMD Zen CPU Microcode Lands In Linux-Firmware Tree
|
Michael Larabel
|
Updated AMD Family 17h (Zen / Zen 2) CPU microcode has quietly landed within the linux-firmware Git tree.
But as is often the case with new additions to linux-firmware.git, the changes to said microcode/firmware binaries aren't usually described in any level of detail. Though with this Zen CPU microcode update it ultimately shouldn't mean too much assuming you are punctual with your motherboard firmware updates that generally ship with the new AMD CPU microcode revisions, in which case the older (in-tree) firmware isn't loaded.
This is though the first AMD Family 17h microcode update to the linux-firmware.git tree since November of last year, so for those relying upon this Linux bundled CPU microcode, it's much newer. The AMD Family 17h microcode version is marked as "2019-10-21" so it is possible that there are some newer bits compared to what is currently bundled with the latest motherboard firmware updates.
I'm reaching out to AMD to try to find out more on any prominent changes to this microcode update / their motivation for now getting the Zen microcode updated for the linux-firmware tree or if it's simply a matter of now having gotten around to it. In fact, it's only the third time the AMD Zen CPU microcode has been updated within linux-firmware.git.
The commit with the updated binary for those interested in fetching it directly. Due to now having the Zen 2 bits and other changes over the past year, the binary size does inflate from 6476 to 9700 bytes.
| 12
| 1,760,719,537.977022
|
https://www.phoronix.com/news/AMD-znver2-Scheduler-LLVM-Pend
|
AMD Zen 2 Improvements For LLVM Have Been Held Up For Months By Code Review
|
Michael Larabel
|
Back in February for LLVM Clang 9.0 was the initial AMD Zen 2 "znver2" enablement, but like the GCC support at the time it was the very basics. With time GCC picked up Zen 2 scheduler improvements and other work while sadly in the case of LLVM the improvements are still pending.
Back in August, AMD's Ganesh Gopalasubramanian sent out the znver2 scheduler model for LLVM for Zen 2 CPUs but a focus on the EPYC 7002 "Rome" processors. "There are few improvements with respect to execution units, latencies and throughput when compared with znver1. The tests that were present for znver1 for llvm-mca tool were replicated. The latencies, execution units, timeline and throughput information are updated for znver2."
That's all great with AMD Ryzen 3000 and EPYC 7002 series processors now shipping for months. But sadly that Zen 2 scheduler model hasn't been merged yet with being held up by needing code review. Hopefully that review can happen soon -- especially with AMD having contributed the code themselves -- in order to make it for the LLVM Clang 10.0 release in early 2020 but a pity that it's been held up while these new AMD processors have already been shipping for months.
For those wishing to have an AMD Zen 2 optimized compiler experience right now without resorting to patching your own LLVM/Clang setup, the AMD Optimizing C/C++ Compiler 2.0 release from August does have all of their LLVM patches in place for the znver2 target.
Moving forward, this is one area where AMD can improve upon with more punctual GCC/Clang compiler support in place. After all, Intel tends to upstream their microarchitecture enablement for the compilers usually at least one year in advance to ensure that the optimized support works its way into released versions of the compilers -- and found within shipping Linux distributions -- by the time their CPUs ship.
| 30
| 1,760,719,538.961722
|
https://www.phoronix.com/news/AMD-RDPRU-Linux-5.5-Feature
|
Linux 5.5 To Advertise RDPRU Support For AMD Zen 2 CPUs Via /proc/cpuinfo
|
Michael Larabel
|
RDPRU is one of the new instruction set extensions of AMD "Zen 2" CPUs that is for reading a processor register that is typically limited to privilege level zero. RDPRU allows for reading select registers at any privilege level. With Linux 5.5, the RDPRU presence will be advertised by the CPU features.
It's still up to user-space for making use of RDPRU, but for software checking from /proc/cpuinfo to see the availability of RDPRU as a supported CPU feature, Linux 5.5 is finally set to advertise it for the Zen 2 CPUs with the Ryzen 3000 series and AMD EPYC 7002 series.
Queued now as part of the x86/cpu changes building up for the Linux 5.5 cycle is the AMD detection for this instruction. The instruction can, of course, be used on earlier kernels this is just about cleanly advertising it through the common cpuinfo area.
This is the only kernel activity around RDPRU besides a KVM change from September for intercepting the RDPRU instruction.
| 7
| 1,760,719,539.39382
|
https://www.phoronix.com/news/AMD-Radeon-RX-5500-Navi-14
|
AMD Announces Navi 14 Based Radeon RX 5500 Series
|
Michael Larabel
|
AMD today lifted the lid on the Radeon RX 5500 series as their first Navi 14 based graphics card. This is the soft launch with no units shipping yet but expected to starting in November.
The Radeon RX 5500 series offers 128-bit GDDR6 video memory, 1408MHz shader clock, a boost clock up to 1845Mhz, 22 compute units, and other features in common to the Navi Radeon RX 5700 series. The Radeon RX 5500 series should offer comparable performance to the Radeon RX 570 Polaris graphics card.
Besides the Radeon RX 5500 desktop model, a Radeon RX 5500M mobile version is coming as well. The specs for the desktop and mobile parts are up at AMD.com.
More details once we get our hands on the Radeon RX 5500, which should hopefully be in the next few weeks. There is Radeon RX 5500 series support in Linux 5.4, but hidden behind its experimental feature flag, and the OpenGL/Vulkan support within mainline Mesa starting with Mesa 19.2 but the best support in Mesa 19.3-devel.
| 43
| 1,760,719,540.519593
|
https://www.phoronix.com/news/Radeon-ROCm-2.9-Released
|
Radeon ROCm 2.9 Released With New "RALI" Library, rocTX
|
Michael Larabel
|
Just one week after the release of Radeon Open eCosystem 2.8, AMD has now released ROCm 2.9 as the newest feature release for this open-source GPU Linux compute stack for Radeon hardware.
Radeon Open eCosystem 2.9 introduces the Radeon Augmentation Library "RALI" for efficient decoding and handling of images from a variety of formats via a programmable processing graph. ROCm 2.9 also introduces rocTX as a new C API for performance profiling.
ROCm 2.9 also brings fp16 and int8 quantization with MIGraphX 0.4, rocSparse csrgemm, Singularity container version 2.5.2, and other changes. Also notable is that Ubuntu 18.04.3 LTS is now supported with ROCm 2.9.
Notably absent is no Navi support with ROCm 2.9. AMD has remained quite about if/when they intend to support the current Navi graphics processors with ROCm. With the upcoming "Arcturus" GPU being Vega derived and appearing to be their next compute accelerator, it appears they aren't in a rush for supporting Navi.
Downloads and more details for ROCm 2.9 via RadeonOpenCompute on GitHub.
| 13
| 1,760,719,541.345737
|
https://www.phoronix.com/news/AMD-FreeSync-2019-Update
|
It Looks Like HDMI FreeSync/VRR For Linux + Wayland Support Will Eventually Come For AMD
|
Michael Larabel
|
AMD provided an update on their Linux FreeSync/Adaptive-Sync support at this week's X.Org Developers Conference event in Montreal. There's good news both for HDMI and Wayland Linux users with Radeon graphics.
Harry Wentland, a longtime member of the AMD driver team and patch wrangler around the DC display code, was the presenter at XDC2019. One of the big complaints of the AMD FreeSync/Adaptive-Sync Linux support to date has been the lack of supporting HDMI outputs... Wentland explained that on Windows, AMD uses a proprietary AMD-developed protocol for enabling FreeSync on HDMI. Obviously that won't fly for the open-source AMDGPU kernel driver. But as for the formal HDMI Variable Rate Refresh (VRR) support, they note it's "pending" but held up by a HDMI VRR conformance test suite being released. So hopefully once that CTS is available, HDMI VRR will be flipped on for Linux users wishing to enjoy Adaptive-Sync/VRR functionality for HDMI displays.
On the user-space side, RadeonSI OpenGL, Orace (proprietary AMDGPU-PRO driver) OpenGL support, and RADV Vulkan driver support. No comments were made on whether they intend to support FreeSync/VRR with their open-source AMDVLK Vulkan driver.
The other notable limitation to the AMD variable rate refresh support right now is that it's only supported on X11/X.Org desktops. There isn't VRR Wayland support at this time but it was mentioned among the "next steps" with possible VRR support for Wayland's Weston / KDE Plasma / GNOME Shell and even potentially ChromeOS. Other next steps include working on FreeSync support for Linux video playback, other power savings possibilities, and potentially exposing a more explicit interface for enabling other future use-cases around the technology.
Those curious about more details on the current AMD FreeSync Linux support can see the PDF slide deck from XDC2019.
| 24
| 1,760,719,542.118612
|
https://www.phoronix.com/news/AMD-EPYC-Linux-5.4-Early
|
An Early Look At The AMD EPYC Performance With The In-Development Linux 5.4 Kernel
|
Michael Larabel
|
While the Linux 5.4 cycle just officially began last week and its feature merge window not even over until this weekend, given there are AMD EPYC load balancing improvements and many other kernel improvements in general, I was eager to fire up the in-development kernel on the EPYC 7002 "Rome" series to see how the performance is looking.
With the Linux 5.4 Git state as of a few days ago, I ran some preliminary benchmarks of Linux 5.4 at the time compared to Linux 5.3.0 and Linux 5.2.16 on the Rome "Daytona" reference platform with the EPYC 7642 and EPYC 7742 processors in both 1P and 2P configurations.
Via the Phoronix Test Suite a range of benchmarks were done for looking at the Linux 5.4 performance at this very early stage. As the Linux 5.4 release candidates begin, tests on more hardware will obviously come -- these are just some early numbers with being benchmarking junkies.
Aside from swapping out Linux 5.2 vs. 5.3 vs. 5.4 branches as well as the different EPYC 7002 series processors, the hardware/software remained the same throughout testing.
While there is the EPYC load balancing change, with no major power management changes or other big ticket items for Linux 5.4, we weren't expecting any dramatic differences especially with workloads not really interacting directly with the kernel. With tests like NAMD, the performance was basically flat between the tested kernels.
With the highly-threaded renderers like Blender and Tungsten, the Rome 2P processors did appear to be lifted ever so slightly by the Linux 5.4 kernel but within the margin of error and we'll need to test on more CPUs for any level of certainty.
While firing up the Stress-NG kernel micro-benchmarks there was a notable increase in performance on the EPYC 7642/7742 processors in their dual socket configuration.
The Apache web server performance did see some improvement off this early Linux 5.4 kernel snapshot as well sans the EPYC 7642 1P.
That's the brief glimpse of the Linux 5.4 performance on AMD EPYC 7002 series in some of the early testing thus far. Stay tuned for more benchmarking as the Linux 5.4 kernel matures.
| 2
| 1,760,719,542.948026
|
https://www.phoronix.com/news/AMD-3rd-Gen-TR-3950X-November
|
AMD Pushes Back 3rd Gen Threadripper & Ryzen 9 3950X Until November
|
Michael Larabel
|
While the Ryzen 9 3950X and 3rd Gen Ryzen Threadripper processors were reportedly on track for launching in October with updates as of a few weeks ago, today AMD announced a slight delay in launching these new processors.
AMD has just shared the following statement:
We are focusing on meeting the strong demand for our 3rd generation AMD Ryzen processors in the market and now plan to launch both the AMD Ryzen 9 3950X and initial members of the 3rd Gen AMD Ryzen Threadripper processor family in volume this November. We are confident that when enthusiasts get their hands on the world’s first 16-core mainstream desktop processor and our next-generation of high-end desktop processors, the wait will be well worth it. So the Ryzen 9 3950X 16-core processor and 3rd Gen Ryzen Threadripper processors will be delayed one month.
AMD also shared a 3rd Gen Threadripper image that does indicate these new HEDT CPUs will start with 24-core options.
| 36
| 1,760,719,543.703506
|
https://www.phoronix.com/news/AMDGPU-Initial-5.4-Fixes
|
AMD Sends In Initial Batch Of Fixes To Linux 5.4 - Includes Dali Support
|
Michael Larabel
|
While just yesterday the big DRM feature pull was sent in for Linux 5.4, AMD has also volleyed out their initial batch of fixes for this next version of the kernel.
This new AMDGPU pull isn't strictly fixes but as anticipated does include the recently reported Dali APU support. Dali along with Renoir -- also newly-supported in Linux 5.4 -- are some of AMD's 2020 APUs. Dali will be targeting the lower-end of the spectrum it's expected for value mobile/embedded. From the driver code, Dali looks like a newer revved version of the current-gen Picasso APUs. Both Dali and Renoir are based on the Vega architecture.
In addition to this AMDGPU fixes pull having the Dali enablement, there are also more Navi 12 and Navi 14 PCI IDs -- including for the Navi "workstation" parts. But the Navi 12/14 devices are now flagged as "experimental" for Linux 5.4 and hidden behind a feature flag.
As for the fixes, this pull does have a fix for bandwidth issues when driving multiple displays on Vega 20, a possible oops in the AMDKFD compute code for Hawaii GPUs, a backlight level fix after suspend-and-resume for some APUs, and a few other random fixes. Fortunately, at least, Navi 10 (Radeon RX 5700 series) doesn't have any big fixes in this pull so it's looking like it should be fairly stable with the current code... I'll be doing some Linux 5.4-rc1 Radeon RX 5700 / 5700 XT tests next week.
The complete list of AMDGPU fixes for this pull can be found on dri-devel.
| 4
| 1,760,719,544.403996
|
https://www.phoronix.com/news/AMD-EPYC-7H12
|
AMD EPYC 7H12 Announced As New 280 Watt Processor For High Performance Computing
|
Michael Larabel
|
From Rome, Italy this afternoon AMD not only announced more than 100 world records have been broken with their new EPYC "Rome" processors, but there is also a new SKU! Meet the EPYC 7H12.
The EPYC 7H12 doesn't quite follow the naming convention of the rest of the EPYC Rome line-up announced back in August as it's a special part. The EPYC 7H12 is more akin to Intel's Cascadelake-AP line-up but with more broad availability and just a higher clocked / higher power part as opposed to tacking on extra dies. But it carries the same focus on delivering maximum HPC performance.
The AMD EPYC 7H12 is a 280 Watt SKU, their most power hungry EPYC processor ever, and as such is primarily designed for server deployments with water cooling. Compared to the 225 Watt top-end AMD EPYC 7742, the 7H12 model raises the base frequency from 2.25GHz to 2.6GHz and the boost frequency is 3.3GHz, which is actually 100MHz lower than the 7742 model. But the rest of the specs besides the clock adjustments and TDP are in-line with the EPYC 7742: 64 cores / 128 threads, 256MB L3 cache, 128 PCIe 4.0 lanes, 8 x DDR4-3200 memory support.
This water-cooled server processor is, of course, designed for the high performance computing (HPC) use-cases and similar. The EPYC 7742 already easily beats the likes of the Intel Xeon Platinum 8280 "Cascade Lake" while the 7H12 with the extra 350MHz on the base frequency should provided an added oomph. In at least Linpack, the EPYC 7H12 delivers around 11% higher performance over the EPYC 7742.
Pricing hasn't been shared yet and, of course, you'll need a specialized server setup to accommodate the 280 Watt TDP.
From the Rome event in Rome, Dell EMC announced their five new PowerEdge servers for these new EPYC CPUs, TSMC themselves are beginning to make use of EPYC processors for R&D, Nokia and IBM Cloud are backing EPYC Rome, and OVHcloud announced they will be providing EPYC 7402P access in the cloud.
| 15
| 1,760,719,545.281771
|
https://www.phoronix.com/news/Linux-5.4-Hwmon-Zen-2-Thermal
|
Linux 5.4 Brings Working Temperature Reporting For AMD Ryzen 3000 Series CPUs
|
Michael Larabel
|
One of the early pull requests for the just-opened Linux 5.4 kernel merge window is the hardware monitoring "hwmon" subsystem changes. This time around the notable change is CPU temperature reporting for Ryzen 3000 series processors.
Due to a combination of poor timing and an oversight at AMD, the CPU temperature reporting under Linux for the Ryzen 3000 series processors isn't in order until this new Linux 5.4 cycle. Back at the Ryzen 3000 series launch event I was told everything was "all good" from the Linux support perspective for thermal monitoring, after having been closely following the situation for past Zen CPUs and ended up myself adding the Linux CPU temperature monitoring support for Threadripper 2 among other hurdles in the past. That all-good though just ended up meaning that there is no Tcontrol offset needed for these new CPUs, which is great news no longer needing the temperature offset by an arbitrary amount. But the oversight was the Family 17h Model 70h ID was never added to the AMD k10temp driver. As a result, temperature monitoring wasn't actually working and took an extra kernel cycle before this trivial addition landed.
With the hwmon changes for Linux 5.4, the ID is in place along with this associated patch that is also a trivial addition. Sadly these patches weren't sent into Linux 5.3 weeks ago as a "fix" but only now going to mainline with the Linux 5.4 merge window. But it's quite possible now the patches will be back-ported to a Linux 5.3 point release. It's frankly silly that AMD couldn't have got these minor additions made as part of their Linux support months ago well ahead of the Zen 2 launch.
I've confirmed with these patches all is well for thermal monitoring on the Ryzen 3000 series processors. Sadly there is no mainline driver for CPU package power reporting with AMD Ryzen processors, but the out-of-tree "Zenpower" driver has worked for reporting this driver.
Anyhow, this k10temp addition is the main change to find with the hwmon updates for Linux 5.4.
| 7
| 1,760,719,546.849259
|
https://www.phoronix.com/news/AMD-Dali-Initial-Linux-Start
|
AMD Begins Posting "Dali" APU Linux Graphics Driver Patches
|
Michael Larabel
|
In addition to AMD's open-source Linux driver developers being busy in recent weeks bringing up the Renoir APU support, today we've seen the first baby steps towards bringing up "Dali" as another upcoming AMD APU.
The Dali codename has been known for a few months now and has been expected to be a value/mobile APU to launch in 2020. Dali is expected in more budget devices while Renoir should carry better performance up the stack.
Today was the first patches to the AMDGPU Linux driver albeit volleyed so far were just the initial DC "Display Core" patches that add in the new ASIC ID and also a second patch limiting the voltage level explicitly for Dali.
Dali is being treated as just a newer revision of the AMD Picasso APU support. The patch messages also confirm that Dali is a new revision derived from Raven 2 / Picasso hardware.
While we're still waiting on more Dali patches to come out, at this stage it's looking to be a near facsimile to current AMD Vega-based Picasso APUs and not featuring any new display or video blocks like we're seeing with the beefier Vega-based Renoir.
Given the timing of these initial AMD Dali patches coming, the support will first come to the Linux 5.5 kernel cycle kicking off late in the year and reaching stable in early 2020. Linux 5.5 is also likely the kernel we'll find in use by the Ubuntu 20.04 LTS OS release in April, so hopefully the Dali (and Renoir) support will be in good shape for that kernel revision.
| 5
| 1,760,719,547.038888
|
https://www.phoronix.com/news/Ryzen-9-3900X-More-Power
|
PowerTop, AMD CPUFreq CPPC & Other Power Tests From The Ryzen 9 3900X On Linux
|
Michael Larabel
|
Continuing on from last week's testing that found the AMD Ryzen 9 3900X + ASUS CROSSHAIR VIII HERO WiFi consuming more power on Linux compared to Windows 10, here are some additional metrics after spending a good deal of time over the weekend on further tests.
From the aforelinked tests last week, a number of Phoronix readers simply suggested the difference be due to the default Linux behavior and that the AMD power consumption would be more in line with Windows if running Intel's PowerTop. So I did.
Idle load was about three Watts less for the overall AC system power consumption, the average power draw under idle and load came out to 2~3 Watts, and the peak AC power draw was just one Watt lower. That's the out-of-the-box state compared to setting all the PowerTop tunables to their "good" state. So it saved 1~3 Watts, but still off the 10~30 Watt differences seen on the same hardware between Ubuntu 19.04 (plus Linux 5.3) and Windows 10 1903.
One interesting discovery made was that the Linux 5.3 power consumption on this system is 8~10 Watts lower than pre-5.3 kernels... BUT my original Windows vs. Linux tests were already done with a Linux 5.3 Git kernel. So in fact the Windows vs. Linux power delta can be even worse than shown last week if using Linux 5.2 stable or older:
Across various workloads, using Linux 5.3 had on average consumed about 8 Watts less while the peak power consumption was down by 11~12 Watts for Linux 5.0 through Linux 5.2. The idle / minimum power draw was largely unchanged.
Linux 5.3 didn't regress the performance, so this is a slight improvement in power efficiency / performance-per-Watt with Linux 5.3. So for those AMD Zen 2 users running pre-5.3 kernels, the stable Linux 5.3 release in two weeks is something to look forward to.
Another thought for the possible Windows vs. Linux difference came down to the CPPC (Collaborative Processor Performance Control) support being in good shape for Windows but not Linux. AMD developers recently posted the new AMD CPUFreq CPPC driver for Zen 2 CPUs but as of Linux 5.3 that isn't merged and even for Linux 5.4 I haven't seen that driver queued in the power management tree. But I did a fresh kernel build with those new amd-cpufreq driver patches to see if it would explain the difference.
In a few workloads using that CPUFreq driver led to a slight performance bump, but most of the time was flat.
Except that using that driver led to an increase in power usage, not lowering it.
So at this stage I haven't found any combination (short of say switching to the "conservative" CPUFreq scaling governor) for lowering the power consumption to make it comparable to Windows. But the anecdotes around better power efficiency on Linux 5.3 will please Ryzen 3000 users (I still have to benchmark it with the power meters on other systems to see if this is a widespread improvement) while PowerTop wasn't too useful for the high-end desktop. The AMD CPUFreq CPPC driver also looks like it may help the performance of some workloads but at least in its default behavior hadn't helped the power usage. For those enjoying the investigative benchmarks and wanting to see more of them, consider showing your support by joining Phoronix Premium.
| 16
| 1,760,719,548.457577
|
https://www.phoronix.com/news/AMD-Zen2-Boost-Optimizations
|
AMD Firmware Update To Bring Boost Performance Optimizations
|
Michael Larabel
|
There has been a lot of talk recently of AMD Ryzen 3000 series processors reportedly not hitting their boost clock frequencies, whether stock coolers are adequate for hitting the boost frequencies, and other concerns around the boost behavior on these new Zen 2 processors. AMD issued a statement today they are rolling out a new BIOS/firmware update to help with boost clock frequency optimizations.
AMD says they found a firmware situation that reduces boost frequency in some situations. They will be working with their motherboard partners to send out the updates providing these boost performance optimizations.
AMD says more information will be available next week.
pic.twitter.com/nX5za4yjH2— AMD Ryzen (@AMDRyzen) September 3, 2019
| 28
| 1,760,719,548.549637
|
https://www.phoronix.com/news/Linux-5.4-Improve-EPYC-Balance
|
Linux 5.4 Kernel To Bring Improved Load Balancing On AMD EPYC Servers
|
Michael Larabel
|
Adding to the growing list of features for Linux 5.4 with its cycle officially kicking off in mid-September is a kernel scheduler optimization designed to improve load balancing on AMD EPYC servers.
The scheduler topology improvement by SUSE's Matt Fleming changes the behavior as currently it turns out for EPYC hardware the kernel has failed to properly load balance across NUMA nodes on different sockets.
AMD EPYC/Zen processors now overrides the node reclaim distance to better account for the CPU's architecture. From one of the code comments, "AMD EPYC machines use this because even though the 2-hop distance is 32 (3.2x slower than a local memory access) performance actually *improves* if allowed to reclaim memory and load balance tasks between NUMA nodes 2-hops apart."
The change goes into more details and is part of the core scheduler changes queued ahead of the Linux 5.4 merge window opening up in two weeks.
| 5
| 1,760,719,549.823329
|
https://www.phoronix.com/news/AMD-Linux-5.4-Next-Fixes
|
AMD Has A Number Of Graphics Driver Fixes To Add For Linux 5.4
|
Michael Larabel
|
In addition to the new hardware support and other features queued already in DRM-Next for the upcoming Linux 5.4 merge window, on Friday AMD sent in a final pull request to DRM-Next of new material ahead of this upcoming kernel cycle.
Though with the Linux 5.4 merge window coming in soon and the soft deadline for new DRM feature work having passed, this new pull request from AMD is mostly about delivering fixes. But there is one big exception to that and it's the Renoir display support for those recently post patches wiring up "DCN 2.1" and other Display Core bits for this next-gen AMD APU. Considering the Renoir APU support is new to Linux 5.4, this late addition to DRM-Next of the display support will most likely be honored.
Besides the ability to drive displays attached to Renoir APUs, this AMDGPU DRM-Next update has fixes concerning GPU VM hardware bug workarounds, support for the SMU11 i2c controller, GPU reset vRAM bug fixes, and other fixes. There is also a number of Navi specific fixes concerning power-gating and other power management behavior.
This pull request outlines all of the fixes in full. The pull is some 91k lines of new code, but that's because the vast majority of that is auto-generated header files around the Renoir DC support with DCN 2.1.
| 21
| 1,760,719,550.066442
|
https://www.phoronix.com/news/AMD-Hiring-For-Coreboot
|
AMD Is Hiring For Coreboot Development, Sponsoring Open-Source Firmware Conference
|
Michael Larabel
|
Back on the AMD EPYC 7002 "Rome" launch day I wrote about how AMD is working to return to open-source BIOS / Coreboot support and now there's further confirmation of their work in that direction.
We were tipped off today that AMD's Head of Platform Firmware, Edward Benyukhis, publicly posted on LinkedIn that he is "looking to hire someone with solid Coreboot and UEFI background." If you have Coreboot experience or know someone who is, see LinkedIn for contacting Benyukhis.
That's exciting itself and certainly noteworthy, but also notable is AMD is now sponsoring next week's Open-Source Firmware Conference. AMD has joined the likes of Amazon AWS, Arm, System76, TrustedFirmware.org, and other companies in sponsoring this conference about Coreboot, LinuxBoot, and related open-source firmware projects.
Exciting times ahead, now let's just hope that the middleware/licensing issues noted in the earlier article are able to be addressed in a timely manner. Besides the prospects of open-source firmware/UEFI/BIOS and Coreboot support exciting many in the server space as part of secure computing initiatives (and desktop users at large, particularly Linux/FLOSS advocates), Coreboot is also a big deal on the Google Chromebook front where they make use of Coreboot and where previous AMD Coreboot efforts have focused on. This also comes at a time where other AMD code work is happening with likely Chromebook/Google connections making for some very interesting possibilities.
| 61
| 1,760,719,551.482961
|
https://www.phoronix.com/news/Ryzen-3900X-Power-Windows-Linux
|
AMD Ryzen 9 3900X Power Usage Is Running Measurably Higher On Linux Than Windows
|
Michael Larabel
|
Frequently brought up following our various Ryzen 3000 "Zen 2" benchmarks like the Ryzen 9 3900X vs. Core i9 9900K gaming benchmarks is how the Ryzen 9 3900X is pulling considerably more power than the similarly equipped Intel Core i9 system and those numbers are higher than what is often cited by Windows reviewers as the difference. I've begun investigating that power difference and indeed quite quickly could see Linux power usage being higher than Windows 10.
When comparing Ubuntu 19.04 to Windows 10 1903 with each operating system having all available updates and even using Linux 5.3 for the Ubuntu tests, the power usage on the same hardware was often measurably higher on Linux over Windows 10. Beyond being the same exact system, the UEFI/firmware settings were also untouched between the Windows/Linux testing and basically at their defaults. Each OS was at their defaults.
Same hardware, same firmware settings. Windows vs. Linux power data on Ryzen 9 3900X. More data and findings soon. pic.twitter.com/bIW5aJ6nmv— Phoronix (@phoronix) August 29, 2019
With some on Twitter initially claiming the difference being due to Windows 10's energy efficient default, I also did a secondary Windows 10 run when switching the power mode to its "best performance" mode too.
The AC system power consumption was monitored by an external system with the Ryzen 9 3900X system connected to a WattsUp Pro power meter and being polled by the Phoronix Test Suite.
First up is a look at the AC power consumption from a cold boot through loading the desktop and then idling for five minutes. During this idle period, Ubuntu 19.04 on this Ryzen 9 3900X system was pulling around 20 Watts more than Windows 10. Windows 10 when switched to its "best performance" power mode during testing only resulted in a Watt or two increase most of the time when not under load.
This is a look at the power usage under load when running the open-source Xonotic first person shooter game followed by Stockfish, Asmfish, encode-mp3, and Blender for a mix of single and multi-threaded workloads. We see that while running Xonotic (the first test), the Windows 10 power consumption was actually higher on Windows than Linux. But when it came to the multi-threaded workloads, Ubuntu was pulling considerably more -- generally 20~30 Watts more than Windows 10. The Windows 10 "best performance" power mode did yield a few Watt increase over the default behavior, but still nothing like was seen on Ubuntu Linux.
One of the possibilities that does come to mind is that there is the support on Windows 10 for the Collaborative Processor Performance Control (CPPC) support of Zen 2 but on Linux not at the moment. AMD did recently post patches for their new CPUFreq CPPC driver and I'll be testing those patches shortly for seeing if that helps lower the AMD Ryzen 9 3900X power usage on Linux. I'll also be doing some other kernel tests/comparisons as well.
So I'm still working to narrow down this significant power difference between the Ryzen 9 on Windows 10 and Linux with next up looking at CPUFreq CPPC, among other kernel options. Stay tuned for updates through the weekend. If you do enjoy my relentless Linux benchmarking, please show your support by at the very least not viewing this web-site with any ad-blocker otherwise joining Phoronix Premium for ad-free site access and multi-page articles on a single page or PayPal tips are also appreciated for helping with these time-consuming tests.
| 36
| 1,760,719,551.769626
|
https://www.phoronix.com/news/HDCP-1.4-AMDGPU-August
|
AMD Prepping Their HDCP 1.4 Content Protection Support For Raven Ridge & Newer
|
Michael Larabel
|
AMD developers have sent out their latest open-source Linux patches doing their kernel driver share for enabling High-bandwidth Digital Content Protection (HDCP) support for version 1.4 and newer.
While seeing HDCP support patches for open-source graphics drivers does irritate many in the community, similar to other open-source drivers supporting HDCP, this is only one part of the content protection puzzle. These patches alone do not impose any restrictions on users or other impairments, but mainly comes down to such proprietary software wanting to make use of HDCP capabilities on Linux. Open-source video players and the like can continue to enjoy GPU-based video acceleration uninterrupted.
So while there is a lot of opposition to Digital Rights Management and the likes of HDCP by the open-source community, having the support within the open-source graphics driver kernel code isn't limiting any current freedoms. But, on the other hand for the longevity of Linux driver support, this is actually good news. Presumably this recent HDCP work for the AMD Linux driver is due to more (or expected) design wins by the likes of Google with their Chromebooks.
Intel's open-source Linux graphics driver only began seeing HDCP work relatively recently when Google engineers were interested with the Intel support in the context of Chromebook support. Now with AMD getting their support in order, hopefully this means we'll be seeing more AMD SoCs within the likes of Chromebooks... Obviously AMD developers wouldn't be investing in this HDCP module for their AMDGPU Linux driver unless there was actual customer need: this latest HDCP addition is more than five thousand lines of new driver code.
For now this HDCP 1.4 implementation at least is targeting Raven Ridge APUs and newer. The patches are currently staged here and given their late timing for Linux 5.4 may very likely end up being material for Linux 5.5 -- as far as the AMDGPU DRM patches are concerned with some DRM-misc code already queued for the upcoming 5.4 cycle.
For free software purists, even though the default functionality shouldn't interrupt your viewing pleasures, the AMDGPU code does spin this addition of the HDCP module out as a new DRM_AMD_DC_HDCP Kconfig option to allow you to continue building a kernel without this functionality. Though based on the message of: "HDCP is not fully finished, so we need to be able to
build and run the driver without it." this Kconfig switch necessarily there for the long-term.
| 17
| 1,760,719,552.950468
|
https://www.phoronix.com/news/Linux-5.3-Disabled-Older-AMD-Rd
|
Linux 5.3 Moves Ahead With No Longer Advertising RdRand On Older AMD CPUs/APUs
|
Michael Larabel
|
Just prior to yesterday's Linux 5.3-rc6 kernel release, a change was pulled into the code-base that disables the advertising of RdRand support on older AMD CPUs/APUs.
The change pulled into the current Linux 5.3 tree (as opposed to being delayed until Linux 5.4) is over hiding RdRand support on AMD Bulldozer and Jaguar era processors. RdRand is the instruction for returning random numbers from the CPU for what began as an Intel extension. The RdRand instruction will continue to function on affected AMD processors if encountered but the CPU ID bit is being cleared so that it won't be advertised for software checking for the presence of this bit. This is being done since those pre-Zen AMD CPUs tend to have issues with their RdRand instruction support following suspend/resume cycles for many motherboards.
On resume, many BIOS implementations are not doing the proper steps for ensuring RdRand support and is evidently widespread enough that AMD felt it necessary to just disable the support by default. Unfortunately there is no way to properly query whether RdRand is behaving fine on AMD hardware and thus for these older systems is more sane to just disable by default. The issue itself has been known for 5+ years where RdRand may effectively just return -1 but is now being addressed by the Linux stable tree as a kernel workaround.
For systems known to be in a good state or not seeing suspend/resume activity, the rdrand=force kernel parameter will override this default change.
So as of yesterday, the change is mainlined for Linux 5.3 and could end up being back-ported as well to existing stable series.
| 10
| 1,760,719,554.371162
|
https://www.phoronix.com/news/AMDVLK-2019.Q3.5-Released
|
AMDVLK 2019.Q3.5 AMD Vulkan Driver Brings Fixes & Optimizations
|
Michael Larabel
|
Two weeks have passed since AMDVLK 2019.Q3.4 while out this morning is the next iteration of this open-source AMD Radeon Vulkan driver derived from the company's official cross-platform driver code-base.
AMDVLK 2019.Q3.5 is built against the Vulkan 1.1.119 headers and now exposes the second revision to the VK_EXT_subgroup_size_control extension.
On the gaming front this driver has continued tuning for Croteam's Talos Principle and dropping an earlier workaround for Total War: Three Kingdoms.
This AMDVLK driver also has improved code generation around FMin/FMax, a hang fix for an HDR test, a semaphore handle leak, and other code improvements.
The latest code (and Ubuntu x86_64 packages) for this official AMD Vulkan open-source driver over Mesa's RADV can be found via GitHub.
| 3
| 1,760,719,555.563075
|
https://www.phoronix.com/news/AMD-Renoir-APU-Gfx-Power-Man
|
AMD Renoir Graphics Power Management Gets Wired Up
|
Michael Larabel
|
While AMD's next-gen Renoir APUs are Vega-based and not Navi, beyond the initial Linux driver enablement seen over the past few weeks coming out a few days ago were a set of patches just getting the power management in order.
The graphics power management support for Renoir is where we're seeing a larger deviation in the driver code than the rest of the driver enablement that mostly pegs it as a Vega/GFX9 Raven Ridge refresh past Picasso. The power features for Renoir has come in at 37 patches amounting to around one thousand lines of new code, more than would be necessary for just a Raven Ridge / Picasso facsimile.
These patches make the many adjustments needed to PowerPlay, enable GFXOFF support for being able to turn off the graphics engine when not needed. updated golden settings, and clock gating support for the various blocks. The new power management code for Renoir can be found on the mailing list.
Given the punctual Linux driver support, hopefully any/all Renoir issues will be squared away prior to launch. From our recent testing of the Picasso Ryzen 5 3400G it's been working out well on Linux, much more so than the original Raven Ridge APUs.
| 0
| 1,760,719,555.841548
|
https://www.phoronix.com/news/AMD-EPYC-Rome-Linux-5.3-Perf
|
Linux 5.3 Kernel Yielding The Best Performance Yet For AMD EPYC "Rome" CPU Performance
|
Michael Larabel
|
Among many different Linux/open-source benchmarks being worked on for the AMD EPYC "Rome" processors now that our initial launch benchmarks are out of the way are Linux distribution comparisons, checking out the BSD compatibility, and more. Some tests I wrapped up this weekend were seeing how recent Linux kernel releases perform on the AMD EPYC 7742 64-core / 128-thread processors.
For some weekend analysis, here are benchmarks of Linux 4.18 through Linux 5.3 in its current development form. All tests were done on the same AMD EPYC 7742 2P server running Ubuntu 19.04 and using the latest kernels in each series via the Ubuntu Mainline Kernel PPA.
Via the Phoronix Test Suite, various benchmarks were run for looking at the performance between the tested kernels.
In most of the CPU bound workloads, there are only slightly variations in the performance, but at least Linux 5.3 is generally edging slightly higher than recent kernels.
Linux 4.18 largely seems to be the recommended minimum kernel version for these new AMD EPYC processors. On older kernels the system will boot and work, but there are various dmesg bits about CPPC not being supported and also x2apic support was added in 4.18. But as I generally recommend with any new hardware: the newer, the better both for features and the best performance.
The various synthetic kernel tests are generally showing the kernel moving in the right direction for performance.
In real-world workloads, there is generally slight performance advantages in many instances to being on Linux 5.2 or newer.
When cutting to the geometric mean for looking at the direction of the kernel performance across dozens of different benchmarks, Linux 5.3 delivered the best performance of the kernels tested on Ubuntu with this dual AMD EPYC 7742 server. I'll be running more benchmarks shortly as well as seeing if some of these Linux 5.3 wins are limited to the latest AMD processors or more widespread optimizations; see our Linux 5.3 feature overview to learn about the changes in this upcoming kernel release. Linux 5.3 should make it out as stable in September and be found by the likes of Ubuntu 19.10 and Fedora 31 while the Linux 5.4 cycle will kick off a new dance.
| 13
| 1,760,719,557.043846
|
https://www.phoronix.com/news/AMD-RdRand-Disable-15h-16h
|
AMD Bulldozer/Jaguar CPUs Will No Longer Advertise RdRand Support Under Linux
|
Michael Larabel
|
Not directly related to the recent AMD Zen 2 BIOS update needed to fix an RdRand problem (though somewhat related in that the original systemd bug report for faulty AMD RdRand stems from these earlier CPUs), but AMD has now decided to no longer advertise RdRand support for Family 15h (Bulldozer) and Family 16h (Jaguar) processors under Linux.
The RdRand instruction will still work on capable CPUs, but the CPU ID bit is being cleared so that it won't be advertised for software explicitly checking for the support. Tom Lendacky of AMD resorted to clearing the RDRAND CPU ID bit for 15h/16h processors (no impact for Zen, etc) due to RdRand issues cropping up after suspend/resume. Those issues have affected some users for a while and originate with the original AMD RdRand systemd bug report over problems following that cycle.
The buggy RdRand is being blamed on BIOS implementations not carrying out the proper steps for ensuring RdRand continues to function. But with apparently enough faulty BIOS out there, the RdRand bit will now be cleared for those CPUs to try to stop software from using it -- though any software still doing so, can though could experience the problematic events. The bug has been known for at least five years though only now being acted upon where RdRand could effectively be returning just -1.
If you don't plan to suspend/resume or your system/BIOS is known to be in a good state, there is a new rdrand_force kernel parameter being added to force-enable this support (a.k.a. maintain the status quo).
In response, at least one upstream developer is causing this a security vulnerability that up until now the RdRand could be spewing non-random data and an issue with AMD's RdRand implementation that it could be insecure if not properly programmed by the BIOS.
This change is currently pending via this patch that is likely to end up in the Linux 5.4 cycle though no word yet on the stable back-port outlook.
| 24
| 1,760,719,558.093128
|
https://www.phoronix.com/news/Ryzen-3000-BIOS-Update-Good
|
AMD Ryzen 3000 Series Playing Nicely With Latest Linux Distros Following BIOS Updates
|
Michael Larabel
|
One month ago we were told that AMD released a BIOS fix to their motherboard partners for addressing the systemd boot issue with Ryzen 3000 series processors that stems from an RdRand instruction issue. Finally over the past week we've seen motherboard vendors pushing out BIOS updates for the prominent motherboards and indeed this takes care of the issue.
With the likes of the ASUS ROG CROSSHAIR VIII HERO WiFi and many other motherboards, since the start of August the updated BIOS releases have begun appearing. In the case of at least ASUS, they are also explicitly mentioning the fix as "Supports Ubuntu 19.04 and other Linux distros."
MSI has also put out BIOS updates last week as a "beta" though without explicitly acknowledging the Linux fix.
I've tested the ASUS BIOS update and indeed it allows Ubuntu 19.04 and other newer Linux distributions to now boot gracefully on the new AMD Zen 2 desktops without any workarounds or patched systemd. So if you were holding off on upgrading to a new AMD Ryzen 3000 series system until the issue was resolved, the BIOS updates are trickling out now.
I'll also be around with some fresh AMD Ryzen 3000 series benchmarks including the 3600X and 3400G benchmarks in the days ahead now with tests done on top of Ubuntu 19.04.
| 24
| 1,760,719,558.600737
|
https://www.phoronix.com/news/AMD-Renoir-APU-Linux
|
AMD Sends Out Initial Linux Driver Support For "Renoir" APUs
|
Michael Larabel
|
AMD is striking well over the past month with their Linux hardware bring-up. In the past month we've seen the Navi 10/12/14 support get in order for Linux as well as support for the future Vega-based Arcturus GPU and now we see the initial enablement patches for their next-generation APUs, Renoir.
Sent out today was the initial 27 patches amount to around two thousand lines of code (roughly half of which is just header files) for bringing up this next-gen AMD APU.
Renoir is the successor to current AMD Picasso APUs and are said to be based on the new Zen 2 architecture while incorporating a Navi GPU... Well, at least originally what rumors/leaks reported. These code patches indicate Renoir to be based on Raven/Vega as opposed to Navi. Not a whole lot is known about Renoir publicly and are not expected to hit the shelves until 2020.
Renoir is based on Navi (or actually Vega?) so it isn't a whole lot of new code from the AMDGPU kernel driver side, but still, we aren't accustomed to seeing this Linux driver support so early in advance of launch (potentially ~6+ months). It could be a sign that there could be some interesting design wins or Chromebooks warranting good Linux support at launch or it could just be that they've hit a good punctual stride on new platform enablement, but whatever the case is it's certainly welcomed, especially after the very bumpy Raven Ridge experience on Linux.
These patches do confirm VCN 2.0 support on Renoir while most of he code changes come down to slight differences around power management / clock gating, golden register settings, and firmware bits. The AMDGPU code though does seem to indicate this as Vega-based rather than Navi as was rumored. So unless they end up marketing it weirdly (just like Kabylake-G with VegaM really being Polaris), that is what's happening.
For now at least Renoir's graphics is just represented by a single PCI ID, 0x1636.
This currently experimental (hidden behind feature flag) AMD Renoir support should be on its way to DRM-Next soon for premiering in Linux 5.4 while over the coming kernel cycles should stabilize and be flipped on by default, hopefully well ahead still of the hardware's expected launch.
| 38
| 1,760,719,559.360601
|
https://www.phoronix.com/news/AMD-EPYC-7742-2P-Summary
|
Summing Up The AMD EPYC 7742 2P Performance In One Graphic
|
Michael Larabel
|
If you didn't have a chance since last night to check out our benchmarks of the AMD EPYC 7742 and EPYC 7502 Linux performance, I certainly encourage you to do so. Even if you aren't a server enthusiast, it's incredible to see the engineering achievement of AMD with Zen 2 and how the race is certainly back on in the CPU space. If you are short on time, here's the quick summary of our initial AMD EPYC 7002 benchmark results.
As all independent reviewers seem to be in consensus, the EPYC 7002 series is a godsend for just about everyone, sans Intel. The EPYC 7002 line-up provides the most competitive option against Intel we've seen in the server space in many years if not ever, customers have these lower-cost CPUs that actually perform better and more power efficient in a majority of real-world workloads, and Rome further benefits consumers by reigniting Intel's engine and they will certainly need to respond in some manner by either (much) better pricing or some card up their sleeves.Better pricing seems about all Intel can do for the next quarter or two at least with Cooper Lake not coming out until next year and even then topping out at 56 cores while the next real battle could be when Icelake server parts finally ship. Intel though also has an upper-hand with a more mature and optimized software stack where they may squeeze more performance potential out of their existing chips as we've seen through their many upstream Linux contributions and the likes of Clear Linux, SVT, MKL, etc where single and even double digit percentage improvements in performance have routinely been achieved.
Anyhow, for those short on time or not yet digging through our pages of benchmark results, here is a summary of the performance between the EPYC 7742, EPYC 7601, and Xeon Platinum 8280 all in dual socket configurations and with each server running at their maximum supported memory channels and optimum frequencies. All servers freshly tested on Ubuntu 19.04 with Linux 5.2 as part of our Rome benchmarking.
As explained yesterday, the only areas where Cascadelake (green-ish line) is beating AMD's Rome is with their own MKL-DNN software (in some of the sub-tests at least), in MariaDB/MySQL strangely but that is being further investigated, and then when testing single-threaded workloads individually without fully taxing them in parallel as would be common in a production server. Cascadelake leading in the single-threaded tests isn't too surprising considering the higher clock speeds on the 8280 and then in some workloads AVX-512 does help out a lot. But as you can see, the EPYC 7742 (purple line) pretty much takes the cake. It's also striking to see how much faster Rome is than the EPYC 7601 2P.
With the dozens of tests run for this comparison, the EPYC 7742 2P has 22% better performance than the dual Platinum 8280 server based on the geometric mean or 56% compared to the 7601 2P.
And an extra graphic...
Here is also a radar chart for the performance-per-dollar graphics based on the CPUs current retail pricing. The chart here is just showing for the higher is better graphs (just a TODO item I need to do with the Phoronix Test Suite to make it show for all). But anyhow in looking at all those results, EPYC clearly wins on performance-per-dollar and even the previous-generation EPYC 7601 represents better value than the Xeon Platinum 8280 that command $10k USD a piece at retail levels.
Now see all of the EPYC 7502 and 7742 benchmarks if not already having done so while more benchmarks forthcoming. Currently wrapping up AOCC 2.0 compiler tests followed by taking Rome on a trip to FreeBSD/DragonFlyBSD, seeing how the performance compares on a few different Linux distributions, looking more at the modern CPU mitigation costs, and other benchmarking fun.
| 23
| 1,760,719,559.949046
|
https://www.phoronix.com/news/AMD-EPYC-Coreboot-Open-Austin
|
How Can AMD EPYC "Rome" 7002 Series Be Even Better? Open-Source BIOS / Coreboot
|
Michael Larabel
|
By now you've likely seen the fantastic performance out of AMD's new "Rome" 7002 series processors. The performance is phenomenal and generally blowing well past Intel's Xeon Cascade Lake processors. So that's all good and it can get even better outside of performance: I asked AMD about the prospects of Coreboot / open-source BIOS support and got a surprising response.
At an event in Austin last month when AMD was talking up Rome, when they weren't talking about the new server CPU's performance potential they were often mentioning the chip's security. That set the stage for bringing up open-source support and Coreboot support without coming across as just an open-source zealot/nerd question. After all, if their lower-level bits were (again) open-source it would ensure a more auditable boot process and ensuring the integrity of their Platform Security Processor (PSP) and the like, which in this day and age is important and trying to ensure no nefarious back-doors to the system. Companies like Facebook and Google are also genuinely interested in this open-source functionality with their work on the likes of Coreboot and LinuxBoot.
So I asked Forrest Norrod, SVP & GM of the Datacenter and Embedded Solutions Business Group, over the prospects of seeing such open-source support or Coreboot possibilities. Forrest though to my surprise said they are basically working on it. He didn't offer up any detailed explanation of expected deliverables or any timeline, but that they are aware of the situation and working on it.
Let's not forget - years ago they in fact provided the support. Back in 2011 was when they were talking of Coreboot support for all future CPUs (and did support some CPUs with Coreboot for a few years) and for a while were providing open-source AGESA releases but in the Zen era we haven't seen any of that from AMD. When they were under financial hardship, those ways changed.
Forrest pointed out that seemingly one hurdle they are working to overcome is the ability to publish the source code. He was quick to explain AMD has all of the source code and that they aren't bundling any arbitrary binary blobs for their processors/chipsets, but that they needed to obtain the rights to be able to publish some of the code. That appeared to be at least one of the hurdles slowing down such an open-source effort and something faced by other organizations as well when it comes to open-sourcing large and complex code-bases often with various sorts of middle-ware.
So it was a fairly broad response, but it does appear AMD is working on it and aware of their customer's needs. Especially now with EPYC 7002 series picking up more industry wins including from the likes of Google for their cloud, it does increase the chances of seeing more open-source code out of the company given Google's clout and work around LinuxBoot/Coreboot and wanting to ensure maximum security. If I find out anything more on AMD's plans, it will certainly be passed along when the time comes.
Worth pointing out that is somewhat related is that the Zen-based Hygon Dhyana is working on Coreboot enablement and AMD Picasso APU enablement for Coreboot has also been seeing code churn the past number of months. On the Picasso consumer front that is being driven by Google's requirements around Chromebook hardware. But aside from that, as it stands today, the AMD Coreboot support is mostly limited to generations-old hardware; years later Opteron boards are still popular with the pure free software crowd and even as recently as this year Opteron motherboards are getting blessed by the Free Software Foundation due to being open-source at the lowest levels.
Also, let's not forget Intel is also working to open-source their FSP code. That was a story Phoronix first reported last December. The last I heard from Raja when up in the mountains of Washington for OSTS 2019 was that it was still being worked on and hopefully we would hear more later in the year. So the outlook is good for returning to days of seeing better Coreboot support for modern (and powerful) server hardware platforms besides POWER9/Raptor though in the case of Intel could still be muddled by their Management Engine (ME) and other bits.Times are certainly interesting moving forward and it's never been so competitive in the server space / data center.
| 28
| 1,760,719,560.995655
|
https://www.phoronix.com/news/AMD-AOCC-2.0-Released
|
AMD Optimizing C/C++ Compiler 2.0 Released With Zen 2 Support
|
Michael Larabel
|
Coinciding with yesterday's glorious AMD EPYC "Rome" 7002 series CPU launch, AMD's software folks released AOCC 2.0 as their LLVM/Clang-based compiler optimized for Zen processors. AOCC 2.0 brings optimized compiler support now for Zen 2 processors not just only the EPYC 7002 line-up but also the Ryzen 3000 series consumer processors.
AMD Optimizing C/C++ Compiler 2.0 is the update adding in their Zen 2 "Znver2" bits and the first update to AOCC since the end of last year. Besides adding in Zen 2 support, they re-based their entire compiler toolchain against LLVM 8.0 and its sub-projects like the Clang 8.0 front-end. They also moved to using FLANG as their default Fortran language front-end.
Some of the other AOCC 2.0 happenings in general include more compiler optimizations, enhanced vectorization, support for LLD as the LLVM linker, optimized libraries including the AMD Math Library (AMDLibM 3.3), and other updates. AOCC 2.0 is a healthy update to AMD's optimized compiler stack though hopefully in the not too distant future they'll be able to contribute the relevant Zen 2 optimizations back into the upstream LLVM code-base.
AOCC 2.0 is available for download from developer.amd.com.
I should have out fresh EPYC 7742 2P benchmarks of GCC vs. LLVM Clang vs. AOCC 2.0 either later today or tomorrow, they've already been in the works from this incredible 128-core / 256-thread server setup on Ubuntu Linux.
| 4
| 1,760,719,561.416201
|
https://www.phoronix.com/news/AMD-Ryzen-9-3900X-SMT-Perf
|
AMD Ryzen 9 3900X SMT Linux Performance Benchmarks
|
Michael Larabel
|
For those wondering what the SMT performance impact is for new Zen 2 processors, here are some tests done using a Ryzen 9 3900X with Ubuntu Linux when testing at the default 12-core / 24-threads and then again when disabling SMT to look at just the twelve physical cores.
A few premium supporters wrote in recently wondering how the SMT performance looks on Linux for the new AMD CPUs and if it differs at all from Windows' SMT performance. Besides the recent Windows vs. Linux Ryzen 9 3900X benchmarks I haven't done a cross-OS SMT comparison yet, but here are some side-by-side tests looking at Ubuntu when toggling SMT from the BIOS.
Obviously toggling SMT was the only change made during this round of Linux benchmarking via the Phoronix Test Suite.
For a mix of different workloads from dozens of benchmarks, the geometric mean of all the results was boosted by just 4% with SMT.
When looking at some of the extremes, in many of the highly-threaded workloads there were significant gains to find out of the Ryzen 9 3900X with SMT. But in some of the multi-threaded tests when becoming over-subscribed the performance did suffer. Interestingly Intel's MKL-DNN math kernel library was one of the workloads suffering the most from SMT.
Those wanting to dig through all of the numbers can do so via this OpenBenchmarking.org result file.
| 14
| 1,760,719,562.466316
|
https://www.phoronix.com/news/AMD-Zen2-k10temp-Patches
|
Linux k10temp Support Pending For AMD Ryzen 3000 Series Processors
|
Michael Larabel
|
Patches are now queued into the Linux kernel's hwmon subsystem staging area that allow the AMD "k10temp" driver to support the Ryzen 3000 "Zen 2" processors.
Back during the Zen 2 press briefings I was told no temperature control offset was needed for Zen 2 and that the Linux thermal driver support was all squared away as a result. It was correct no Tctl is thankfully needed but the Linux driver support still wasn't exactly squared away. It turns out the new PCI IDs were missing from the Linux driver and thus no temperature monitoring support under Linux.
But now interested users have volleyed the patches and they are queued in the hardware monitoring tree. Sadly it's post Linux 5.3 merge window so at this stage it's in the Git tree for Linux 5.4, but we'll see if they end up getting sent in as "fixes" for the current Linux 5.3 cycle or will end up waiting until the Linux 5.4 merge window only at that point to then see it back-ported to the stable tree.
For those interested the support boils down to this one line patch and this four line patch for adding in the new IDs. That's all that is needed thanks to no longer needing an offset, but a pity it took until post-launch for the bits to be added to the kernel. Then again, with all the Zen processor launches to date it's taken until after launch for having the Linux temperature support.
On a related note, I have been lobbying AMD to either provide driver support or documentation for getting the (estimated?) CPU package power consumption reporting under Linux. During the Bulldozer days was the fam15h_power driver for reporting its CPU power consumption numbers under Linux though that driver doesn't work with Zen processors. Under Windows for Zen CPUs there is support for reporting CPU power consumption though how accurate it is a matter of some debate. Intel CPUs on Linux also expose their power consumption number too through RAPL and the like. I'm hopeful AMD will eventually provide this support on Linux for newer CPUs but still no definitive word yet.
| 19
| 1,760,719,563.268466
|
https://www.phoronix.com/news/Linux-5.2-5.3-Ryzen-9-3900X
|
Linux 5.2/5.3 Kernel Performance On The AMD Ryzen 9 3900X
|
Michael Larabel
|
With yesterday's Windows 10 vs. Ubuntu 18.04 LTS Linux benchmarks for the AMD Ryzen 9 3900X, some suggested that the Linux performance could have been better if using a Linux 5.x kernel. Well, here are some benchmarks comparing the performance of Ubuntu 18.04.2 LTS with its Linux 4.18 kernel compared to Linux 5.2 stable as well as the brand new Linux 5.3 development kernel.
On the same Ryzen 9 3900X system, these three kernel releases (Ubuntu 18.04's stock 4.18 kernel, Linux 5.2 stable, Linux 5.3 Git) were compared across a variety of workloads.
Sans graphics and I/O for any areas explicitly improved upon in recent kernel releases, the CPU performance itself went pretty much unchanged.
Obviously if you are using the open-source Intel/Radeon graphics in particular, I certainly recommend the newest kernel as possible for the updated drivers, but in terms of Zen 2 desktop CPU performance it was largely the same using the stock Ubuntu 18.04 LTS kernel.
Those wanting to look through the dozens of benchmarks individually, they are available via 1907278-HV-3900XLINU13 on OpenBenchmarking.org.
Looking ahead to Linux 5.4 late in 2019, that's where there is the possibility of Zen 2 performance changes due to AMD's new CPPC CPUFreq driver as one performance-sensitive change pending, but we'll see what else comes about and other optimizations now that more Linux developers are getting their hands on these impressive processors.
| 12
| 1,760,719,563.995309
|
https://www.phoronix.com/news/GCC-10-Zen2-Scheduler-Model
|
GCC 10 Compiler Picks Up New Scheduler Model & Cost Tables For AMD Zen 2 Processors
|
Michael Larabel
|
While AMD developers published their "Znver2" compiler patches for Zen 2 originally back in November, months ahead of the recent Ryzen 3000 series launch, this compiler support was incomplete as it re-used the existing scheduler model and costs table of Znver1. Now though one of SUSE's compiler experts who often works in cooperation with AMD has published the new Znver2 scheduler model and costs table for Zen 2.
The updated costs table better reflects the "costs" of moving and loading various registers and different instructions compared to Znver1 so the compiler can make wiser decisions for the most efficient usage. With these updated costs to reflect faster multiplication and 256 vector paths, there is better GNU C Library performance in particular and SUSE developer Jan Hubicka noted that the memory copy performance "wins" even for small blocks.
The Zen 2 scheduler model is largely based on the Zen 1 model is refined to better match AMD's new microarchitecture. Hubicka noted though there is little effect on overall performance but I'll certainly be firing off my complete benchmark set shortly.
The initial "Znver2" support made it into GCC 9 as released earlier this year and on the LLVM/Clang side into version 9.0. That timing alone is a bit less than ideal considering LLVM 9.0 isn't being released until next month and the GCC 9.1 compiler shipped in May and thus not yet found in the likes of the latest Ubuntu stable releases, so those early customers of AMD Zen 2 processors are missing out on this optimized support unless using a rolling-release Linux distribution or rolling their own compiler. But beyond that, it's been missing out on the reworked scheduler model and updated costs table for the Zen 2 microarchitecture. Now in turn, while that hit the GCC code-base today it won't be found in a released version of the GNU Compiler Collection until GCC 10.1 releasing in April~May of 2020 or LLVM 10.0 on the Clang side in February~March.
Getting out the updated compiler support well ahead of launch has been an area where Intel has been very punctual over the years in realizing the (performance) benefits to getting their compiler support available and tuned so that by the time their processors are shipping the supported compilers are found throughout the Linux ecosystem for ensuring optimal performance and also seeing all new instruction set extensions are available. Hopefully with time AMD will be able to improve this process and provide more timely compiler support; it's certainly something I've been bringing up with AMD several times this year. Even with today's scheduler model and cost table work, it still appears GCC and Clang lack support for some Zen 2 extensions like RDPRU.
I'll be firing up some fresh Znver2 compiler benchmarks on the Ryzen 9 3900X shortly.
| 3
| 1,760,719,564.8212
|
https://www.phoronix.com/news/ASUS-CROSSHAIR-VIII-HERO
|
ASUS ROG CROSSHAIR VIII HERO Testing On Ubuntu 18.04 Linux
|
Michael Larabel
|
For those in the market for an AMD X570 high-end motherboard for use with the new Zen 2 processors, the ASUS ROG CROSSHAIR VIII HERO was one of the boards sent out as part of the reviewer's kit and it's been working out quite well.
The ASUS ROG CROSSHAIR VIII HERO is quite feature packed with dual M.2 drives, USB 3.2 Gen2, active chipset heatsink, 2.5 Gbps Ethernet and Gigabit Ethernet (requires kernel patches for the 2.5G controller), and plenty of other connectivity. This motherboard does cost a pretty penny though at around $380 USD.
The ASUS ROG CROSSHAIR VIII HERO has been working out well in our tests sans the Linux systemd/RdRand issue that will soon see a new BIOS release with the fix in place.
This motherboard has also been working out well in the testing we've done so far with FreeBSD 12.
As some reviewers have found some performance differences between the pre-release/reviewer BIOS and the latest production BIOS now available for the Zen 2 launch, I did run some Linux tests. At least in the case of this ASUS motherboard, there hasn't been much of a performance difference between the two BIOS revisions:
It's usually a few percent swing if that in either direction, but the older BIOS was favored in a number of cases.
This may be due to Linux differences around the CPU frequency scaling driver and other kernel factors compared to Windows, but I am currently carrying out more BIOS tests on different AMD X570 motherboards. This Linux BIOS comparison testing from the VIII HERO in full can be found via this OpenBenchmarking.org result file.
There's another week of interesting Zen 2 benchmarks coming up on Phoronix.
| 25
| 1,760,719,565.344221
|
https://www.phoronix.com/news/AMD-Picasso-Coreboot-Ready
|
AMD Picasso Support For Coreboot Appears Finally Ready
|
Michael Larabel
|
Back in April I wrote about Coreboot seeing AMD Picasso APU enablement work as the first Zen/Ryzen processor target being handled by this open-source BIOS alternative. It now looks like that Picasso support is all squared away and ready for use by future AMD-powered Google Chromebooks.
Following various other merge requests landing with this Picasso enablement, Picasso is now added to the Kconfig build system configuration for being able to flip the support on. In other words, the rest of the support has landed so now it's ready for allowing users to toggle the support for building it. This Picasso work continues to be largely done by Google engineers in the context of Chromebook support.
Hopefully we'll see more AMD Coreboot support moving forward.
| 7
| 1,760,719,566.577014
|
https://www.phoronix.com/news/AMD-Releases-Linux-Zen2-Fix
|
AMD Releases BIOS Fix To Motherboard Partners For Booting Newer Linux Distributions
|
Michael Larabel
|
AMD has just alerted us that they have released a BIOS fix to their motherboard partners that takes care of the issue around booting newer Linux distributions on the new Zen 2 processors.
Earlier this week I mentioned AMD would be working on a BIOS fix to address the fundamental problem with booting newer systemd-using Linux distributions on their new Ryzen 3000 series processors. However, I hadn't expected the fix to make it to motherboard vendors in less than one week!
RdRand issue looks like will be fixable by a BIOS update.— Michael Larabel (@michaellarabel) July 8, 2019
The problem is the RdRand issue colliding with systemd that is making use of the RdRand instruction directly and not jiving with the expected behavior. There's been a patch in systemd since May but that hasn't been found in a released version yet. But for newer Linux distributions like Ubuntu 19.04, Clear Linux, Fedora Workstation 31, Arch Linux / Manjaro, and others, it's meant not being able to boot the distribution due to all systemd services failing to start.
See my Ryzen 7 3700X / Ryzen 9 3900X Linux review from launch day for more details. These CPUs have been running great on Linux besides this embarrassing fault that made it in due to AMD focusing their tests on Ubuntu 18.04 LTS and not vetting Ubuntu 19.04 or others pre-launch.
I just received the following official statement from AMD:
AMD has identified the root cause and implemented a BIOS fix for an issue impacting the ability to run certain Linux distributions and Destiny 2 on Ryzen 3000 processors. We have distributed an updated BIOS to our motherboard partners, and we expect consumers to have access to the new BIOS over the coming days.
Hopefully it won't take too long for motherboard vendors to release new BIOS updates. But considering it also has a fix for the Desinty 2 Windows game rather than "just a Linux fix" plus with the rush of other motherboard BIOS updates now that more consumers are getting their hands on these exciting new processors, hopefully most motherboards will be patched in the very near future. Stay tuned and I also have more AMD Zen 2 Linux benchmarks on the way.
| 95
| 1,760,719,567.08001
|
https://www.phoronix.com/news/AMD-New-CPUFreq-CPPC
|
AMD Posts New CPUFreq Driver For CPPC Support With Zen 2 CPUs
|
Michael Larabel
|
AMD Zen 2 CPUs support ACPI's Collaborative Processor Performance Control (CPPC) for tuning the system to energy and/or performance requirements. AMD has now published a new CPUfreq driver for handling their CPPC implementation and the new controls found with their new processors.
The AMD CPPC support with Zen 2 desktop/server/mobile CPUs can be optionally enabled and allows setting min/maximum performance along with desired performance and other knobs for tuning via sysfs.
This new AMD CPU frequency scaling driver is called "amd_cpufreq" and can be flipped on under supported kernels with the amd_cpufreq.cppc_enable=1 kernel flag or amd_cpufreq=enable. But given the timing of these new patches being posted, sadly it's unlikely to see this new driver queued for the currently open Linux 5.3 merge window and thus will have to wait for Linux 5.4 later in the year.
AMD will be documenting different performance/power options within performance/optimization guides. They also plan to develop a Linux user-space tool for generating CPPC profiles for a target workload. Likewise, there is no generic/default CPPC policy at this time.
The ACPI CPPC specification allows for the operating system to better manage the power/performance of the processor based on an abstract performance scale. This amd_cpufreq driver exposes more knobs than what is supported by the existing Linux kernel's acpi_cppc driver. That existing kernel CPPC code has largely been focused on Arm's support for this functionality while it's great with Zen 2 there is now the support on the AMD front.
Once this code is mainlined it should provide for interesting tests. The initial yet-to-be-reviewed patches can be found on the LKML.
| 16
| 1,760,719,568.145052
|
https://www.phoronix.com/news/Ryzen-3K-RdRand-Systemd-Maybe
|
The Ryzen 3000 Boot Problem With Newer Linux Distros Might Be Due To RdRand Issue
|
Michael Larabel
|
As outlined yesterday, AMD's Ryzen 3000 processors are very fast but having issues booting newer Linux distributions. The exact issue causing that boot issue on 2019 Linux distribution releases doesn't appear to be firmly resolved yet but some are believing it is an RdRand instruction issue on these newer processors manifested by systemd.
For those that missed my launch day article, check it out for more details and then all the benchmarks for when running very performant on the likes of Ubuntu 18.04 LTS where the issue does not occur. What people are jumping on today are the old reports of AMD RdRand problems for using this hardware RNG instruction causing issues on older pre-Zen2 processors. It is possible though the RdRand support regressed even further and thus in worse shape with Zen 2, but I haven't seen that officially acknowledged by AMD nor if it firmly addresses the issue.
It does make some sense though as outlined yesterday newer Linux 5.0~5.2 kernels do boot on older Ubuntu 18.04 LTS era distributions, which are on older systemd releases. There is a change in the latest systemd that does help that was merged in May to eat bad RdRand values on AMD CPUs.
I was aware of the AMD RdRand issue on older CPUs and among my attempted debugging of the problem pre-launch was trying to blacklist the x86_64 crypto modules, AESNI_Intel, and related bits to no avail. But it's possible that this change made by systemd in May could workaround the issue. There hasn't been a new systemd release since April which would explain why rolling-release distributions like Arch Linux and Clear Linux still aren't able to boot these newest AMD processors.
Anyhow, right now it's looking like borked RdRand support is a feasible explanation if indeed accurate. There is the systemd workaround in systemd Git but hopefully it's something that can be addressed via a firmware/microcode update as otherwise no modern Linux distributions will be booting properly for these new AMD processors until the next rounds of releases -- it's unlikely we'd see say a Fedora or Ubuntu re-spin simply with the new systemd build integrated. But at least it it can be worked around in the microcode/firmware would allow existing Linux distributions to begin playing nicely with these shiny and powerful new AMD boxes.
More details when I have them; I'm away the next two days but will begin resuming more tests ASAP to try to confirm the RdRand behavior.
| 121
| 1,760,719,568.783316
|
https://www.phoronix.com/news/AMD-Navi-RADV-Mesa-19.2
|
RADV Vulkan Driver Manages Launch-Day Support For AMD Navi 10/12/14 GPUs
|
Michael Larabel
|
Leading up to today's Radeon RX 5700 "Navi" series launch it was looking like there wouldn't be any support within Mesa's Radeon "RADV" Vulkan driver for this community-maintained open-source implementation. But the open-source developers at Valve managed to not only deliver Navi 10 support but also Navi 12 and Navi 14 are also supported with this new Mesa 19.2 code.
Various folks at AMD didn't believe the "community" RADV developers at the likes of Valve and Red Hat were provided with samples or documentation in advance, but however it turned out, Valve developer Samuel Pitoiset along with Google developer Bas Nieuwenhuizen managed to land the Navi/GFX10 enablement code just minutes ago into Mesa 19.2 Git.
The code is just about one thousand lines on top of the existing RADV GFX9 support. Interestingly, some 600 lines of that code is just for bringing up the NGG (Next Generation Geometry) support for vertex shaders. So overall at least from the Vulkan driver side it wasn't much new code for driving Navi, granted, it also leverages the Navi support within the AMDGPU kernel driver and LLVM shader compiler.
It's not clear at this stage how well optimized RADV is for the Radeon RX 5700 and RX 5700 XT if it will be like RadeonSI where a lot of additional optimizations are pending. At least though this means Mesa 19.2 due out around the end of August / early September will have both open-source OpenGL and Vulkan support for these new GPUs.
Also somewhat entertaining is that this RADV Navi support has beat out AMD's official Vulkan Linux driver "AMDVLK" for Navi support as that is still pending and expected in the days ahead once finishing up its internal/legal reviews.
I'll have up some new RadeonSI+RADV benchmarks on the RX 5700 and RX 5700XT to allow for any imminent optimizations/fixes to land.
| 20
| 1,760,719,569.66154
|
https://www.phoronix.com/news/AMD-SEV-Vulnerability-Micro-Up
|
AMD Releases Firmware Update To Address SEV Vulnerability
|
Michael Larabel
|
A new security vulnerability has been made public over AMD's Secure Encrypted Virtualization (SEV) having insecure cryptographic implementations. Fortunately, this AMD SEV issue is addressed by a firmware update.
CVE-2019-9836 has been made public as the AMD Secure Processor / Secure Encrypted Virtualization exposes an insecure cryptographic implementation.
AMD has responded by providing an updated firmware that is said to address these issues. The comment we heard from an AMD representative is:
At AMD, security remains a top priority and we continue to work to identify any potential risks for our customers. Through ongoing collaboration with industry researchers AMD became aware that, if using the user-selectable AMD secure encryption feature on a virtual machine running the Linux operating system, an encryption key could be compromised by manipulating the encryption technology’s behavior. AMD released firmware-based cryptography updates to our ecosystem partners and on the AMD website to remediate this risk.
That's all we know for now, but will update on hearing anything else.
| 15
| 1,760,719,570.078965
|
https://www.phoronix.com/news/AMD-Neil-LLVM-No-More
|
One Of AMD's Leading LLVM Compiler Experts Jumped Ship To Unity
|
Michael Larabel
|
AMD has lost one of their leading LLVM compiler developers as well as serving as a Vulkan/SPIR-V expert with being involved in those Khronos specifications.
Neil Henning has parted ways with AMD and is now joining Unity Technologies. Neil was brought to AMD to improve the performance of their LLVM compiler, in particular their LLVM Pipeline Compiler (LLPC) used by the likes of their official AMD Vulkan driver in order to make it competitive with their long-standing, proprietary shader compiler currently used by their binary graphics drivers. While at AMD, he was able to increase the performance of their LLVM shader compiler stack by about 2x over the past year. He also implemented various Vulkan driver extensions into their stack.
AMD's LLVM compiler stack continues getting more mature and comparable in performance/features to their shader compiler but we haven't heard anything new recently about when they may migrate their proprietary OpenGL/Vulkan components to using it.
But he acknowledges he didn't enjoy internal politics, the internal systems at the company, and other factors. As a result, Neil is now joining Unity to work on their Burst compiler. Their Burst compiler effort is what they are doing and based on LLVM with a goal of making C# faster than C++. With Neil onboard, the Burst prospects have become a lot more exciting.
As part of this move, Neil is also parting ways with The Khronos Group and his work on Vulkan/SPIR-V. More details via this blog post.
| 44
| 1,760,719,571.02172
|
https://www.phoronix.com/news/VKHR-Vulkan-Hair-Rendering
|
VKHR - An AMD-Backed Open-Source Hair Renderer In Vulkan
|
Michael Larabel
|
VKHR is an open-source, real-time hybrid hair renderer written in Vulkan and developed under the support of AMD/RTG.
AMD previously worked on some great hair rendering tech with TressFX but now it's being taken to a whole new level with VKHR. VKHR is being led by Erik Jansson of AMD as a real-time hybrid hair renderer "written 100% from scratch in Vulkan" and using C++17 code. VKHR has a built-in ray-tracer based on Intel's Embree technology. And there's even a built-in benchmark for comparing the project's hair rendering performance.
Those wanting to learn more about all the technical details can do so via the project's GitHub while for those wondering the end result, here are some samples:
In addition to Erik Jansson, others from AMD/RTG are also involved as well as university researchers. A paper on their research can be viewed here.
| 10
| 1,760,719,571.485153
|
https://www.phoronix.com/news/AMD-Zen-2-New-Instructions
|
AMD Zen 2 CPUs Come With A Few New Instructions - At Least WBNOINVD, CLWB, RDPID
|
Michael Larabel
|
During the AMD Zen 2 + RDNA launch event they highlighted some of the new instructions to find with the Zen 2 processor but there is at least one more.
During the Zen 2 briefings they covered CLWB for Cache Line Write Back and WBNOINVD for Write Back and Do Not Invalidate Cache. CLWB writes back to memory the cache line specified. WBNOINVD is similar to WBINVD but handles the write back and does not invalidate the cache. WBNOINVD will write the modified cache lines from the CPU cache to system RAM without flushing (invalidating) the internal caches.
Those were just the cache instructions talked about during the LA event. But in recalling the AMD znver2 compiler patches to GCC that were published at the end of last year, there was also the RDPID instruction. RDPID is the instruction to read the processor ID in a faster manner than the likes of RDTSCP that also includes the time-stamp counter.
Curious, I asked around and found out that RDPID is in fact present for Zen 2 processors. That's good news and originally Intel plumbed the RDPID support for Cannonlake/Icelake, to which they got the support ready but for lack of Cannonlake shipping at scale and not yet Icelake, AMD is able to take advantage of that infrastructure work.
One of the areas the Linux kernel is already plumbed to make use of RDPID when present is within the __getcpu code. Given the getcpu system call for determining the CPU node/core where a thread has been active in terms of placement, this could have some interesting implications. At the very least, Intel developers who only support this instruction with Icelake have said this instruction to be much faster than the existing instructions for the task. When I have my hands on the new Ryzen 3000 processors (and review embargo expired), I'll certainly run some benchmarks to see if it makes any meaningful difference as at least Intel's RDPID support was mentioned in a code comment of being much snappier.
When finding out if RDPID was indeed in Zen 2, I also heard that there are likely "some" other new instructions as well with Zen 2 processors but couldn't be immediately recalled by one of the Zen architects. That's exciting to hear but wasn't covered at the AMD event as they may be waiting for the Hot Chips conference to talk about them more, but I'll hopefully be getting a list soon of these new instructions, and hearing they didn't think the "press" would be interested in ISA details at this AMD event plus obviously limited time.
The only downside of hearing there are more instructions at stake is that sadly the GCC and LLVM Clang compilers with their initial "znver2" support only expose the cache instructions and RDPID as part of this Zen 2 target. Thus having to add more instructions to be exposed by znver2 will just delay the time by which developers / code builders will find these new instructions flipped on when building with -march=znver2. That's one of the areas Intel has been good at meanwhile in generally providing new ISA support and compiler targets for forthcoming CPUs generally up to years in advance due to the annual release cadence of the GCC and six month releases for LLVM/Clang. So for any Znver2 compiler improvements we'll likely not see them until GCC 10 next year or LLVM Clang 9 in September.
I also raised the issue of the znver1 scheduler / cost tables sometimes not being optimal and the znver2 compiler support mostly reusing the znver1 code without being tuned yet and how that's unfortunate given the slow cadences for getting this support into new compilers. Hopefully by the time of Zen 3's launch we'll see good znver3 support in released compilers next year with proper tuning and all new instruction set extensions.
Anyhow, after 7 July I'll certainly be running some Zen 2 compiler benchmarks with GCC and Clang on Phoronix along with the AMD Optimizing C/C++ compiler whenever AMD updates that LLVM/Clang-forked compiler for Znver2.
| 24
| 1,760,719,572.585683
|
https://www.phoronix.com/news/AMD-Samsung-RDNA-License
|
AMD Licensing RDNA Graphics IP To Samsung For Smartphones & More
|
Michael Larabel
|
AMD today announced a new licensing deal with Samsung around low-power, high-performance graphics technologies.
AMD is licensing to Samsung "custom graphics IP based on the recently announced, highly-scalable RDNA graphics architecture to Samsung for use in mobile devices, including smartphones, and other products that complement AMD product offerings."
Basic details were shared today via this press release.
Given Samsung's Linux/Android usage, it will be interesting to see the driver support evolve likely based off AMD's existing open-source graphics driver stack and with time should ultimately provide a look at how close their RDNA graphics will be to the Radeon hardware itself, assuming Samsung doesn't close up their Linux driver bits.
| 42
| 1,760,719,572.929495
|
https://www.phoronix.com/news/ROCm-RET-Enablement-Tool
|
The ROCm Enablement Tool Makes It Easier To Setup AMD's Open-Source Compute Stack
|
Michael Larabel
|
While there are the Debian/RPM packages offered of the Radeon Open eCosystem (ROCm) stack for Linux users on supported distributions, the new "ROCm Enablement Tool" could assist in setting up this GPU compute stack on supported Linux distributions and elsewhere.
The ROCm Enablement Tool, or RET for short, is a currently experimental tool for setting up the ROCm driver stack as well as associated software like TensorFlow.
The ROCm Enablement Tool ideally wants to make it easy from setting up the drivers to dependencies and various machine learning frameworks and benchmarks. At this point though the number of machine learning framework benchmarks is fairly limited, but hopefully AMD developers will continue improving the utility with time.
Those wanting to check out this early stage / experimental RET component can find it on GitHub.
| 19
| 1,760,719,574.080442
|
https://www.phoronix.com/news/AMD-Ryzen-3000-RX-5700
|
AMD Announces Ryzen 3000 Series, Radeon RX 5700
|
Michael Larabel
|
At Computex today AMD announced their initial 3rd Gen AMD Ryzen desktop processor line-up launching on 7 July along with the Radeon RX 5700 series.
The 3rd Gen Ryzen line-up based on their Zen 2 core announced today starts with the Ryzen 5 3600 at 6 cores / 12 threads and 65 Watt TDP with 4.2GHz boost frequency and 35MB cache for $199 USD. At the top end is the Ryzen 9 3900X at 12 cores / 24 threads with 105 Watt TDP, 3.8GHz base frequency / 4.6GHz boost, 70MB cache, and $499 USD retail price. A step below the Ryzen 9 is the Ryzen 7 3800X with 8 cores / 16 threads, 4.5GHz turbo / 3.9GHz base, 36MB cache, and $399 retail price. There is also a Ryzen 5 3600X at $249 and Ryzen 7 3700X at $329. All of these new Zen 2 CPUs support 40 PCI Express 4.0 lanes. These CPUs are shipping on 7 July.
The Radeon RX 5700 series meanwhile is the first of their new "RDNA" gaming architecture and up to 1.25x higher performance-per-clock and 1.5x performance-per-Watt over GCN. The Radeon RX 5700 series will feature GDDR6 memory and PCI Express 4.0 while offering performance in line with the GeForce RTX 2070.
More details at AMD.com. We'll have launch-day benchmarks in July.
| 68
| 1,760,719,574.229483
|
https://www.phoronix.com/news/AMD-Raven1-Skip-The-DMCU
|
AMD Staging Another Fix To Try Correcting Some Raven Ridge Systems On Linux
|
Michael Larabel
|
AMD Raven Ridge APUs have been out for more than one year now and at least under Linux can still be quite problematic depending upon the particular motherboard BIOS and other factors. Fortunately, while Raven 2 and Picasso APU support is appearing to be in better shape, the AMD open-source developers haven't forgot about these problematic Raven 1 systems.
Out today is the latest patch trying to help those with original Raven Ridge systems. This latest hopeful fix is now skipping over loading the DMCU firmware for Raven Ridge. DMCU in this context is the Display Micro-Controller Unit and is the micro-controller used for Panel Self Refresh (PSR) and similar functionality.
The AMDGPU Linux DRM driver has supported loading the DMCU firmware for a while now but the developers seem to believe that skipping over loading it on these Raven systems is better off rather than trying to load it and ignoring any errors. In particular, when loading the DMCU firmware on some systems with a "bad" system BIOS, it can still lead to users unable to boot their systems.
The patch is out there for review and will see if it ends up in the Linux stable series or gets tossed around some more before merging. In response to the patch, there has been a request already to white/black-list the functionality or at least offer the DMCU firmware loading as a module parameter. Without the ability to load the DMCU firmware, one missing feature is Automatic Backlight Management (ABM) for Raven Ridge laptops.
| 37
| 1,760,719,575.564319
|
https://www.phoronix.com/news/AMD-GCN-GCC10-Constructors
|
AMD GCN GPU Target Continuing To Improve For The GCC 10 Compiler
|
Michael Larabel
|
With the recent release of the GCC 9 stable compiler there is the initial "AMD GCN" GPU target/back-end merged. However, for this GNU Compiler Collection release the AMD GCN target isn't all that useful but continued work on it gives us hope of seeing it in good shape for next year's GCC 10 release.
With the GCC 9.1 release, the AMD GCN back-end can only handle running basic single-threaded programs... Not exactly useful for graphics cards. The GCC 9 code supports targeting the Fiji and Vega 10 GCN instruction set architecture.
Still being worked on for mainline GCC with the AMD GCN GPU targeting is the ability to handle OpenACC/OpenMP GPU device offloading with this back-end to allow for real-world code-bases to run on the Radeon graphics cards. We're still not there yet with the GCC development code, but work is progressing in that direction.
Besides other AMD GCN activity, the most recent addition is now handling constructors/deconstructors with GCN. Not all that useful by itself, but we're certainly excited every time more AMD GCN support hits GCC.
This code continues to be worked on by CodeSourcery / Mentor Graphics under contract for AMD. Within Mentor Graphics' CodeBench Lite AMD software is the bleeding-edge GCC back-end work that does allow for GCN offloading to Radeon Instinct GPUs with OpenMP and OpenACC. Hopefully mainline GCC is caught up by the next major release, GCC 10, albeit unfortunate it has taken so long for the GNU compiler to see this level of AMD GPU support.
| 0
| 1,760,719,575.675796
|
https://www.phoronix.com/news/Linux-5.2-AMD-Zen-P2P-DMA
|
Linux 5.2 To Allow P2P DMA Between Any Devices On AMD Zen Systems
|
Michael Larabel
|
With the Linux 5.2 kernel an AMD-supplied change by AMDGPU developer Christian König allows for supporting peer-to-peer DMA between any devices on AMD Zen systems.
The patch by Christian allows for peer-to-peer DMA to happen on modern AMD systems even between different root ports, expanding the current assumptions by the kernel. The patch explains:
The PCI specs say that peer-to-peer DMA should be supported between any two devices that have a common upstream PCI-to-PCI bridge. But devices under different Root Ports don't share a common upstream bridge, and PCIe r4.0, sec 1.3.1, says routing peer-to-peer transactions between Root Ports in the same Root Complex is optional.
Many Root Complexes, including AMD ZEN, *do* support peer-to-peer DMA even between Root Ports. Add a whitelist and allow peer-to-peer DMA if both participants are attached to a Root Complex known to support it. The change already landed in Linux Git as part of the PCI updates for Linux 5.2.
| 8
| 1,760,719,577.016546
|
https://www.phoronix.com/news/Threadripper-GCC9-Optimized-Run
|
AMD Ryzen Threadripper Compiler Tuning/Optimization Benchmarks With GCC 9, PGO
|
Michael Larabel
|
For those interested in compiler optimization/tuning with AMD Ryzen Threadripper hardware, here are some follow-up benchmarks to Tuesday's GCC 9 vs. Clang 8 C/C++ Compiler Performance On AMD Threadripper, Intel Core i9.
The tests today are of GCC 9 at different tuning/optimization levels for reference purposes. Additionally there is a run when using Profile Guided Optimizations (PGO) for looking at the performance impact on GCC 9.
To cut to the chase here are the levels tested and jumping straight to the geometric mean of all the test results:
For all the individual metrics across dozens of benchmarks, jump to this OpenBenchmarking.org result file.
As for the PGO impact, it's quite light. That's using the Phoronix Test Suite's PGO testing module. All those details for those interested in compiler tuning via this OpenBenchmarking.org result file.
| 3
| 1,760,719,577.107455
|
https://www.phoronix.com/news/AMD-CPUs-RdRand-Suspend
|
Some AMD CPUs Might Lose RdRand Randomness Following Suspend/Resume
|
Michael Larabel
|
Systemd developers are sounding the alarms that some AMD processors might lose randomness (yielding non-random data) via the RdRand instruction following a suspend/resume alarm. However, initial indications don't appear for this to be some glaring widespread issue and might be limited to the older AMD CPUs and/or BIOS/motherboard combination.
Systemd lead developer Lennart Poettering of Red Hat tweeted today, "So AMD CPUs implement an RDRAND operation that doesn't actually return randomness (after your first suspend/resume cycle that is)."
So AMD CPUs implement an RDRAND operation that doesn't actually return randomness (after your first suspend/resume cycle that is). https://t.co/FNsqJb4TEu— Lennart Poettering (@pid_eins) May 7, 2019
Referenced is this bug report outlining that the RdRand instruction on an older AMD A6-6310 processor isn't properly behaving following a suspend/resume. By avoiding RdRand usage on the system as part of generating a UUID, the reported systemd issue no longer happens.
Also referenced is this Linux kernel bug report that is still open after five years. That bug report cites RdRand failing after resume on AMD CPUs. In this case, OpenSSL was failing to generate keys after a kernel suspend/resume. The belief back then was that it may be due to a BIOS bug but the issue not fully investigated since OpenSSL ended up disabling RdRand usage in the process and thus working around the problem experienced by the end-user.
Fortunately, by default the Linux kernel doesn't exclusively rely upon RdRand as a source of entropy but ends up being mixed in with other data / entropy sources. Regardless, now that Red Hat developers are involved and other upstream developers, hopefully they'll be able to figure out this issue in short order to come up with an effective solution.
I haven't encountered this issue myself and so far the only reports I've seen are from those using older Excavator era processors and not any newer AMD "Zen" Ryzen/EPYC processors.
| 22
| 1,760,719,578.536598
|
https://www.phoronix.com/news/AMD-Frontier-Supercomputer
|
AMD EPYC + Radeon Instinct To Power "Frontier" 1.5 Exaflop Supercomputer
|
Michael Larabel
|
AMD and Cray announced today they won a contract to provide the US Department of Energy with the hardware to the "Frontier" supercomputer that is expected to go online in 2021 and deliver 1.5 exaflops of compute power.
The US DoE Frontier supercomputer is the fastest planned supercomputer right now and will be built around custom EPYC CPUs and Radeon Instinct graphics processors. There will be four Radeon GPUs per EPYC CPU and feature a custom Infinity Fabric implementation and this supercomputer is built around Cray's Shasta architecture.
Making this Frontier supercomputer win even more exciting is that the CPU/GPU programming will be built around the Radeon Open eCosystem "ROCm" stack, marking a big win for open-source/Linux. It will be interesting to see ROCm on super-computers and a pleasant change of pace compared to the proprietary CUDA/compute used by today's NVIDIA-based supercomputers.
More details on the Frontier supercomputer at AMD.com.
| 0
| 1,760,719,579.931653
|
https://www.phoronix.com/news/Linux-5.2-AMD-Zen-2-EPYC-EDAC
|
Next-Gen AMD EPYC Changes To EDAC Driver Sent In For Linux 5.2 Kernel
|
Michael Larabel
|
The notable change with the "EDAC" changes for Linux 5.2 comes down to the "Zen 2" support for the new AMD EPYC processors launching later this year.
EDAC is for the kernel's subsystem of Error Detection And Correction device support to allow the collecting and reporting of events pertaining to ECC memory and other errors. With the in-development Linux 5.2 kernel, the EDAC feature pull adds in the Zen 2 support patches we originally wrote about back in February.
Added to the AMD64 EDAC driver code is the new IDs for Family 16h Model 30h and newer parts, which are the next-generation Zen processors. There are also Zen 2 updates now for handling more than two unified memory controllers to a new maximum of eight memory controllers per die.
These Zen 2 EDAC support changes are part of this pull request. While we can't yet officially confirm, it does look like AMD should have the Zen 2 support generally in good standing in time for the next-generation Ryzen and processors sans potentially some missing bits like the EDAC reporting depending upon what kernel you plan to run around launch time, but all critical kernel bits should be into shape as well as associated software support like the "znver2" support within the new GCC 9 compiler. But as is usually the case for major new hardware platforms, running the newer the Linux kernel, generally is the best bet for optimal performance and functionality.
| 8
| 1,760,719,580.068194
|
https://www.phoronix.com/news/AMD-MP2-I2C-Linux-5.2
|
Linux 5.2 Should Resolve Many AMD Ryzen Laptop Touchscreens/Touchpads Not Working
|
Michael Larabel
|
Linux 5.2 is certainly going to be a big kernel release... On top of many new features and other changes, AMD Ryzen laptops will be better supported by this kernel update to be released as stable around July.
A number of AMD Ryzen laptops up to this point haven't had working touchscreens/touchpads when running on the mainline kernel. That's been due to a lack of AMD PCIe MP2 I2C controller support, but a new driver is being merged for Linux 5.2 to properly support that controller.
The i2c-amd-mp2 driver is currently queued in i2c-next for introduction with the Linux 5.2 merge window, which should open tomorrow assuming Linux 5.1 releases as expected.
This long overdue driver is based on code AMD released back in 2018 but took a rewrite and going through several rounds of review to get it into shape for merging to mainline. But now it's queued and ready for use with Linux 5.2 mainline.
If your touchpad is currently working on an AMD Zen notebook, it's possible the original AMD MP2 driver is at play. Ubuntu has been one of the distributions carrying a patch of AMD's original driver code. Now once upgrading to Linux 5.2+, the support should be even better thanks to various functionality improvements over the original driver patched into Ubuntu kernels. Other AMD laptop owners have resorted to using a DKMS version of this driver in order to make use of their laptops.
Among the many laptops this new AMD MP2 driver should help for working touchpads or touchscreens are the Dell Latitude 5495, Lenovo Yoga 530, Lenovo Ideapad 530s, Acer Nitro 5, Huawei Matebook D, and many others sporting this new controller.
| 22
| 1,760,719,581.169083
|
https://www.phoronix.com/news/AMD-Gold-Edition-50-Years
|
AMD Rolls Out Gold Edition Ryzen & Radeon VII Products For 50th Anniversary
|
Michael Larabel
|
Confirming recent leaks, AMD today announced the Ryzen 7 2700X Gold Edition and Radeon VII Gold Edition products in marking the 50th anniversary of Advanced Micro Devices.
These two Gold Edition products will come with a game bundle of Tom Clancy's The Division 2 and World War Z. Those aren't native Linux games but at least can become relevant with Steam Play on Linux. The Gold Edition Ryzen 7 2700X features Lisa Su's signature on the processor cover and gold "AMD50" packaging. The Radeon VII Gold Edition meanwhile features a red shroud and gold AMD50 packaging.
These Gold Edition products will be offered for a limited time and have no specification changes over the conventional SKUs.
More details on AMD.com.
| 15
| 1,760,719,581.363191
|
https://www.phoronix.com/news/Ryzen-Embedded-R1000
|
AMD Ryzen Embedded R1000 Pairs Dual Core Zen CPU + Vega 3 Graphics @ 12~25 Watts
|
Michael Larabel
|
AMD today announced the Ryzen Embedded R1000 series with the inaugural R1505G and R1606G SoCs.
Both of these R1000 embedded launch products are dual core / quad thread Zen CPUs with Vega 3 graphics and capable of driving three displays and both offering dual 10Gb Ethernet ports.
Both of these Ryzen Embedded SoCs offer a TDP range of 12 to 25 Watts, base clock frequencies 2.4~2.6GHz, and up to a 3.3~3.5GHz boost frequency while supporting DDR4-2400 memory.
These Ryzen Embedded R1000 SoCs will soon begin appearing in products from a variety of OEMs/ODMs. More details on AMD.com.
| 12
| 1,760,719,582.533205
|
https://www.phoronix.com/news/AMD-2nd-Gen-Ryzen-PRO-Mobile
|
AMD Announces 2nd Gen Ryzen PRO Mobile Processors
|
Michael Larabel
|
While we are quite looking forward to Ryzen 3000 (Zen 2) desktop processors this summer, for those in the market for AMD mobile processors in business notebooks/laptops, AMD today announced the 2nd Gen Ryzen PRO mobile processors.
HP and Lenovo will begin introducing new notebooks this quarter featuring the 2nd Gen Ryzen PRO processors while more notebook designs are expected later in the year.
The Ryzen PRO 3000 series mobile processors are slated to offer 16% better multi-threaded performance than Intel's competition while promising up to 10~12 hour battery life and up to 14% faster content creation for these 12nm CPUs.
The new parts are the Athlon PRO 300U, Ryzen 3 PRO 3300U, Ryzen 5 PRO 3500U, and Ryzen 7 PRO 3700U. All of these Ryzen PRO processors feature Vega graphics.
More details at AMD.com.
| 17
| 1,760,719,583.887229
|
https://www.phoronix.com/news/AMD-EPYC-Linux-5.1-Performance
|
AMD EPYC Is Running Well On Linux 5.1 Too - Performance Wins
|
Michael Larabel
|
Last week I passed along some initial benchmark results after finding Intel Cascade Lake offering up some performance improvements when using the in-development Linux 5.1 kernel. The exciting news is this doesn't appear to be Cascadelake-specific or even Intel specific as with the Dell PowerEdge EPYC 2P server I am also seeing some nice performance improvements in the same benchmarks.
I am still in the midst of conducting more Linux 5.1 kernel benchmarks albeit perpetually short on time but should have some additional Linux 5.1 data out next week. But in being curious whether Linux 5.1 is also looking up on AMD hardware, I ran some quick Linux 5.0.7 stable benchmarks against the latest Linux 5.1 Git kernel...
On this powerful Dell PowerEdge R7425 server with dual EPYC 7601 processors, I was elated to see Linux 5.1 running faster.
Equally exciting is that in my Linux 5.1 kernel testing to date I have yet to see any measurable performance regressions compared to Linux 5.0.
So far I am quite happy with how Linux 5.1 is panning out on at least Intel and AMD systems. While waiting for more Linux 5.1 benchmark, see the Linux 5.1 feature overview to learn about all that has been added/changed with this kernel. Linux 5.1.0 should be released by early May.
| 3
| 1,760,719,584.410549
|
https://www.phoronix.com/news/AMD-30h-EDAC-Bits
|
AMD Zen 2 EDAC Support Being Readied, Preps For Up To 8 Memory Controllers Per Die
|
Michael Larabel
|
AMD developers continue working out the open-source enablement bits for Linux to handle the upcoming Zen 2 processors.
We've seen more and more Zen 2 related kernel patches come out in recent months and the latest is a revision to their EDAC patches for the kernel's Error Detection And Correction code in reporting ECC errors, etc. The patches extend the existing AMD64 EDAC driver for supporting Family 17h Model 30h and beyond.
Of interest, the patches do prepare for future AMD CPUs moving to more than two memory controllers per die. One of the patches spells out clearly, "The first few models of Family 17h all had 2 Unified Memory Controllers per Die, so this was treated as a fixed value. However, future systems may have more Unified Memory Controllers per Die."
A follow-up patch also notes that these future AMD CPUs could have as many as eight memory controllers per die, "Increase the size of the struct chip_select array to eight, which is the largest number of controllers per die currently supported on AMD systems."
The revised patches are coming in a bit late so they might not make it in time for the Linux 5.1 kernel merge window opening next week, but otherwise will be material to look forward to with Linux 5.2.
| 6
| 1,760,719,585.23451
|
https://www.phoronix.com/news/AMD-Znver2-LLVM-Clang-9
|
AMD Zen 2 "znver2" Support Lands In LLVM Clang 9.0
|
Michael Larabel
|
While it didn't make it in time for the soon to be released LLVM 8.0, the latest LLVM/Clang 9.0 development code has just added support for the Zen 2 "znver2" processors.
Back in October is when AMD published the Znver2 compiler patch for GCC that builds atop the existing Zen "znver1" support while adding in the new instructions of Cache Line Write Back (CLWB), Read Processor ID (RDPID), and Write Back and Do Not Invalidate Cache (WBNOINVD). It was the first-cut support and still leveraged the same cost tables and scheduler data from the current-generation Zen processors. That support was quickly merged, making it for the upcoming GCC 9.1 stable compiler release, so that when these next-generation processors hit it will be possible to use -march=znver2 for generating optimized code for these 7nm AMD CPUs.
Over in the LLVM/Clang space, as of Tuesday the necessary bits landed in LLVM and Clang. Like the GCC 9 state, the same scheduler model as Znver1 is currently used and enables the CLWB / RDPID / WBNOINVD over what's found with Znver1.
As LLVM Clang 9.0 won't be shipping until September or so, there still is plenty of time for getting this Zen 2 CPU support better tuned. AMD Zen 2 CPUs are expected to begin rolling out around the middle of this year.
| 3
| 1,760,719,586.37521
|
https://www.phoronix.com/news/AMD-SEV-Linux-Firmware-Git
|
AMD SEV Firmware Added To Linux-Firmware Tree For Easier Updating
|
Michael Larabel
|
For those making use of Secure Encrypted Virtualization for secure VMs running on AMD EPYC platforms, the firmware bits required for supporting SEV have now been added to the linux-firmware.git tree to allow for easier updating to this virtualization security feature.
AMD Secure Encrypted Virtualization, the extension of Secure Memory Encryption (SME) to allow virtual machines to provide page-granular memory encryption support using their own per-VM secure keys, requires firmware support to function. Up to now the AMD SEV firmware updates have been distributed as part of the system BIOS images. But now to allow AMD to more easily update the SEV firmware outside of server/motherboard BIOS releases and those vendors' release schedules, AMD added support for distributing these binaries independently and offering them via the de facto Git tree for Linux firmware binaries.
Now next time you (or your distribution) pulls from Linux-Firmware.Git for updating available firmware binaries, it will include the latest AMD SEV firmware files. This should be useful for any new SEV features requiring firmware support and/or security updates to more quickly and easily be distributed to Linux users rather than being held up by system BIOS updates.
| 2
| 1,760,719,586.607582
|
https://www.phoronix.com/news/Vega-10-More-PowerPlay-Controls
|
Vega 10 & Newer Getting More Fine-Grained PowerPlay Controls On Linux
|
Michael Larabel
|
With the upcoming Linux 5.1 kernel cycle, discrete Radeon graphics cards based on Vega 10 and newer will have fine-grained controls over what PowerPlay power management features are enabled and the ability to toggle them at run-time.
Queued into the work-in-progress AMDGPU code for the eventual Linux 5.1 kernel cycle is now a ppfeatures for sysfs. This new "ppfeatures" file on sysfs will allow for querying the PowerPlay features state and toggling them individually. This includes features like GFXOFF (the ability to turn off the graphics engine when idling), automatic fan control, LED display for GPU activity, the dynamic power management state for the various blocks, and other features. Up to now the PowerPlay features couldn't be toggled individually but just a blanket enable/disable.
Also for Vega 10 and newer dGPUs with this next Linux kernel will be pp_dpm_socclk for being able to read and adjust the SOC clock power levels. Likewise, pp_dpm_fclk but that interface will be just for Vega 20 and newer. Rounding out these PowerPlay additions is a pp_dpm_dcefclk for adjusting its value on Vega 10 and newer. Though without any full-featured AMDGPU GUI control panel for Linux, users will be left to manually tinkering with their values via the sysfs files.
Outside of the PowerPlay space, the AMDGPU code being worked on for Linux 5.1 also adds support for self IRQ on Vega 10 with the processing of rings one and two. This work and more (including dropping the old Sea Islands dynamic power management code) is currently building up in drm-next-5.1-wip.
| 5
| 1,760,719,587.628222
|
https://www.phoronix.com/news/Noctua-NH-U9-TR4-SP3-2019
|
Noctua's NH-U9 TR4-SP3 Is Still The Best 4U EPYC / Threadripper Cooler I've Found
|
Michael Larabel
|
If you are in the market for an AMD Ryzen Threadripper or AMD EPYC heatsink that fits within 4U height requirements, the Noctua NH-U9 TR4-SP3 is still easily the best option available. I'm now running the NH-U9 TR4-SP3 in five different EPYC/Threadripper systems in the racks and they work out splendid.
I've already covered the NH-U9 TR4-SP3 multiple times before, but with having picked up another one of these coolers this past week and being satisfied with the results, just wanted to give another shout-out to Noctua and pass along the latest thermal results. For this latest build, the NH-U9 TR4-SP3 is cooling an EPYC 7351P 16-core / 32-thread CPU that tops out at 3.9GHz.
I've looked at the Noctua cooling performance in past rounds with Cooling AMD EPYC With Noctua Coolers (several different models tested) and the original Noctua NH-U9 TR4-SP3 Threadripper testing from 2017. There was also the Noctua Air Cooling With The AMD Ryzen Threadripper 2970WX testing.
With having tested the NH-U9 TR4-SP3 with Threadripper 2 and EPYC 7551 CPUs, to little surprise the cooler worked out great with the EPYC 7351P that was running within a 4U Rosewill rackmount chassis.
I ran some tests with the cooler at the default auto/PWM behavior and then again when the Noctua fans were running at their forced full-speed, which makes the system slightly louder while with the default PWM behavior the fans are extremely quiet. But even with these fans at full-speed, still much quieter than normal 1U/2U cooling...
The 7351P EPYC was operating at about 37 degrees during the demanding test of compiling the Linux kernel or it was about three degrees cooler on average with the fans forced to their full-speed.
Even during the much longer process of compiling the entire GNU Compiler Collection, the 7351P was operating very well with the Noctua 4U-compatible cooler and quiet with the default PWM control.
With multi-threaded video encoding, the 7351P stuck to 40 degrees or less on average.
Blender was one of the most demanding workloads where the EPYC 7351P topped out at just 46 degrees in the stock configuration.
Lastly is a look at the cooling performance over the span of several hours with various single and multi-threaded workloads for this system. With the stock configuration in this 4U chassis, the EPYC CPU never even hit 50 degrees. Running with the fans always at full-speed led to a slightly noisier setup but was about four degrees cooler.
Those looking for Socket TR4/SP3 cooling solutions can find the Noctua NH-U9 TR4-SP3 and others via NewEgg.
| 9
| 1,760,719,588.064646
|
https://www.phoronix.com/news/AMD-CES-2019-Keynote
|
AMD CES 2019 Keynote: Say Hello To Radeon 7 As 2nd Gen Vega
|
Michael Larabel
|
While Monday was the Intel keynote at CES 2019, this morning AMD took the spotlight with Lisa Su's keynote in Las Vegas.
Highlights from the AMD CES 2019 keynote included:
- 2nd Gen AMD Ryzen Mobile CPUs as announced earlier this week.
- The first Acer and HP AMD-powered Chromebooks are coming out soon.
- The Radeon 7 (Radeon VII) was announced today as their new high-end GPU. Radeon 7 is their first consumer desktop 7nm GPU. 60 compute units, 16GB of HBM2 memory. 1TB/s memory bandwidth.
- Radeon 7 will have up to 30% better performance over the RX Vega 64 in content creation workloads like Blender and up to 60% better for OpenCL. Windows gaming performance is 25~42% faster over RX Vega 64 with 4K gaming.
- Radeon VII will launch 7 February for $699 USD.
- Google Project Stream (cloud video game streaming) will be powered by Radeon Vega graphics.
- 2nd Gen EPYC with 7nm Zen 2 seeing 2x performance per socket, 4x floating performance per socket.
- One 7nm AMD EPYC CPU was shown to outperform dual Xeon Platinum 8180 processors for NAMD.
- 2nd Gen EPYC CPUs are on track for shipping in the middle of 2019.
- 3rd Gen Ryzen (Zen 2) desktop CPUs were previewed. A demo shown on Windows put a early sample of 3rd Gen Ryzen outperforming an Intel Core i9 9900K while having ~30% lower power consumption.
- 3rd Gen Ryzen is using a chiplet design with 8C/16T and I/O die.
- 3rd Gen Ryzen will support PCI Express 4.0.
| 74
| 1,760,719,589.350835
|
https://www.phoronix.com/news/Ryzen-3000-Series-Mobile
|
AMD Announces Ryzen 3000 Series Mobile Processors, 7th Gen A-Series For Chromebooks
|
Michael Larabel
|
Ahead of the big AMD keynote on Tuesday, AMD kicked off CES week today by announcing the Ryzen 3000 Series Mobile Processors.
The AMD Ryzen 3000 Series Mobile Processors range from the Athlon 3000U at 2 cores / 4 threads with a 15 Watt TDP and 3.3GHz boost clock speed all the way up to the Ryzen 7 3750H at 4 cores / 8 threads with 4.0GHz boost frequency (2.3GHz base) and a 35 Watt TDP. The new Ryzen Mobile processors all have Vega graphics ranging from 3 to 10 GPU cores with GPU clock frequencies up to 6MB. These new Ryzen Mobile 3000 Series processors are quite interesting and hopefully we'll see some interesting notebook designs soon with these Zen+Vega processors.
AMD also announced the 7th Gen AMD A-Series processors that will be powering Acer and HP Chromebooks to be announced later this month. The AMD A4-9120C is a two core / two thread processor with 2.4GHz boost frequency and 1.6GHz base frequency with GCN 1.2 "Radeon R4" graphics. There is also an AMD A6-9220C that is also a dual-core / dual-thread part with 2.7GHz boost frequency and 1.8GHz base frequency while having Radeon R5 GCN 1.2 graphics.
More details on the new AMD parts announced today via this press release.
| 31
| 1,760,719,589.562966
|
https://www.phoronix.com/news/AMD-Radeon-Linux-News-2018
|
The Most Popular AMD/Radeon Linux News Of 2018
|
Michael Larabel
|
After looking yesterday at the most popular Intel Linux news of 2018, the tables have turned and this article is looking at the most popular AMD/Radeon news for the year on Phoronix.
This year was very interesting for AMD with the successful launch of the Threadripper 2 series, new Ryzen CPUs, EPYC processors continuing to pickup sever marketshare, AMD being less affected by Spectre/Meltdown vulnerabilities compared to Intel, the Radeon graphics driver stack continuing to advance a great deal (finally having FreeSync is the latest feature ending out the year!), early hardware enablement signs for Zen 2 / next-gen EPYC, continued znver1 compiler tuning, and much more happened in 2018.
Of 454 AMD Linux/open-source original news articles on Phoronix this year, the most popular included:
AMD Cuts Ryzen Prices, Confirms New Hardware, New Ryzen CPUs With Vega While Intel announced their new CPUs with Radeon Vega M graphics, AMD had a host of announcements on their own for getting CES 2018 started with some excitement. NVIDIA RTX, AMD On Linux & Distro Performance Dominated Linux Discussions In October During the month of October on Phoronix there were 330 original news stories and 26 featured articles / Linux hardware reviews penned by your's truly. For Now At Least AMD CPUs Are Also Reported As "Insecure" Right now with the big mysterious security vulnerability causing the rush of the x86 Page Table Isolation work that landed in the Linux kernel days ago, it's believed to be a problem only affecting Intel CPUs. But at least for now the mainline kernel is still treating AMD CPUs as "insecure" and is too taking a performance hit. Linux Will End Up Disabling x86 PTI For AMD Processors - Update: Now Disabled While at the moment with the mainline Linux kernel Git tree AMD CPUs enable x86 PTI and are treated as "insecure" CPUs, the AMD patch for not setting X86_BUG_CPU_INSECURE will end up being honored. Sony Is Working On AMD Ryzen LLVM Compiler Improvements - Possibly For The PlayStation 5 One of Sony's compiler experts has taken to working on some tuning for the AMD Ryzen "znver1" microarchitecture support within the LLVM compiler stack. This begs the question why Sony is working on Ryzen improvements if not for a future product. Hygon Dhyana: Chinese x86 Server CPUs Based On AMD Zen While there are the VIA/Centaur-based Zhaoxin desktop CPUs targeted for the Chinese market, it turns out there is another x86 Chinese CPU effort but this time is a collaboration with AMD. AMD PSP Affected By Vulnerability While all eyes have been on Intel this week with the Spectre and Meltdown vulnerabilities, a disclosure was publicly made this week surrounding AMD's PSP Secure Processor in an unrelated security bulletin. Seven Reasons To Already Get Excited For Linux 4.17, Especially For AMD/Radeon Users While Linux 4.16 is coming in the next few days, I am already quite excited about the upcoming Linux 4.17 kernel cycle and the changes it will bring. Linux 4.17 To Enable AMDGPU DC By Default For All Supported GPUs Since the introduction of the AMDGPU DC display code (formerly known as DAL) in Linux 4.15, this modern display stack has just been enabled by default for newer Radeon Vega and Raven Ridge devices. With Linux 4.17 that is changing with AMDGPU DC being enabled by default across the board for supported GPUs. AMD Retpoline Benchmarks From FX To Threadripper & EPYC For those curious about the performance impact of the Retpoline patches as found in the latest Linux 4.15 kernel, here are some benchmarks on an assortment of old and new AMD Linux systems. AMD FreeSync 2 HDR Coming To The Linux Kernel In 2019 Next year is when all of the pieces of the open-source puzzle for fully supporting FreeSync/Adaptive-Sync/VRR displays with AMD graphics cards should be in place for allowing out-of-the-box support. AMD Did NOT Disable Branch Prediction With A Zen Microcode Update With the plethora of software security updates coming out over the past few days in the wake of the Meltdown and Spectre disclosure, released by SUSE was a Family 17h "Zen" CPU microcode update that we have yet to see elsewhere... It claims to disables branch prediction, but I've confirmed with AMD that is not actually the case. This Chart Shows How The Radeon RX 580 vs. GeForce GTX 1060 Now Compete Under Linux It was just last year that open-source RadeonSI/RADV developers were trying to get the Radeon RX 580 "Polaris" GPU to be competitive with the GeForce GTX 1060 as it is under Windows given each GPU's capabilities. We've seen the RX 580 and GTX 1060 dancing under Linux the past few months and yesterday's 20-way GPU comparison with Rise of the Tomb Raider was quite significant -- perhaps most surprising being how well the RX 580 performed. Heck, just one or two years ago it was an accomplishment seeing any official Radeon driver support at-launch for new Linux game releases. So here are some extensive tests looking closer at the GTX 1060 vs. RX 580 battle in this latest Vulkan-powered Linux game port. AMD Zen CPU Microcode Added To Linux-Firmware Tree, Bulldozer Updated Ensuring your CPU microcode is kept up-to-date for Zen processors is now a little bit easier with the microcode files being added to the linux-firmware.git collection. A Look At The Linux vs. Windows Power Use For A Ryzen 7 + Radeon RX Vega Desktop Recently I have been posting a number of Linux laptop battery benchmarks including how the power consumption compares to Windows 10. If you are curious how these numbers play out on the desktop side and when using AMD hardware, here are some results for your viewing pleasure with a Ryzen 7 2700X and Radeon RX Vega 64 desktop system. Radeon Vega 20 Will Have XGMI - Linux Patches Posted For This High-Speed Interface It has been expected that Vega 20 would feature XGMI as a high-speed GPU interconnect alternative to PCI Express and that was firmed up today thanks to a new set of AMDGPU Linux driver patches. AMD Posts Open-Source Driver Patches For Vega 12 It's been a while since last hearing anything about the rumored "Vega 12" GPU but coming out this morning are a set of 42 patches providing support for this unreleased GPU within the mainline Linux kernel. AMD's Raven Ridge Botchy Linux Support Appears Worse With Some Motherboards/BIOS With my launch testing of the Raven Ridge desktop APUs with the Ryzen 5 2400G and Ryzen 3 2200G there were some stability issues to report and some hangs within games and mode-setting issues. It appears those issues are exacerbated with some motherboards: the past few days with two different AMD B350 motherboards have been a real pain getting the current AMDGPU driver stack working -- and even Linux 4.17 AMDGPU WIP code -- on either of these Raven Ridge APUs. AMD Finally Rolls Out New Linux Patches For Adaptive-Sync / VRR (FreeSync) While the open-source AMD Radeon Linux graphics stack has gotten into particularly good shape the past two years or so, one of the areas that has left the red Linux gamers unsatisfied is the lack of FreeSync support (or DisplayPort Adaptive-Sync / HDMI Variable Refresh Rate) when using the fully open-source stack. It looks like that could be changing soon with the new set of patches under review. Linux Kernel Expectations For AMD Threadripper 2 If you have already pre-ordered your AMD Threadripper 2990WX processor or just planning to be an early customer of that high-end desktop processor or the Threadripper 2950X, you may be wondering about Linux requirements from these new high-end AMD CPU offerings. Here's the gist of the Linux support state of AMD Zen+ CPUs for those wanting to get ready for Threadripper 2. And our most viewed reviews/benchmarks:
AMD Threadripper 2990WX Linux Benchmarks: The 32-Core / 64-Thread Beast Whether you are compiling a lot of code, rendering models with Blender, or running various scientific workloads with OpenMP or MPI, the AMD Threadripper 2990WX is capable of delivering immersive Linux performance with its 32-cores and 64 total threads. While coming in at $1800 USD, the AMD Threadripper 2990WX can deliver better performance than the more expensive Intel Core i9 7980XE. Beyond being mesmerized about the performance today with this high-end desktop/workstation processor with the many thread-happy Linux workloads we encounter daily, this 32-core Zen+ processor has us even more eager to see AMD's next-generation Zen2-based EPYC CPUs next year. A Look At The Windows 10 vs. Linux Performance On AMD Threadripper 2990WX Complementing the extensive Linux benchmarks done earlier today of the AMD Threadripper 2990WX in our review (as well as on the Threadripper 2950X), in this article are our first Windows 10 vs. Linux benchmarks of this 32-core / 64-thread $1799 USD processor. Tests were done from Microsoft Windows 10 against Clear Linux, Ubuntu 18.04, the Arch-based Antergos 18.7-Rolling, and openSUSE Tumbleweed. POWER9 Benchmarks vs. Intel Xeon vs. AMD EPYC Performance On Debian Linux For several days we've had remote access to one of the brand new Raptor Talos II Workstations that is powered by POWER9 processors and open-source down through the firmware. For those curious how these latest POWER processors compare to AMD EPYC and Intel Xeon processors, here are some benchmarks comparing against of the few other systems in house while all testing was done from Debian GNU/Linux. Ryzen 5 2400G Radeon Vega Linux OpenGL/Vulkan Gaming Benchmarks Here are our initial performance figures for the Vega graphics found on the newly-released Ryzen 5 2400G "Raven Ridge" APU under Linux and testing both OpenGL and Vulkan graphics benchmarks. CPU tests as well as benchmarks of the Ryzen 3 2200G under Linux are forthcoming on Phoronix. AMD Ryzen 5 2600X + Ryzen 7 2700X Linux Benchmarks The embargo on the Ryzen 5 2600X and Ryzen 7 2700X processors has expired now that these Ryzen+ CPUs are beginning to ship today. We can now talk about the Linux support and the initial performance figures for these upgraded Zen desktop CPUs. A Look At The Windows vs. Linux Scaling Performance Up To 64 Threads With The AMD 2990WX This past week we looked at the Windows 10 vs. Linux performance for AMD's just-launched Ryzen Threadripper 2990WX and given the interest from that then ran some Windows Server benchmarks to see if the performance of this 64-thread CPU would be more competitive to Linux. From those Windows vs. Linux tests there has been much speculation that the performance disparity is due to Windows scheduler being less optimized for high core/thread count processors and its NUMA awareness being less vetted than the Linux kernel. For getting a better idea, here are benchmarks of Windows Server 2019 preview versus Ubuntu Linux when testing varying thread/core counts for the AMD Threadripper 2990WX. AMD Ryzen 7 Performance On Windows 10, Windows Server 2016, Six Linux Distributions Our latest Windows vs. Linux benchmarking interest has been seeing how the AMD Ryzen 7 performance compares with the latest operating systems / Linux distributions. We have recently posted some Windows 10 vs. Windows WSL vs. Windows Linux benchmarks, relative Spectre/Meltdown mitigation impact tests on Windows vs. Linux, and other benchmarks but has mostly been done with Intel or server hardware. For those curious, today's tests were done with an AMD Ryzen 7 1700 platform. AMD Athlon 200GE: Benchmarking The $60 Zen+Vega Chip At the high-end of the AMD desktop CPU 2018 spectrum is the insanely fast Threadripper 2990WX while at the opposite end of that spectrum is the recently announced Athlon 200GE. For just $60 USD is this Zen+Vega chip that we have begun testing and have our initial Linux performance benchmarks out today compared to a range of lower-end and older desktop CPUs as well as integrated graphics test results, power consumption data, and performance-per-dollar metrics. Windows 10 vs. Ubuntu Linux With Radeon / GeForce GPUs On The Latest 2018 Drivers Given how fiercely the latest open-source AMD Linux driver code is running now up against NVIDIA's long-standing flagship Linux GPU driver, you might be curious how well that driver stacks up against the Radeon Software driver on Windows? Well, you are in luck as here are some fresh benchmarks of the Radeon RX 580 and RX Vega 64 as well as the GeForce GTX 1060 and GTX 1080 Ti while being tested both under Microsoft Windows 10 Pro x64 and Ubuntu 16.04 LTS while using the latest AMD/NVIDIA drivers on each platform. NVIDIA GeForce vs. AMD Radeon Linux Gaming Performance At The Start Of 2018 Here is a fresh look at the NVIDIA GeForce and AMD Radeon Linux graphics card performance as we start 2018. Testing was done using the latest Linux 4.15 Git kernel -- including the KPTI page table isolation support -- as well as using the newest Mesa 17.4-dev driver code for RadeonSI/RADV and on the NVIDIA side is their brand new 390.12 beta driver. AMD Vega 8 Graphics Performance On Linux With The Ryzen 3 2200G Yesterday I posted the initial Ryzen 5 2400G Vega 11 Linux graphics benchmarks while for your viewing please today -- as well as this morning's 21-way Intel/AMD CPU Linux comparison that featured these new Raven Ridge APUs -- the results now completed are initial OpenGL and Vulkan performance figures for the Vega 8 graphics found on the Ryzen 3 2200G. The NVIDIA/AMD Linux GPU Gaming Benchmarks & Performance-Per-Dollar For July 2018 In part with GPU demand by crypto-currency miners waning a bit, NVIDIA GeForce and AMD Radeon graphics card availability at retailers has been improving in recent weeks as well as seeing less inflated prices than just recently had been the case. Given the better availability and stabilizing prices, here is a fresh look of the current line-up of GeForce and Radeon graphics cards under Ubuntu Linux using the newest AMD/NVIDIA drivers and also providing performance-per-dollar metrics given current retail prices. AMD Ryzen 3 2200G + Ryzen 5 2400G Linux CPU Performance, 21-Way Intel/AMD Comparison Yesterday I posted some initial Linux benchmarks of the Ryzen 5 2400G Raven Ridge APU when looking at the Vega 11 graphics, but for those curious about the CPU performance potential of the Ryzen 5 2400G and its ~$100 Ryzen 3 2200G sibling, here are our first CPU benchmarks of these long-awaited AMD APUs. These two current Raven Ridge desktop APUs are compared to a total of 21 different Intel and AMD processors dating back to older Kaveri APUs and FX CPUs and Ivy Bridge on the Intel side. Raptor Talos II POWER9 Benchmarks Against AMD Threadripper & Intel Core i9 For those curious about the performance of IBM's POWER9 processors against the likes of today's AMD Threadripper and Intel Core i9 HEDT processors, here are some interesting benchmarks as we begin looking closer at the POWER9 performance on the fully open-source Raptor Talos II Secure Workstation. This open-source, secure system arrived for Linux testing with dual 22-core POWER9 CPUs to yield 176 total threads of power. AMD Ryzen 5 2600 / Ryzen 7 2700 Benchmarks On Linux, 9-Way Ubuntu CPU Comparison Last month we delivered launch-day Linux benchmarks of the Ryzen 5 2600X and Ryzen 7 2700X for these new "Zen+" processors while recently we received the non-X Ryzen 5 2600 and Ryzen 7 2700 processors for Linux testing as well. In this article are benchmarks of these new AMD Ryzen processors as well as other Intel/AMD CPUs for delivering a fresh nine-way Linux distribution comparison using the very latest software components. AMD Ryzen 7 2700X: Windows 10 vs. Linux Performance Recently there have been several Linux distribution benchmark comparisons on Phoronix to test the latest Linux OS releases, including several comparing to the current Microsoft Windows 10 performance. Those recent tests have all be done with various Intel CPUs, but for those curious about the AMD Windows vs. Linux performance, here are some fresh benchmarks as we approach the end of July. Radeon GPUs Are Increasingly Competing With NVIDIA GPUs On Latest RadeonSI/RADV Drivers As it's been a few weeks since last delivering a modest Linux GPU comparison and given the continuously evolving state of the Linux kernel Git tree as well as the Mesa project that houses the RadeonSI OpenGL and RADV Vulkan drivers, here are our latest benchmarks showing the current state of the AMD Radeon open-source Linux graphics driver performance relative to NVIDIA's long-standing and high-performance but proprietary driver using several different graphics cards. 36-Way Comparison Of Amazon EC2 / Google Compute Engine / Microsoft Azure Cloud Instances vs. Intel/AMD CPUs Earlier this week I delivered a number of benchmarks comparing Amazon EC2 instances to bare metal Intel/AMD systems. Due to interest from that, here is a larger selection of cloud instance types from the leading public clouds of Amazon Elastic Compute Cloud, Microsoft Azure, and Google Compute Engine. The Current Linux Performance On 22 Intel / AMD Desktop Systems For your Linux benchmark viewing pleasure today are test results from twenty-two distinct Intel / AMD systems when running a recent release of the performance-optimized Clear Linux distribution and the hardware spanning from old AMD FX and Intel Core i3 Haswell CPUs up through the high-end desktop Core i9 and Threadripper processors. Windows Server 2019 vs. Linux vs. FreeBSD Performance On A 2P EPYC Server When Microsoft rolled out their Windows 10 October 2018 Update they also released Windows Server 2019. Now over the slower holiday period I am finally getting caught up in benchmarking Windows Server 2019. For this initial benchmark comparison is a look at the Microsoft Windows Server 2019 performance against a handful of Linux distributions as well as FreeBSD 12.0 for seeing how this latest Windows Server performance compares on a dual AMD EPYC 7601 server.
| 6
| 1,760,719,590.924731
|
https://www.phoronix.com/news/AMD-F17-M30h-Linux-Temp-Changes
|
AMD Zen 2 Temperature Monitoring Changes Sent In For Linux 4.21
|
Michael Larabel
|
When AMD Zen CPUs originally rolled out, the ability to monitor the CPU core temperatures under Linux didn't roll out until months later. Fortunately, for Zen 2 the AMD Linux CPU temperature driver looks like it will be ready in time.
Back in November when patches first emerged I commented on AMD already working on temperature driver support for Zen 2, or AMD Family 17 Model 30h and newer. Special changes are needed with the new processors having multiple roots per Data Fabric / SMN interface.
That code is now sent in as x86/amd-nb changes for the Linux 4.21 kernel.
Updates the data fabric/system management network code needed to get k10temp working for M30h. Since there are now processors which have multiple roots per DF/SMN interface, there needs to some logic which skips N-1 root complexes per DF/SMN interface. This is because the root complexes per interface are redundant (as far as DF/SMN goes). These changes shouldn't effect past processors and, for F17h M0Xh, the mappings stay the same. This temperature monitoring support is on top of other "Zen 2" Linux patches we've begun seeing in Q4'2018 for ensuring these next-gen AMD processors will be fully supported by Linux by the time they officially launch in 2019.
| 1
| 1,760,719,591.216934
|
https://www.phoronix.com/news/Improved-AMD-Microcode-Loading
|
Improved AMD CPU Microcode Handling On Deck For Linux 4.21
|
Michael Larabel
|
With CPU microcode updates having become increasingly important over the past year in light of the Spectre vulnerabilities and other security updates, the Linux 4.21 kernel is bringing several improvements to the AMD CPU microcode update handling.
The AMD CPU microcode update handling now does more verification work to ensure it's not corrupted, the microcode loading code has been cleaned up and seen more unification of the code paths, and other improvements.
Borislav Petkov commented in the pull request, "The result is a set of verification routines which validate the microcode blobs before loading it on the CPU. In addition, the code is a lot more streamlined and unified. In the process, some of the aspects of patch handling and loading were simplified."
| 0
| 1,760,719,592.321075
|
https://www.phoronix.com/news/AMD-QoS-Landing-Linux-4.21
|
AMD Platform QoS Support For Next-Gen EPYC Processors Landing In Linux 4.21
|
Michael Larabel
|
The AMD Platform QoS support talked about a few months ago on Phoronix is landing for the upcoming Linux 4.21 kernel. While not officially confirmed, this Quality of Service system resource work appears almost surely for the next-generation 7nm EPYC processors coming out in the months ahead.
The AMD QoS platform support is for monitoring the usage of different system resources as well as for allowing limits to be placed on these different resources. Among the initial functionality exposed is around L3 cache monitoring and limiting, L3 data prioritization, and memory bandwidth enforcement.
This AMD QoS code is similar to Intel RDT (Resource Director Technology) and in fact re-uses some of that code within the Linux kernel. AMD previously documented this processor extension here though that technical document now appears to have been removed. Given the timing of these patches, it seems logical this will be supported by the upcoming 2019 EPYC processors.
With re-using the common bits of the Intel RDT code, Intel's INTEL_RDT Kconfig switch for this functionality has been renamed to "RESCTRL" for controlling whether both Intel RDT and AMD Platform QoS support is built into the kernel.
This Platform QoS code has been called for pulling into the upcoming Linux 4.21 merge window via the x86/cache pull request.
| 2
| 1,760,719,592.661067
|
https://www.phoronix.com/news/AMD-Two-More-200-Series
|
AMD Rolling Out Two More Sub-$75 Zen + Vega 3 Processors
|
Michael Larabel
|
It's too late to get as a stocking stuffer, but AMD announced today that two more Athlon 200 series parts with two Zen cores and Vega 3 graphics.
At the start of the quarter, AMD announced the Athlon 200GE as a dual-core Zen part with SMT to provide four threads, a 3.2GHz clock frequency, and Vega 3 graphics while having a 35 Watt TDP... All for just $55 USD. You can see our Athlon 200GE Linux review. It's not all that bad if you are on an extremely tight budget or just want a very low-end, low-power modern CPU/APU.
Adding to the family is now the Athlon 220GE and 240GE. The Athlon 220GE at $65 USD retail is still a dual-core / quad-thread part with Vega 3 graphics but with a 3.4GHz clock speed... So $10 more for a 200MHz bump and sticking to the 35 Watt TDP. The Athlon 240GE meanwhile has a 3.5GHz clock speed for $75 USD while the other specs are similar to the rest of the Athlon 200 series line-up.
These new budget x86_64 processors will be available for ordering today but it doesn't look like they will ship until Q1. More details on the new chips at AMD.com.
| 15
| 1,760,719,593.700419
|
https://www.phoronix.com/news/AMD-ACP3-Linux-4.21-Sound
|
AMD Audio Co-Processor 3.x Support Coming For Linux 4.21
|
Michael Larabel
|
Coming to the next version of the Linux kernel is new sound driver support for the AMD ACP3x (Audio Co-Processor 3.x).
This new AMD ALSA SoC PCM driver is used initially for AMD Raven Ridge APUs with the Audio Co-Processor 3.x (ACP3) but appears to be used by future AMD hardware too. This ACP3x driver is a bit late for supporting audio on Raven Ridge hardware with an I2S codec, but better late than never. Going back years there has been Linux support for support for older versions of the ACP.
The new code is queued now in the sound subsystem's "for-next" branch ahead of the Linux 4.21 kernel opening up next week.
On the Intel side meanwhile this next Linux kernel cycle will bring support for Intel SoCs with an RT5660 I2S audio codec. Today's sound merge also adds support for the Asahi Kaesi AKM4118, Meson AXG S/PDIF inputs, and various Qualcomm and Xilinx parts.
| 4
| 1,760,719,594.189416
|
https://www.phoronix.com/news/Linux-Firmware-AMD-Christmas
|
Vega 12 Firmware Lands Along With RX 590 Polaris Bits, Updated Zen CPU Microcode
|
Michael Larabel
|
A big serving of AMD firmware/microcode updates landed on Friday in the linux-firmware.git canonical tree for both AMD Zen processors and Radeon graphics processors.
On the CPU side, the recent AMD Zen CPU microcode update I wrote about at the end of November is now merged. Though there still isn't any public change-log that explains what has changed by this microcode update for Family 17h processors.
On the Radeon GPU firmware side is SMC firmware and MC firmware needed for properly initializing the Radeon RX 590. These firmware blobs plus what will be Linux 4.20 or newer allow this newest Polaris variant to properly work under Linux. The AMDGPU changes are pending through "-fixes" updates for clearing up the RX 590 support with Linux 4.20. See the Radeon RX 590 Linux benchmarks for more details on the current state.
Also new is Vega 12 firmware blobs were added to the linux-firmware.git tree for all the relevant hardware blocks for Vega 12.
Rounding out the linux-firmware.git activity is updating the Polaris 10/11, Vega 10, and Raven Ridge firmware against the new Radeon Software 18.50 Linux driver release.
| 14
| 1,760,719,595.0733
|
https://www.phoronix.com/news/AMD-Vega-2-Trademark
|
AMD Files Trademark For Vega II
|
Michael Larabel
|
It looks like AMD could be announcing Vega II as new 7nm Vega GPUs soon complementing the recently announced Vega 20 Radeon Instinct MI50 / MI60 accelerators.
On Friday afternoon we reported of AMD quietly adding a number of new PCI device IDs to their Linux driver that are Vega 10/20 parts. With it coming to light today about a recent trademark filing on Vega II, it's likely those new PCI IDs could be the parts going under this new branding.
The filing on their new image trademark can be found via Justia.com and USPTO. The new logo is a revision to the original Vega logo but with two marks comprising the V.
It's possible AMD could be announcing the new Vega parts next month at CES.
| 6
| 1,760,719,595.624633
|
https://www.phoronix.com/news/AMD-Zen-November-Microcode
|
Updated AMD Zen CPU Microcode Posted
|
Michael Larabel
|
AMD has just dropped a new Family 17h "Zen" microcode file for Linux users.
Just posted to the kernel mailing list is the latest CPU microcode file for Family 17h. Unfortunately there is no change-log/notes and obviously with it being binary-only there isn't anything to glean by itself.
Changes for Zen 2? Spectre-related security updates? Performance? Unfortunately no word yet, but I have fired off an inquiry to AMD to hopefully learn more about this first AMD 17h microcode update in six months. I wouldn't be surprised if it's pertaining to Zen 2 support considering the other recent open-source/Linux Zen 2 patches we've been seeing recently.
This new firmware should soon be available via the linux-firmware Git tree for then appearing on the various Linux distribution updates.
| 15
| 1,760,719,596.43852
|
https://www.phoronix.com/news/Stoney-Ridge-ALSA-HDA-Linux
|
Linux's ALSA HDA Code Finally Seeing AMD Stoney Ridge Support
|
Michael Larabel
|
While Stoney Ridge was AMD's 2016 APU platform with Excavator CPU cores and GCN 1.2 graphics, the Linux support in some regards is still being settled in some areas.
It was just earlier this year that AMD CPU temperature driver added Stoney support and now with the upcoming Linux 4.21 cycle but to be back-ported to the existing stable series is the ALSA HDA audio support.
Back during the Linux 4.15 kernel cycle was work on Stoney Ridge support but it turns out a PCI device ID was left out for supporting Stoney Ridge within the HDA audio driver.
As of today that's queued in the sound-next code branch ahead of Linux 4.21. Just the PCI ID was needed to be added and it also works out with D3 run-time power-savings to help lower the power drain, unlike some earlier AMD hardware that was problematic in that regard. The simple patch is marked for back-porting to the stable series, so eventually AMD Stoney audio will work on Linux 4.20 and friends.
This patch actually comes courtesy of Canonical with their developers having noticed the support missing.
| 9
| 1,760,719,597.142001
|
https://www.phoronix.com/news/Linux-4.20-AMD-MCE-Fix
|
Linux 4.20 To Receive Fix That Prevented Some AMD Raven Ridge Systems From Booting
|
Michael Larabel
|
It looks like the mainline Linux 4.20 kernel within a few days will be playing nicely on more AMD hardware. In particular, the Raven Ridge Zen+Vega APUs that have been rather troublesome depending upon the BIOS/motherboard since their launch almost one year ago.
Queued in this morning in the x86/urgent branch, which should be sent into Linus Torvalds as part of fixes for the Linux 4.20 kernel soon, is fixing the thresholding machinery initialization order in the AMD-specific x86 MCE code. The commit explains, "Currently, the code sets up the thresholding interrupt vector and only then goes about initializing the thresholding banks. Which is wrong, because an early thresholding interrupt would cause a NULL pointer dereference when accessing those banks and prevent the machine from booting. Therefore, set the thresholding interrupt vector only *after* having initialized the banks successfully."
The bug introducing this issue to the AMD MCE kernel code was merged in 2016. Of bug reports related to this issue, the Ryzen 2500U / Raven Ridge hardware seems to be a common trait. But depending upon the motherboard and particular BIOS version, the issue may or may not present itself. A workaround up until now has been booting the affected kernels with mce=off to avoid this null pointer dereferencing that would halt the boot process.
It's good that this issue is now being cleared up and one of my Raven Ridge systems was running into kernel boot issues with its latest BIOS that I'll now try again once this commit lands into Linux 4.20. Presumably this change will also get backported to current stable/LTS series.
| 22
| 1,760,719,597.862652
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.