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-SEV-V8-Encrypted-Virt
|
AMD Secure Encrypted Virtualization Patches Updated For Linux
|
Michael Larabel
|
AMD Linux developers today sent out the latest revision to their big set of patches adding in support for Secure Encrypted Virtualization (SEV) to the Linux kernel.
Landing with the Linux 4.14 kernel that will be released next weekend is AMD Secure Memory Encryption support for use with new EPYC processors and the related AMD Secure Processor support. Building off that is Secure Encrypted Virtualization (SEV) that allows for virtual machines to have their memory encrypted and secured in a way that only the guest itself can access the unencrypted data. Each VM backed by SEV on the AMD EPYC servers has its own unique encryption key and ties into the AMD Secure Processor.
AMD Secure Encrypted Virtualization has a lot of potential for better securing public clouds and other practical use-cases for better safeguarding VMs. Today the eighth version of the SEV patches were published for the Linux kernel.
More details on this work can be found via the kernel patch series and AMD's SEV whitepaper.
With the Linux 4.15 kernel merge window opening next week, unfortunately it's looking quite tight for getting SEV into Linux 4.15, but we'll see for sure soon enough. SEV support also requires patches against QEMU too, so long story short it will likely be a few months still before this SEV support begins appearing prominently on AMD EPYC servers -- especially for the enterprise distributions.
| 4
| 1,760,719,681.752763
|
https://www.phoronix.com/news/Monero-Mining-Threadripper-DIMM
|
Monero/XMR Mining On Threadripper With Multi-Channel Memory
|
Unknown
|
Phoronix reader Thomas Frech has shared with us another article on Monero/XMR cryptocurrency mining performance with AMD Threadripper. Thsi follows his recent guest posts of mining Ethereum with Threadrippers and AMD Vega GPUs, Ethereum and Monero mining on the same systems, and the AMDGPU-PRO 17.40 crypto mining boost. In one of my last articles I went from one RAM DIMM per system to two DIMMs for the 1920X. This resulted in a performance jump from 500H/s to 1400H/s. Because of this I bought 5 more RAM DIMMs to test if we maybe get even better performance.
But not so much going from two to four DIMMs for the 1920X resulted in no better result than going from one to four DIMMs on the 1900X system resulted in very little improvement, maybe 50-100H/s better.
This implies that Monero on Threadripper wants dual channel memory per Threadripper system but there is no need for more. This also proved that 1900X do not need many RAM DIMMs to perform good.
I do not know the result for the 1950X maybe the full 16 cores need four DIMMs but we do not know this yet.
Some people also pointing out that because Threadripper is a NUMA system with two NUMA blocks I need to split the XMR worker thread into 2 instances per core. I will test this Later.
I also wait for my Power meter from Amazon to get some efficiency results.
| 6
| 1,760,719,682.290547
|
https://www.phoronix.com/news/AMD-EPYC-Sched-Linux-4.15
|
Linux 4.15 Will Have A Scheduler Change To Benefit AMD EPYC
|
Michael Larabel
|
Linux 4.15 will be exciting for AMD Zen systems not only for working temperature reporting (finally) being in place for Ryzen/EPYC, but AMD EPYC CPUs should also benefit from a scheduler topology improvement.
Suravee Suthikulpanit of AMD has been working on a patch for Family 17h / EPYC systems where currently the kernel's setup of scheduler domains is less than optimal with not being properly balanced across the NUMA nodes.
The current kernel's behavior can lead to cases where tasks aren't properly balanced across logical NUMA nodes. In other cases, threads are scheduled for the same CPU while other CPU cores could be idling, yielding an inconsistency in performance.
With the patch now being tested via linux-next, it introduces the concept of a NUMA identity node sched domain to address the potential for improper load balancing. All the details within this patch that's pending for Linux 4.15.
If you haven't already, be sure to check out our many AMD EPYC Linux benchmarks. More are on the way.
| 22
| 1,760,719,683.377816
|
https://www.phoronix.com/news/AMDKFD-More-For-Linux-4.15
|
AMDKFD For Linux 4.15 Adding Usermode Events, Dropping Radeon DRM Support
|
Michael Larabel
|
Building off an earlier update in DRM-Next of upstreaming more AMDKFD changes for Linux 4.15, a second batch of feature work was proposed today for merging into DRM-Next.
Oded Gabbay has sent in a second batch of AMDKFD driver changes for Linux 4.15. AMDKFD, of course, being the HSA kernel driver. With this second batch of changes, it still is mostly focused on upstreaming APU work and not yet the discrete GPU bits needed.
The main additions with this pull request are usermode events and dropping Radeon DRM support. The usermode events support is cleaned up and ready for mainline. The Radeon DRM support from AMDKFD is being dropped since Kaveri APUs are "fully supported" with the AMDGPU DRM driver and current/future versions of the HSA user-space libraries will mandate the use of the AMDGPU DRM driver. Because of this, the Radeon DRM interface support is being dropped from AMDKFD and users are encouraged to migrate to using AMDGPU. See our recent AMDGPU vs. Radeon DRM graphics tests if you are behind on the DRM driver situation.
There are also bug fixes and cleanups as outlined in the AMDKFD -next pull request. Hopefully for Linux 4.16 we see more of the AMD changes upstreamed, especially for dGPU support.
| 6
| 1,760,719,683.715358
|
https://www.phoronix.com/news/AMDGPU-PRO-17.40-Stable
|
AMDGPU-PRO 17.40 Linux Hybrid Driver Promoted To Stable
|
Michael Larabel
|
Two weeks ago AMD released an AMDGPU-PRO 17.40 driver intended for cryptocurrency mining systems while now that v17.40 series driver has been promoted to being their general purpose stable Linux hybrid driver.
The headline feature of the AMDGPU-PRO 17.40 beta was introducing large page support intended to help blockchain compute workloads with OpenCL. Sure enough, AMDGPU-PRO 17.40 is much faster for cryptocurrency mining like Ethereum though we haven't found many performance improvements in other OpenCL compute workloads.
This large page support remains the primary feature being advertised for today's 17.40 stable driver. The release notes also mention there are stability fixes for Indigo Bench, Blender rendering fixes, and a FireRender problem. There aren't any other mentioned new features or other prominent fixes in this AMDGPU-PRO 17.40 update, which is marked as amdgpu-pro-17.40-492261.
The amdgpu-pro-17.40-492261 update only officially supports RHEL/CentOS 7.3, RHEL/CentOS 6.9, Ubuntu 16.04.3 LTS, and SLED/SLES 12 SP2. There is not yet any support for RHEL 7.4 nor Ubuntu 17.10 (as AMD now focuses their hybrid driver just on Ubuntu LTS releases, but should come indirectly once Ubuntu 16.04.4 support is eventually done).
This latest AMDGPU-PRO 17.40 Linux hybrid driver is available for download from AMD.com.
| 3
| 1,760,719,685.21277
|
https://www.phoronix.com/news/GCC-8.0-znver1-20171030
|
Running Some Fresh GCC 8.0 Compiler Benchmarks On AMD EPYC With "znver1"
|
Michael Larabel
|
As SUSE has been working in conjunction with AMD on more tuning for AMD Zen CPUs under the GCC compiler, here are some fresh benchmarks of the GCC 8 compiler code being tested on an AMD EPYC system.
For looking at the "znver1" improvements and general work that's ongoing for GCC 8 as the next annual GNU Compiler Collection update due out in 2018, I ran some fresh benchmarks of this GCC 8 compiler code after building it from source yesterday using its latest SVN/Git tip. The GCC 7.2 compiler was then built from source and configured in the same manner for showing how GCC 8.0 is currently comparing to GCC 7.2.
Ubuntu was running on this TYAN server with AMD EPYC 7601 during testing. During the Phoronix Test Suite testing process, the CFLAGS/CXXFLAGS were set to "-O3 -march=znver1" when testing both GCC 7.2 and GCC 8.0 SVN.
FFTW is running slightly faster with GCC 8.
SciMark's Monte Carlo is significantly faster now under GCC 8. Additional follow-up benchmarks will look more closely if it's due to znver1 optimizations or general improvements to the compiler code in GCC 8.
In most tests, the gains are quite small but measurable in a number of scenarios.
C-Ray seems to frequently become faster as a result of GCC optimizations.
LAME MP3 encoding is slightly faster according to the Phoronix Test Suite.
As outlined via this result file there were also some cases where GCC 8 isn't currently any different than GCC 7. More compiler benchmarks forthcoming on different AMD Zen processors as well as other compiler tuning tests in general as well as similar treatment under LLVM Clang.
| 9
| 1,760,719,686.025648
|
https://www.phoronix.com/news/AMDGPU-DC-25-Okt-Patches
|
New AMDGPU DC Patches Published, More Work Towards FreeSync
|
Michael Larabel
|
AMD's Linux team working on the AMDGPU DC display code sent out a set of 29 more patches this week.
Due to the DRM-Next cutoff for new feature material being this week and Alex Deucher previously indicating the final 4.15 pull of their Radeon work, these latest AMDGPU DC patches are most likely early material that will then arrive with Linux 4.16. But the initial AMDGPU DC display code is still on track for debuting with Linux 4.15, should Linus Torvalds not have any objections.
These 29 additional AMDGPU DC patches touch around one thousand lines of code and have a diverse range of changes. The work includes removing an "annoying" FreeSync warning, fixing FreeSync and amd-stg that had regressed in recent patches, fixing an issue when plugging in displays during the S3 suspend mode, many fixes found during the Raven Ridge bring-up, and various other Raven Ridge fixes and improvements.
These latest AMDGPU DC patches can be found here. It will be interesting to see how the Raven Ridge Linux support is once the laptops begin shipping in the weeks ahead.
| 3
| 1,760,719,686.628519
|
https://www.phoronix.com/news/AMD-NPT-GPU-Pass-Through
|
AMD/Ryzen NPT Fix Discovered For Better Pass-Through Graphics Performance
|
Michael Larabel
|
One area where AMD Ryzen users have encountered Linux issues with virtualization is when trying to setup pass-through support for a graphics card to allow the virtual machine direct access to the GPU. When NPT (Nested Page Tables) are enabled, performance can become severely degraded.
GPU/PCI pass-through problems have affected the small number of Ryzen Linux users trying to setup such a configuration, mostly for gaming, when NPT is enabled. Some have thought it was a hardware bug, etc, but the good news is a fix is in the works.
A Phoronix reader pointed out this mailing list post by Geoffrey McRae, "I have identified the issue! With NPT enabled I am now getting near bare metal performance with PCI pass through. The issue was with some stubs that have not been properly implemented. I will clean my code up and submit a patch shortly. This is a 10 year old bug that has only become evident with the recent ability to perform PCI pass-through with dedicated graphics cards. I would expect this to improve performance across most workloads that use AMD NPT."
With his pending fix for the Linux IOMMU code, there's nearly a 5x improvement in graphics performance when doing PCI pass-through of his GeForce GTX 1080 Ti paired with a Ryzen 7 processor.
| 31
| 1,760,719,687.660093
|
https://www.phoronix.com/news/Ryzen-Mobile-Today
|
Looking Ahead To AMD Ryzen Mobile On Linux
|
Michael Larabel
|
Following AMD on Twitter teasing new Ryzen announcements the past few days, today is expected to be the launch day for the new Ryzen Mobile hardware up to now known as "Raven Ridge".
We weren't briefed ahead of time on the Ryzen Mobile launch and are not under any NDA for today's event. We do know that it's the Ryzen Mobile 5 2500U and Ryzen Mobile 7 2700U launching today and they are what was known as the "Raven Ridge" codename up until now with these mobile parts featuring Zen CPU cores and Vega graphics.
Today's announcement also comes just days after HP accidentally slipped out their first Raven Ridge design in the form of the ENVY x360.
These new Mobile Ryzen parts are said to offer up to twice the CPU power of AMD 7th Gen APUs and 128% faster graphics performance while up to 50% less power.
Windows numbers show the new mobile Ryzen 7 chip competing with Intel's latest Core i7 8550U on the processor front. On the graphics front, however, the Raven parts are running much faster than Intel's latest UHD Graphics. Hopefully this will motivate Intel to release some 8th Gen chips with Iris Graphics...
Expectations from the Linux front are that AMD's new mobile hardware should play fine on recent Linux distributions from the CPU front, just as we've seen for the past few months with all our Zen-based processor testing from Ryzen 3 to EPYC. The area that's a bit more sticky are the onboard Vega graphics. Here you'll either need to be using the AMDGPU-PRO driver or running Linux 4.15+. Not to be confused with the upcoming 4.14 kernel in a few weeks, only with Linux 4.15 is the AMDGPU DC display code for supporting Vega graphics and also proper Raven Ridge display support and many Raven-specific fixes. The Linux 4.15 kernel won't be officially released until roughly the end of January. Of course, if going the mainline Linux / open-source route, you'll want to be using the latest Mesa+LLVM code too, if not using AMDGPU-PRO. With Linux 4.15 is also when the Zen temperature reporting is in place as tacked onto the k10temp kernel driver.
It will be interesting to see what Ryzen Mobile laptop designs come from the prominent vendors in the weeks ahead... I know there are many Phoronix readers that have been waiting for this long-awaited Zen+Vega APU. With that said, I've been saving up, so do hope to pick up a Raven Ridge laptop once finding a model from Lenovo/ASUS/Razer that's interesting and to use as my next main production system. So I will be delivering Ryzen Mobile Linux benchmarks in time as it doesn't look like any review samples are on the way, but unless you want to be using AMDGPU-PRO or -next code for pre-4.15 code, you won't want to rush out picking up a new laptop quite yet.
| 34
| 1,760,719,688.153737
|
https://www.phoronix.com/news/Feral-RADV-AMD_shader_info
|
Feral Adding AMD_shader_info To RADV Vulkan Driver
|
Michael Larabel
|
As further sign of Feral Interactive continuing to pursue Vulkan for their Linux games, a Feral developer today posted a patch for implementing the brand new AMD_shader_info extension for the RADV Mesa driver.
Alex Smith of Feral posted the patch today wiring in AMD_shader_info. This AMD-developed shader information extension debuted just days ago with the Vulkan 1.0.64 update. This Vulkan extension provides a means of querying information about a compiled shader from the shader disassembly to statistics, but the extension itself places no mandates about what is exposed.
With Feral's RADV VK_AMD_shader_info support, it dumps various shader statistics like the number of spilled registers, code size, waves, scratch memory used, and more. It also is able to provide the disassembly of the shaders.
Alex mentions that this extension support works in conjunction with the latest development code for RenderDoc, a popular Vulkan debugging application, to then easily see the shader disassembly of a capture. This should help a lot for allowing game/app developers like Feral Interactive to be able to quickly analyze the disassembly output when trying to debug a problem.
Great to see all the work continuing to build up around RADV but a shame AMD hasn't yet opened up any more of their own Vulkan driver.
| 16
| 1,760,719,689.198788
|
https://www.phoronix.com/news/AMDGPU-DC-Last-4.15-Batch
|
AMDGPU DC Gets A Final Batch Of Changes Before Linux 4.15
|
Michael Larabel
|
The AMDGPU DC display code has a final batch of feature updates that were sent in this weekend for DRM-Next staging and is the last set besides fixes for the "DC" code for the 4.15 target.
This latest AMDGPU DC update has 13,902 lines of new code and 6,076 lines of code removed. This now bumps the AMDGPU DC code size to over 130k lines of code just for this new display stack that provides Vega/Raven display support, atomic mode-setting, HDMI/DP audio, and other modern display features including the groundwork for FreeSync.
This newest batch of AMDGPU DC work includes memory leak fixes, S3 fixes, hotplug fixes, DisplayPort Multi-Stream Transport (DP MST) fixes, DML display updates for Raven Ridge, code clean-ups, and other bug fixes.
This latest batch of AMDGPU DC changes can be found via dri-devel. Now all that's left is to hope Linus Torvalds approves of merging AMDGPU DC for the upcoming Linux 4.15 merge window.
| 14
| 1,760,719,689.521591
|
https://www.phoronix.com/news/ZenStates-Linux
|
ZenStates Allows Adjusting Zen P-States, Other Tweaking Under Linux
|
Michael Larabel
|
ZenStates is an independent effort to offer P-States-based overclocking from the Linux desktop of AMD Ryzen processors and other tuning.
ZenStates-Linux is an open-source Python script inspired by some available Windows programs for offering Ryzen/Zen CPU overclocking from the desktop by manipulating the performance states of the processor.
The zenstates.py program allows dynamically toggling P-State, listing available P-States of the processor, setting the FID/DID/VID voltages, enabling C-State C6 mode, and other tunables.
ZenStates works by reading/writing to the Zen-specific model-specific registers (MSRs). As such, the small Python program must be run as root and requires the MSR kernel module be loaded.
Try out ZenStates-Linux at your own risk. The code is hosted via GitHub.
| 18
| 1,760,719,691.012749
|
https://www.phoronix.com/news/AMDGPU-4.15-Ctx-TTM-PP
|
More AMDGPU Changes Queue For Linux 4.15
|
Michael Larabel
|
Adding to the excitement of Linux 4.15, AMD has queued some more changes that were sent in today for DRM-Next.
Already for Linux 4.15, the AMDGPU Direct Rendering Manager driver should have the long-awaited "DC" display stack that brings Vega/Raven display support, HDMI/DP audio, atomic mode-setting and more. Other pull requests have also brought in a new ioctl, UVD video encode ring support on Polaris, transparent huge-pages DMA support, PowerPlay clean-ups, and many fixes, among other low-level improvements.
In what is likely the last AMDGPU feature pull request to DRM-Next for Linux 4.15 are a few more changes. The latest work includes a buffer object flag to let buffers opt-out of implicit synchronization, a context priority setting interface, more PowerPlay clean-ups, vRAM work related to GPU reset, TTM memory management support for huge-pages, and various other fixes.
Great to see all the work going on recently around huge-pages. Also worth noting is the context priority setting interface: this is the effort led by Valve in trying to get priority scheduling for the AMDGPU driver as part of their effort for ensuring good AMDGPU performance for virtual reality (VR) use-cases.
This AMDGPU priority scheduling has been in the works since last year and great to see the Valve Linux developers moving ahead and in the right direction for improving Linux VR gaming support/performance.
The latest batch of AMDGPU changes for DRM-Next can be found listed on amd-gfx.
| 11
| 1,760,719,691.07665
|
https://www.phoronix.com/news/AMDKFD-Linux-4.15-Changes
|
AMDKFD Preps More Carrizo/Kaveri Code For Linux 4.15 While dGPU Lags
|
Michael Larabel
|
The AMDKFD kernel driver that is a component of HSA support on Linux for Radeon GPUs is seeing more upstreaming work in Linux 4.15, but only for older APUs.
Recently AMD has been focused on upstreaming more of their changes to this HSA kernel driver. Initially it's been focused on Carrizo/Kaveri APU support and with Linux 4.15 it's continued that way.
There are improvements around suspend/resume, process termination, and various other code clean-ups. But Oded Gabbay mentions in the pull request to DRM-Next, "The patches here are relevant only for Kaveri and Carrizo. Still no dGPU patches."
The pull request can be found on dri-devel. Hopefully it won't be too much longer before seeing more of the discrete GPU support land as well as focusing on Raven Ridge.
| 4
| 1,760,719,692.596246
|
https://www.phoronix.com/news/Radeon-GPU-Profiler-1.03
|
Radeon GPU Profiler 1.03 Released
|
Michael Larabel
|
AMD's GPUOpen initiative has put out a new release of their Windows and Linux supported Radeon GPU Profiler program for profiling Vulkan (and Direct3D 12) games.
Radeon GPU Profiler went public over the summer as a means of helping Vulkan game developers optimize their titles, at least as far as the AMD Vulkan driver is concerned as well as D3D12 on Windows. The Radeon GPU Profiler works on Polaris/Fiji/Vega GPUs and provides low-level GPU timing data on barriers, events, pipeline state, stalls, and more.
Radeon GPU Profiler 1.03 is the new version out today and it exposes more system information but that is currently Windows-only and aside from that are various user-interface and user experience improvements. A headless version of the Radeon Developer Service binary was also added along with various bug/stability fixes.
More details on Radeon GPU Profiler 1.03 can be found via GPUOpen.com. The project continues to be hosted on GitHub but as of writing the program's source code remains not yet publicly available and thus just the Windows/Linux binaries.
| 3
| 1,760,719,692.607481
|
https://www.phoronix.com/news/Linux-4.14-AMD-CPB
|
Linux 4.14 Ensures The "Core Performance Boost" Bit Gets Set For AMD Ryzen CPUs
|
Michael Larabel
|
Recently making waves in our forums was talk of a kernel patch to address a case where the AMD CPB (Core Performance Boost) isn't being exposed by Ryzen processors. Here's more details on that and some benchmarks.
Being talked about recently is f7f3dc0: "CPUID Fn8000_0007_EDX[CPB] is wrongly 0 on models up to B1. But they do support CPB (AMD's Core Performance Boosting cpufreq CPU feature), so fix that."
Basically the bit for indicating Core Performance Boost support indicates it's disabled for Ryzen CPUs while it should be enabled for hitting the boost frequencies on these processors. This patch isn't for Linux 4.15 but was already merged as a fix for Linux 4.14.
I confirmed the change with a Ryzen 7 1800X that is stepping 1 and indicated from /proc/cpuinfo via the flags line you can look for the presence of "cpb." (For reference, EPYC is stepping 2 and thus unaffected by this error.) Booting to Linux 4.13, the CPB string wasn't shown but is in fact reported when booting to the Linux 4.14 kernel.
Via the Ryzen 7 1800X setup I then proceeded to run some benchmarks to see if the performance was any difference with the new kernel:
But with a range of tests run, there isn't any real performance change to find out of the variety of tests executed.
| 17
| 1,760,719,694.43297
|
https://www.phoronix.com/news/AMDGPU-PRO-14.40-Mining
|
AMD Releases AMDGPU-PRO Beta Driver For Mining / Compute Customers
|
Michael Larabel
|
AMD this week has quietly released an updated AMDGPU-PRO 17.40 hybrid driver targeting mining and GPGPU compute Linux customers.
The sole listed change with this AMDGPU-PRO 17.40 update is large page support for blockchain compute workloads. This is advertised as the "AMDGPU-PRO Beta Mining Driver 17.40" driver.
The release page mentions, "Introduces Large Page Support for Blockchain Compute Workloads. This driver is provided as a beta level support driver which should be considered "as is" and will not be supported with further updates, upgrades or bug fixes. This driver is not intended for graphics or gaming workloads."
While the driver has large page support, it requires also booting the kernel with e.g. "amdgpu.vm_fragment_size=9" for 2MB page support.
Those mining cryptocurrencies like Ethereum rely upon the AMDGPU-PRO driver when running on Radeon GPUs due to the newer ROCm OpenCL compute stack not yet fully working against upstream open-source components (e.g. the Linux kernel and LLVM) while the open-source Clover OpenCL stack isn't good enough for a majority of compute workloads. Hopefully in the months ahead ROCm will begin working on more upstream components.
| 5
| 1,760,719,695.336358
|
https://www.phoronix.com/news/HP-ENVY-x360-Raven-Ridge
|
HP Rolls Out The First "Raven Ridge" Zen+Vega APU Notebook
|
Michael Larabel
|
AMD has announced the world's first "Raven Ridge" APU with this notebook being powered by Ryzen 5 CPU cores paired with Vega graphics.
The HP ENVY x360 15-bq101na is this first notebook featuring the Zen CPU cores paired with Vega 8~10 graphics. The HP ENVY x360 15-bq101na is listed as having a Ryzen 5 2500U quad-core processor with 2.0GHz base frequency and 3.6GHz boost frequency. On the graphics side are integrated "AMD Radeon Vega M" graphics. This notebook also has 256GB of NVMe storage, 8GB DDR4-2400 memory, and ships with Windows 10. The power adapter is 45 Watts and at least under Windows is expected to offer up to a ten hour battery life.
More details via this data sheet. Pricing and availability details not yet available. Once more Raven Ridge hardware becomes available, I hope to get access to some for Linux testing. The Raven Ridge testing has been up and coming already within the AMDGPU code the past few cycles: Raven Ridge does require AMDGPU DC support so you'll be looking for a Linux 4.15+ based kernel.
| 32
| 1,760,719,695.827134
|
https://www.phoronix.com/news/AMD-GCC-More-Zen-Patches
|
More AMD Zen Tuning Patches Posted For GCC
|
Michael Larabel
|
A few days back I initially wrote about a SUSE developer working on Zen tuning patches for GCC. That work has continued with more compiler patches coming for optimizing the GNU's compiler for Ryzen / Threadripper / EPYC processors.
Jan Hubicka of SUSE has continued posting more patches for tuning Zen for GCC and underlying code improvements for GCC in dealing with CPU-specific optimizations.
Among the work includes simplifying some code, avoiding 512-bit memcpy/memset for AVX-128 optimal targets, disabling Bulldozer dispatch scheduling, and better managing around CPU specific optimizations.
I'll be running some fresh GCC benchmarks on Zen via Ryzen or EPYC shortly, once seeing what other patches Hubicka might land in the next few days. This work is being queued in GCC 8, which will be released as stable in H1'2018 as GCC 8.1.
| 6
| 1,760,719,696.830262
|
https://www.phoronix.com/news/AMD-EPYC-More-Epic-Benchmarks
|
More AMD EPYC Processors Arrive For Linux Performance Testing
|
Michael Larabel
|
Yesterday a new batch of AMD EPYC processors arrived for testing at Phoronix.
Thanks to AMD for sending out some additional EPYC processors for Linux testing as well as TYAN for their platform support.
To complement our existing AMD EPYC testing with the EPYC 7601 is now the:
EPYC 7251 - The lowest-end EPYC part an 8 core / 16 thread server processor for just under $500 USD while having a 2.1GHz base frequency and 2.9GHz peak frequency. This is the lone EPYC CPU right now having only DDR4-2400 support rather than DDR4-2666, but still supports eight memory channels.
EPYC 7351P - The 7351P is a $750 USD CPU that boasts 16 cores / 32 threads, 2.4GHz base frequency, and 2.9GHz boost frequency. There's a lot of value with the 7351P especially in comparison to the current Xeon Scalable line-up.
EPYC 7401P - The EPYC 7401P boasts 24 cores / 48 threads with 2.0GHz base frequency and 3.0GHz maximum boost frequency. This CPU goes for just over $1000 USD and should be another interesting test candidate.
EPYC 7551 - A step below the EPYC 7601. This $3400 USD processor is a 32-core / 64-thread CPU while having a base clock of 2.0GHz and clocks up to 3.0GHz. Like the 7601, it's a 180 Watt TDP part, 64MB L3 cache, etc.
Full reviews and benchmarks on these four more EPYC processors will be published in the days ahead plus many other interesting EPYC Linux tests (and I happen to be finishing up EPYC 7601 BSD tests right now as well) over the coming weeks.
Part of the purpose of this notice is also to solicit any other test/workloads you'd like to see benchmarked with these EPYC processors. Aside from the many benchmarks ran already, if there are any other workloads of interest to you, be sure to let me know right away. If they are not already part of the tests on Openbenchmarking.org, be sure to link / provide reference to any documentation for being able to run the tests in an automated, default sane manner... Or ideally a bash script already to start from, considering I am not a domain expert on everything as well as being a time saver, any helpers are useful in being able to ensure coming up with good real-world tests. Basically if the test is multi-threaded and can easily be run in a standardized manner, happy to add it to the test queue for these processors. Or any other EPYC test requests, feel free to post away on the forums or via Twitter; of course, if you are a Phoronix Premium supporter, you should already know my email to personally contact me with any benchmark requests and feedback.
| 16
| 1,760,719,697.372125
|
https://www.phoronix.com/news/Vega-TR-Ethereum-Mining
|
Mining Ethereum With AMD Threadrippers Paired With Four RX Vega 64 GPUs
|
Unknown
|
Phoronix reader Thomas Frech has shared with us an article he wrote about his new Ethereum mining work on two systems using AMD Threadripper processors and a total four Radeon RX Vega 64 GPUs under Linux. In this article I will write about my new protect to build 2 dual-use systems means Desktop+Ethereum-Mining.
If you do not have time to read the full article here the short version in one sentence: No, it is not practical to run Ethereum on AMD Vega 64 graphics cards on Linux because the ROCm OpenCL stack in the AMDGPU-PRO 17.30 is very slow in Ethereum so the whole system with four Vega64 cards make 16-23Mh/s and even a single old AMD-RX470 with the old Closed-source OpenCL AMDGPU-PRO stack run ~20+ Mh/s... This all could be the effect of the exponential function Ethereum-ICE-AGE Bomb...
But now the long version:
I ran into unexpected building problems in the time of build this 2 systems. The ASRock X399 Taichi has two short screws on the three-screw CPU socket this results in the problem that I was unable to screw in the screws by hand. So I needed a electric drill and relatively high pressure to get the first traction and then I used the hand-screw tool from the Threadripper 1900X/1920X CPU. The MSI X399 gaming pro carbon instead had a perfect mechanical socket and I was able to use the hand screw tool to mount the CPU perfectly. So the clear win in mechanical quality goes to MSI.
Both mainboards have WiFi included but only the ASRock X399 Taichi has onboard WLAN and the MSI does only have a WLAN PCIE card but if you use all four PCIE 16 slots with dual-high AMD Vega cards the result is that the MSI is without wireless because there is no space left to install the PCIe WiFi card. So ASRock is the clear WiFi/wireless/WLAN winner.
I ran into big problems because of the BIOS version of the boards: The 1900X was released one month after the 1920+1950 so you need a more expensive CPU to install the BIOS updates because 1900X is only supported in later revisions. Also I ran into problems because the AMD-VEGA is also not supported by earlier BIOS versions and also 2/3 of the RAM DIMMS I bought where also not supported and I even bought a more expensive 1920X cpu and three different kinds of RAM DIMMs to perform a BIOS update. Atfer the updates, all the RAM DIMMs worked fine as well as the VEGA GPU and 1900X CPU. I did not even try to make the same on the ASRock so O bought the ASROCK with Bios-update from the store.
This means if you are a beginner and you do not want to run into "BIOS" incompatibility problems it is highly recommended to buy the mainboard with pre-installed bios updates from the store.
The X399 mainboards all look like you can use four AMD Vega cards but they run into thermal overheating problems because the space between the cards in the middle is very tiny and even with mechanical tricks to expand the space between the cards I lost ~20-30% mining power of these expensive Radeon Vega cards. AMD really needs to bring mining edition cards to the market with passive copper cooling blocks without monitor HDMI plugs to make it possible to install external cooling fans to push the heat out of the PC-case.
I also ran into problems because the ASRock Taichi has no M.2 SSD slot screws and because of this I had to buy these screws for 7,99€ on eBay and the MSI X399 board had a perfect set of screws for every M.2 SSD slot on the mainboard. So the MSI beats the ASRock if you want to use the M.2 SSD slots.
The ROCm Stack in the AMDGPU-Pro 17.30 driver is very slow: a AMD VEGA 64 has results like 40Mh/s on the old Closed-source Catalyst/FGLRX/AMDGPU driver but now the ROCm stack is the new default for Linux installs on the LTE 16.04 ubuntu releases. I only needed like 2 minutes of clicking around with this (irony)quality(/irony) closed source driver then the KDE-info-center crashes with segmentation faults.
The 17.30 driver had no working default OpenCL because it was optional to install but the installer did not show any information about this optional openCL driver part. But AMD's John Bridgman wants to bring this info into the installer of the AMDGPU-pro driver installer or fix the default that the users get a working OpenCL after the install:
"I don't know why compute/OpenCL was left as optional but am asking around. Not sure we will be able to change that quickly but if not then at least I want to see us get a big message at the top of the install instructions noting that OpenCL/compute is not installed by default but that instructions are further down the page." Bridgman
But for sure this extra information and steps bring the OpenCL ROCm stack to work:
Installing the Optional ROCm Component This AMDGPU-Pro driver package incorporates the ROCm component that can be optionally installed for running Compute/OpenCL applications. You can install the component by issuing the following command:
sudo apt install -y rocm-amdgpu-pro
Configuring the Optional ROCm Component
The LLVM_BIN environment variable needs to be set prior to running ROCm applications.
To set it temporarily when running an individual ROCm command, such as clinfo, use:
env LLVM_BIN=/opt/amdgpu-pro/bin /opt/amdgpu-pro/bin/clinfo
To set it permanently for all bash and other sh-like shell users, you can use the following command:
echo 'export LLVM_BIN=/opt/amdgpu-pro/bin' | sudo tee /etc/profile.d/amdgpu-pro.sh
See the Ubuntu Environment Variables Community Help for more information. More details via AMD Support.
Now comes the good part I asked for support here and the AMD-Linux support by Bridgman is legendary good.
And it really looks good for the future so we will get a closed-source OpenCL-only installer for the RadeonSI FOSS driver this should fix the segmentation fault of the KDE infocenter because with RadeonSI all desktop features work perfect:
"[...]but if you mean "install openCL and use it alongside radeonsi" then yes we are working on that.
It will be packaged and tested with the DKMS/KCL kernel driver (copied & pasted that last word) rather than upstream but we are also trying to support newer kernel/X releases so should work OK. That won't be in 17.40 though... ", wrote Bridgman.
According to Bridgman the performance problems should be fixed by the AMDGPU-PRO 17.40 driver at the end of this month:
"Initial feedback is that latest ROCm (1.6.3) has the performance fixes, but there will be a 17.40 release soon which also includes them. We are trying to arrange an early release of 17.40 which should be what you need. If you check airlied's drm-next branch you will see kfd commits starting to show up frequently; IIRC that goes into 4.15. We're going to try to also get that backported to upcoming LTS and enterprise distro releases."
As we can see AMD works on upsteaming the kernel features into the future 4.15 Linux kernel. The two performance relevant features will be 2MB pages and memory interleaving HBM banks
I do not know yet how much performance we will earn from these performance optimizations. But he also still wait for AMD staff to get to know what is the recommended driver stack for Ethereum:
"I'll find out what driver stack we are recommending for ethminer - IIRC we were seeing performance drops with latest workloads and made a couple of driver changes to address them. One was use of 2MB pages (which is not in 17.30 IIRC) and another was related to memory interleaving between HBM banks... each of them made a noticeable difference. The answer might be "use the ROCm stack in the short term" but will find out." Bridgman
A funny side note is that I used a 64-bit Windows 10 installation DVD to try Windows-mining too but my MSI X399 system was unable to start the windows 10 installer DVD. so if you think the Ethereum 6Mh/s from the ROCm Stack on linux is slow think about the 0 Mh/s on Windows! ;)
Also remember that Mega Hashrates on Ethereum is not always the same because of the Ethereum Ice Age mechanism according to this source:
"The Ethereum Ice Age is a difficulty adjustment scheme that was put in place to ensure that everyone has an incentive to move to the new blockchain once the hard-fork is implemented," from here.
This means on 18 October 2017 is the reset of the ICE-AGE bomb then we will see normal results again. Right now the difficulty adjustment goes very high because the exponential function ICE-AGE Bomb...
This mechanism is there to force the miners into using the newest software to make sure their development of Ethereum is unified.
Because of this I will write a second article after the 18.10-2017 Ethereum ICE-AGE-BOMB is disarmed then I will do some Energy bill Measurements and I will bring fresh numbers of the 17.30 vs 17.40 drivers.
Right now according to the Ethermine.org pool statistic my MSI mining PC show that my PC would make 40-50dollar a month if running 24 hours a day right now it only run at day time and I turn it off for sleeping. This means the ~4000€ PC would never pay out with 17.30 ROCm and ICE-AGE bomb...
Also the CPU usage is 0% so the 1920X Threadripper 12-core CPU brings no benefit over a 1900X 8core cpu. Right now I also use only a single 16GB RAM DIMM but the PC use less than 4GB RAM.
To be continued...
| 47
| 1,760,719,698.329123
|
https://www.phoronix.com/news/AMD-Hiring-For-ROCm
|
AMD Is Hiring LLVM Compiler Engineers To Work On ROCm
|
Michael Larabel
|
AMD is currently looking to hire more LLVM compiler engineers to work on their ROCm open-source compute stack.
The Radeon Open eCosystem Group is looking to hire more developers to work on ROCm with their OpenCL, HCC C++, and HIP compute efforts. Currently they are willing to look at all candidates regardless of experience level but need a background in compilers, C/C++ programming, parallel programing, and related fields.
Those potentially interested in joining AMD to work on open-source GPGPU efforts can see this mailing list post or head directly to jobs.amd.com.
| 4
| 1,760,719,698.628223
|
https://www.phoronix.com/news/AMDGPU-Linux-4.15-1
|
AMD Submits Initial AMDGPU DRM Updates Slated For Linux 4.15
|
Michael Larabel
|
Alex Deucher has submitted the initial AMD Direct Rendering Manager updates for pulling into DRM-Next that in turn will hit the Linux 4.15 kernel.
The feature work ready so far for queuing into DRM-Next includes per-VM buffer object support, many PowerPlay clean-ups, PowerPlay support for GCN 1.1 Sea Islands hardware, the "pasid" manager for the AMD KFD driver, interrupt infrastructure for recoverable page faults, SR-IOV fixes, initial GPU reset support for Vega 10, DRI PRIME mmap() support, TTM page table debugging improvements, and many bug fixes.
Great to see the continued improvements for Vega and the seemingly never-ending work on PowerPlay. It's also good that there is now PowerPlay support for GCN 1.1 Sea Islands support, another item to mark off the list before GCN 1.0/1.1 GPUs could theoretically default to the AMDGPU DRM in the future rather than Radeon DRM, but not for 4.15. And before anyone asks, there's no update yet on whether we will see the AMDGPU DC display code for Linux 4.15.
There's been no final call for review of DC yet while DC patches have continued flowing. We'll see in a few weeks whether this new display stack looks like it will be ready for merging in 4.15 or will be another kernel release without AMDGPU HDMI/DP audio, Vega display support, FreeSync, etc.
The list of AMDGPU changes so far for Linux 4.15 can be found via dri-devel. Given that 4.14-rc2 was just released this weekend, there's about four more weeks that new feature material will be accepted into DRM-Next before closing the window for 4.15.
| 18
| 1,760,719,699.809123
|
https://www.phoronix.com/news/Linux-4.14-ROCm-KV-CZ
|
Linux 4.14 + ROCm Might End Up Working Out For Kaveri & Carrizo APUs
|
Michael Larabel
|
It looks like the upstream Linux 4.14 kernel may end up playing nicely with the ROCm OpenCL compute stack, if you are on a Kaveri or Carrizo system.
While ROCm is promising as AMD's open-source compute stack complete with OpenCL 1.2+ support, its downside is that for now not all of the necessary changes to the Linux kernel drivers, LLVM Clang compiler infrastructure, and other components are yet living in their upstream repositories. So for now it can be a bit hairy to setup ROCm compute on your own system, especially if running a distribution without official ROCm packages. AMD developers are working to get all their changes upstreamed in each of the respective sources, but it's not something that will happen overnight and given the nature of Linux kernel development, etc, is something that will still take months longer to complete.
The good news for at least Carrizo and Kaveri APUs, the upstream Linux 4.14 kernel may end up being good enough for at least some ROCm OpenCL workloads. Former AMD compute developer Tom Stellard who is now working for Red Hat on open-source GPGPU efforts inquired about recent commentary by AMD's Felix Kuehling about getting ROCm working on upstream kernels.
Felix clarified that for now it's only Carrizo and Kaveri APUs. As well, you still need custom Thunk code and a simple hack is needed to the user-space code to get Kaveri working. The kernel ABI is still being worked out as part of the upstreaming process.
In a follow-up message Felix added that amd-staging-drm-next is working as the "upstream" source while the necessary patches are also working their way into Linux 4.14.
Hopefully in time for Linux 4.15 we'll see more AMD ROCm-related upstreaming work.
| 12
| 1,760,719,700.087409
|
https://www.phoronix.com/news/AMD-CrossFire-mGPU
|
AMD Rebrands CrossFire As Just mGPU
|
Michael Larabel
|
While not particularly relevant to Linux gamers at this point in time, AMD is dropping their CrossFire branding in favor of just calling it their mGPU technology.
The mGPU term is just short for multi-GPU. They are dropping their multi-GPU "CrossFire" branding due to a shift in approach in how games make use of multiple GPUs. Traditional CrossFire, similar to NVIDIA's SLI, has relied upon game profiles for leveraging rendering across multiple graphics cards. That was with Direct3D 11 and prior (and OpenGL) while with Direct3D 12, the responsibility of multi-GPU handling is punted off to the developer with the D3D12 APIs for implicit and explicit multi-adapter support. As such, it's really not "CrossFire" anymore but just mGPU for future games.
Vulkan is similar to Direct3D 12 in that the responsibility of multi-GPU handling is left to the game developer via extensions for managing resources and rendering across multiple GPUs.
When it comes to CrossFire support on Linux, the RadeonSI+AMDGPU or AMDGPU-PRO stacks have not supported it in the traditional sense. With the former Catalyst/fglrx driver stack there was OpenGL CrossFire support, but it was largely useless and not many game profiles supported on Linux and so the benefits were minimal. But with the Vulkan support, there is said support, but it's a matter of when game developers begin adding multi-GPU Vulkan support to their renderers...
Back during the Radeon HD 4000 days, CrossFire worked on Linux with the notorious fglrx driver. There is nothing blocking anyone from adding OpenGL CrossFire support to the open-source AMD Linux graphics stack, but it will probably never come. AMD obviously doesn't have motivation now to add the support with Vulkan being around, the AMD developers are busy as-is with a ton of other projects, and it just doesn't make too much sense. Even NVIDIA SLI for Linux gaming isn't too practical.
Khronos added its initial multi-device support to Vulkan earlier this year with their v1.0.42 GDC update. These extensions are for the external memory, device groups, device group creation, etc.
It could be a good question for Linux games when comes the multi-GPU benefits given there aren't many Vulkan titles right now and game developers are more concerned about seeing the games work across the Linux drivers and performant than seeing it working for the small percentage of Linux gamers using multiple GPUs.
| 26
| 1,760,719,701.269765
|
https://www.phoronix.com/news/AMD-Zen-Temps-Hwmon-Next
|
AMD Zen Temperature Monitoring On Linux Is Working With Hwmon-Next
|
Michael Larabel
|
If you want CPU temperature monitoring to work under Linux for your Ryzen / Threadripper / EPYC processor(s), it's working on hwmon-next.
The temperature monitoring support didn't make it for Linux 4.14 but being published earlier this month were finally patches for Zen temperature monitoring by extending the k10temp Linux driver.
The work ended up being quite small for adding Zen temperature monitoring to the existing k10temp Linux kernel driver, but missed the Linux 4.14 merge window. This support should work for all AMD Family 17h processors.
Those k10temp patches by Guenter Roeck are now living within hwmon-next. That means it is being staged for merging then come the Linux 4.15 kernel cycle.
I tried hwmon-next today on my Threadripper 1950X system and indeed the temperature reporting is working. A single CPU temperature is reported for the package via k10temp (not sure under Windows if they support per-core temperature reporting?).
For those wanting to check on their Zen CPU temperatures under Linux, I uploaded this Debian/Ubuntu kernel built today from hwmon-next that in turn is currently based on the 4.14-rc1 kernel state. Use at your own risk, but should allow Ryzen/TR/EPYC temperature reporting to at least work.
| 23
| 1,760,719,701.451662
|
https://www.phoronix.com/news/AGESA-1.0.0.6b-Update
|
AGESA 1.0.0.6b Might Fix The Ryzen Linux Performance Marginality Problem
|
Michael Larabel
|
Motherboard vendors have begun pushing out BIOS updates for Ryzen motherboards using the AMD AGESA 1.0.0.6b revision and it's reported that it does resolve the "Performance Marginality Problem" affecting early Ryzen Linux customers.
While newer Ryzen CPUs are running great on Linux without the performance marginality problem as described by AMD, since yesterday I have begun receiving unconfirmed reports from Phoronix readers that the recent AGESA 1.0.0.6b revision does address the issue via this software update.
But unfortunately details on AGESA 1.0.0.6b are light. Most of the web pages / threads referencing AGESA 1.0.0.6b are actually inquiring about the changes to be found in this update, but few actual details on this update that has been surfacing in the past two weeks via BIOS updates.
After beginning to receive reader statements since yesterday that AGESA 1.0.0.6b addresses the Linux "performance marginality problem", I reached out to some AMD contacts when waking up this morning. Unfortunately, as of ending out the day, I have yet to receive any comment from any of them whether this update is expected to address the problem.
Since then as well I have received a comment from the reader of being able to reproduce the performance marginality problem under Windows. There is now the kill-ryzen-win script that also seeks to illustrate Ryzen issues when using Visual Studio C/C# compilation. But we'll see if the latest AGESA update takes care of that too or if hearing anything else from AMD.
| 78
| 1,760,719,702.742993
|
https://www.phoronix.com/news/AMDGPU-DC-September-11
|
AMDGPU DC Display Code Tacks On Another 28 Patches
|
Michael Larabel
|
The big undertaking of the rewriting/modernizing of the AMDGPU DRM driver's display code stack has out now another 28 patches.
This AMDGPU DC display stack has been well over one thousand lines of code and in development for years in trying to better synchronize the AMDGPU Direct Rendering Manager's display code with that of their Windows driver. AMDGPU DC is what's needed for HDMI/DP audio on modern Radeon GPUs, HDMI 2.0 support, atomic mode-setting, FreeSync, and other modern display features. More recently, it's now needed for driving physical displays/monitors attached to Radeon Vega graphics cards.
We're still waiting to see if AMDGPU DC will be ready for Linux 4.15 (it's already too late for Linux 4.14) or whether it's back to the long waiting game. We'll see in the next few weeks if Linux 4.15 looks realistic depending upon whether AMD sends out the massive patch series for another round of public review on dri-devel. As it stands now, we haven't seen any indication for or against of whether it's looking like it could be merged for Linux 4.15.
But today we did get 28 more patches to the display code touching around one thousand lines. These latest patches by AMD developers have a number of Raven Ridge fixes, no longer power down clock sources during the atomic check, removes some more abstractions, and fixes display unplugs during S3 or fbcon modes.
More details via this patch series.
| 60
| 1,760,719,703.891748
|
https://www.phoronix.com/news/Lenovo-ThinkPad-AMD-A-Series
|
Lenovo Announces New ThinkPads With AMD APUs
|
Michael Larabel
|
For the many of you Linux users that have been desiring an AMD laptop, things could get interesting with Lenovo having just announced the ThinkPad A-Series.
The ThinkPad A-Series is powered by AMD PRO hardware and initially consists of the ThinkPad A275 and A475. The A275 comes in at 2.9 lbs / 1.31 kg with a 12.5-inch display while the A475 is at 3.48 lbs / 1.57 kg with a 14-inch display. Both models offer up to 512GB NVMe SSD or 1TB HDD, USB Type-C, and other modern features.
But before getting too excited right now, the initial A275 and A475 models are powered by "7th gen AMD PRO A12 processors with AMD Radeon R7 integrated graphics." Yes, the Carrizo APUs of 2015 with Excavator CPU cores. A bit unfortunate considering the Raven Ridge launch with Zen cores and Vega graphics should just be months away. Presumably though this is Lenovo getting comfortable with AMD-powered ThinkPads and to test the waters: I definitely hope they will be timely releasing a Raven Ridge laptop and I may be interested myself in getting such a laptop.
For those interested in these 7th Gen AMD PRO laptops, the A275 pricing begins at $869 with shipping in October while the A475 starts at $849 and shipping this month. The laptops will be loaded with Windows, but at least in these current models being based off Carrizo PRO, they presumably should play nicely with modern Linux distributions.
More details at Lenovo.com.
| 37
| 1,760,719,704.770511
|
https://www.phoronix.com/news/Linux-4.14-Crypto-AMD-SP
|
AMD Secure Processor Support In Linux 4.14
|
Michael Larabel
|
The crypto subsystem updates have been pulled in for the Linux 4.14 kernel and it includes more complete AMD Secure Processor support, among other changes.
The mainline Linux kernel has already supported the AMD Cryptographic Coprocessor (CCP) for encryption, hashing, and other features. With Linux 4.14 the CCP support is still around along with the Platform Security Processor (PSP) device support. This support is still exposed via the CRYPTO_DEV_CCP Kconfig switch and new CRYPTO_DEV_SP_CCP while is now advertised as just "Support for AMD Secure Processor" rather than just "AMD Cryptographic Coprocessor."
Among the uses for the AMD Secure Processor are for use with the AMD Secure Encrypted Virtualization (SEV) for key management and for Trusted Execution Environments (TEE; there's a new TEE subsystem in Linux as of 4.12). Among the CCP crypto additions in Linux 4.14 is support for RSA and then for v5 CCP hardware is allowing RSA, XSTS-AES-128, and XTS-AES-256. Also queued up for Linux 4.14 as well is the AMD Secure Memory Encryption (SME) support as found on EPYC CPUs.
The AMD Secure Processor is built off ARM TrustZone technology with an ARM Cortex-A5 being embedded into recent APUs and CPUs. With the new AMD EPYC 7000 series there is an AES-128 engine support and the SME and SEV support, among other capabilities.
Other crypto work for the Linux 4.14 kernel include a STM32 HASH module, Microchip / Atmel ECC driver, Freescale RNGC hardware random number generator support, and other changes. The updated code is here.
| 20
| 1,760,719,705.650878
|
https://www.phoronix.com/news/Clang-6-SVN-Tripper
|
More Benchmarks Of AMD's Threadripper With LLVM Clang 6.0 SVN
|
Michael Larabel
|
With AMD a few days ago having landed an updated scheduler model for Zen CPUs within LLVM, I ran some fresh compiler benchmarks to see how the performance compares.
This weekend I ran tests of LLVM Clang 4.0, Clang 5.0, and Clang 6.0 SVN after the landing of the recent znver1 scheduler model update. Just some quick tests from the AMD Threadripper 1950X Linux system while a more formal compiler comparison is in the works.
And a quick look at the many C/C++ benchmarks used:
You can find all of the details in full from this OpenBenchmarking.org result file. There are some minor changes in performance in both directions with Clang 6, but nothing too major. LLVM Clang 6.0 will still be under development for about the next five months before it will be released as stable.
| 8
| 1,760,719,706.302811
|
https://www.phoronix.com/news/Zen-Thermal-4.14-Where
|
It Doesn't Look Like A Ryzen/EPYC Thermal Driver Will Make It For Linux 4.14
|
Michael Larabel
|
While the Ryzen CPUs have been available for a few months now and the higher-wattage Threadripper and EPYC processors are now available too, the Linux thermal driver remains missing in action and it's looking less likely that it will materialize for Linux 4.14.
The Linux 4.14 merge window should open this weekend, unless the Linux 4.13 cycle is unexpectedly stretched by an additional week.
The various subsystem maintainers are prepping their "-next" trees for Linux 4.14 merging, including hwmon maintainer Guenter Roeck. Unfortunately, there still is no Zen/Ryzen/Threadripper/EPYC thermal driver in the tree.
A glance in hwmon-next shows the material queued so far, including a Lantiq cpu temperature sensor support (used by the XRX200), but nothing on the AMD front.
There's also been no posting of any driver for review on the kernel mailing list yet. I'm still holding out hope we could maybe see one arrive for Linux 4.14, but last I heard from AMD contacts is they were still getting the necessary documentation out there.
For most users a Zen thermal driver isn't necessary unless you plan on overclocking your system, are concerned about the air cooling with Threadripper, compare cooling solutions (in fact, one of the reasons I've been waiting for this driver), or want to feel otherwise safe while running especially EPYC/Threadripper. Aside from the missing thermal sensor support, the Ryzen/Threadripper support has been working out well on Linux. It would be nice to see it too for Linux 4.14 considering this is a Long Term Support (LTS) kernel release likely to be used by Ubuntu 18.04 among other early 2018 Linux distributions.
| 14
| 1,760,719,707.06352
|
https://www.phoronix.com/news/AMDGPU-DC-77-More
|
77 More Patches For AMDGPU's DC Display Stack
|
Michael Larabel
|
Seventy-seven new patches were posted today for the AMDGPU DC (formerly DAL) display code that reworks around six thousand lines of this massive codebase.
AMDGPU DC or just "display code" as they now prefer calling it is a massive undertaking that isn't going to see the mainline kernel now until at least Linux 4.15 as it's too late to merge in DRM-Next for 4.14. We'll see in the next few weeks if the public review begins again to see if it's looking likely to happen for 4.15. We're growing increasingly anxious since this is the code that's blocking the display support for Radeon RX Vega for the mainline kernel, HDMI/DP audio for newer GPUs, HDMI 2.0 features, FreeSync, and more.
These newest AMDGPU DC patches rework some parts of the driver to be more like the other atomic DRM drivers, flatten out some more of their abstracted code, reworking hardware object code, Raven Ridge APU fixes, and various other fixes.
The long list of these additional changes here. This new display stack is well over one hundred thousand lines of code.
| 13
| 1,760,719,707.797417
|
https://www.phoronix.com/news/Znver1-Scheduler-Update-LLVM6
|
Updated AMD Zen Scheduler Model Lands For LLVM 6.0
|
Michael Larabel
|
With the soon-to-be-released LLVM 5.0 there is the initial AMD Zen scheduler model for the compiler to benefit Ryzen / EPYC processors. But now already hitting the LLVM development code for LLVM 6.0 is a revised scheduler model.
AMD's Ganesh Gopalasubramanian posted an updated scheduler model for "znver1", the first-generation AMD Zen processors. Today it's now been merged in the development code to be released as part of LLVM 6.0 in another six months. Changes mentioned include:
1) Regex based Instruction itineraries for integer instructions.
2) The instructions are grouped as per the nature of the instructions (move, arithmetic, logic, Misc, Control Transfer).
3) FP instructions and their itineraries are added which includes values for SSE4A, BMI, BMI2 and SHA instructions. It looks like I'll be working on some updated Ryzen/Threadripper LLVM Clang 4/5/SVN benchmarks this weekend and possibly some fresh GCC numbers as well.
| 15
| 1,760,719,708.715948
|
https://www.phoronix.com/news/AMD-Threadripper-1900X
|
AMD Rolls Out The Threadripper 1900X: 16 Thread, 4.0GHz Boost
|
Michael Larabel
|
Following the rumors of an eight-core / sixteen-thread Threadripper, the 1900X is now officially available beginning today.
While the Threadripper 1950X has 16 core / 32 threads and the 1920X has 12 core / 24 threads, the Threadripper 1900X has just eight cores and sixteen threads. The 1900X has a 3.8GHz base frequency and 4.0GHz boost frequency.
You may be wondering why there is a Threadripper with just 8 cores / 16 threads when that is the same count as the Ryzen 7 series... It's namely for those wanting to use the X399 platform for its 64 PCI Express lanes and quad-channel DDR4 support. The Threadripper 1900X will come in at $549 USD while the 1920X remains at $799 USD and the 1950X at $999.
More details on the new chip at AMD.com. In case you've missed it, we've been able to deliver a number of 1950X Threadripper Linux benchmarks with more on the way.
| 78
| 1,760,719,709.406128
|
https://www.phoronix.com/news/LLVMpipe-SWR-Threadripper
|
LLVMpipe & OpenSWR OpenGL Riding Off Threadripper
|
Michael Larabel
|
One of the unique test requests coming in as part of our Threadripper on Linux testing is to see how well the LLVMpipe and OpenSWR CPU-based OpenGL implementations within Mesa perform for this 16 core / 32 thread single-socket processor. Here are those results.
A few days back I did a similar LLVMpipe/SWR comparison on a 80 thread Intel system so check out those numbers if you are unfamiliar with these CPU-based OpenGL drivers... This testing is done mostly for curiosity about the viability of LLVMpipe/SWR on CPUs with high core counts.
I tested LLVMpipe on Ubuntu 17.04's default stack (Mesa 17.0.7) and then tested LLVMpipe and SWR when built from Mesa 17.3-dev Git paired with LLVM 6.0 SVN.
An intentionally awkward CPU/GPU combination to say the least...
And for putting the results into perspective, comparing these software numbers to some low-end Radeon hardware... A Radeon RX 460 and even with that being many times faster than LLVMpipe, also a vintage Radeon HD 4650.
An interesting round of benchmarks to say the least.
The LLVMpipe/SWR performance on the Threadripper 1950X is much better than you would find with older desktop CPUs with lower core counts, cache sizes, and those CPUs without AVX/AVX2, but still it's pretty slow.
LLVMpipe is mostly used as a means of driver fallback for debugging along a vendor-neutral code-path or also is being commonly used as a fallback for various Linux desktop environments when no GPU driver is available. OpenSWR is along similar lines while Intel appears to have actual customers using it for some server-side rendering of visualizations and other off-screen rendering tasks.
LLVMpipe and SWR also only support OpenGL 3.3.
More benchmarks via this OpenBenchmarking.org result file. OpenGL is still best kept on the GPU, even as core counts continue to rise and software drivers make use of AVX, etc.
For those wondering about the CPU utilization on Threadripper when using LLVMpipe/SWR, those curiosity numbers via this OpenBenchmarking.org page. Back to the real-world tests now...
| 20
| 1,760,719,710.12382
|
https://www.phoronix.com/news/AMD-TR-1950X-Compile-Times
|
Keeping The Ryzen Threadripper Busy With An Array Of Compiler Benchmarks
|
Michael Larabel
|
While there are an array of interesting AMD Ryzen Threadripper 1950X Linux benchmarks in this morning's review, after hitting a 36 second Linux kernel compilation time with this 16 core / 32 thread processor, I spent this afternoon seeing what I was getting for some other compile times of popular programs.
A 36 second defconfig Linux x86-64 kernel build isn't quite the fastest I've ever encountered with recently getting closer to 20 seconds with this dual Xeon Gold + Tyan 1U combination, but it is the fastest build time of my single socket CPU tests ever... It will be interesting to see how EPYC compares with its compile times once I get my hands on those parts, hopefully in just a matter of days. It used to take plenty of time to enjoy a coffee (or beer) while the Linux kernel compiled, but with the modern high-end processors from AMD and Intel, it can be easily done in less than one minute.
The @AMDRyzen #Threadripper 1950X managing a #Linux kernel build in about 36 seconds, not bad! pic.twitter.com/vQJwZKpakJ— Phoronix (@phoronix) August 24, 2017 See this morning's article for a plethora of Threadripper 1950X benchmark results compared to various AMD/Intel systems on Ubuntu while this article is just some complementary data, mostly looking at different compile times. GCC 4.9.4, GCC 6.3, GCC 7.2, GCC 8.0, LLVM Clang 4.0, and LLVM Clang 6.0 SVN were the compilers tested on this Threadripper box for these additional reference points... Mostly driven out of curiosity for seeing how much more I could reasonably squeeze out of the build times on this high-end desktop processor, short of dropping the compiler optimization levels or other big differences.
I did also try some runs when the tests were installed to tmpfs/zram, but that showed the Corsair MP500 NVMe SSD wasn't actually the bottleneck that the build times weren't getting reduced in any meaningful way. In a follow-up article in the days/weeks ahead I'll have a formal article looking at various compiler tuning / GCC/Clang benchmarks for Threadripper besides today's attempt at seeing how fast Threadripper can build some popular open-source/Linux programs.
All of these compiler benchmarks were facilitated in a fully-automated and reproducible manner using the open-source Phoronix Test Suite benchmarking software.
First up was the timed Apache HTTPD web-server build times.... LLVM Clang has gotten slower in some recent releases while GCC has also worked to improve its build times, but these Apache build times show LLVM Clang is still coming out up to a few seconds faster than the GNU Compiler Collection. GCC 8.0 SVN was the slowest compiler tested while Clang SVN was the fastest.
Next up is a look at how quickly the GCC releases could build out the GNU Compiler Collection 7.2 release. The oldest release tested, GCC 4.9.4, was the quickest at building the GNU compiler given over time more optimization passes and other features get added to GCC that often times slow it down. Fortunately though going from GCC 7.2 to GCC 8.0 is currently yielding a small build time improvement.
When building the ImageMagick graphics program, LLVM Clang could build it in 18 seconds while GCC was at 20 seconds and beyond. The older GCC 4.9.4 was the fastest GCC Compiler release tested.
Most exciting to many is the Linux kernel compilation times... Sadly, the mainline LLVM/Clang compiler can still not build yet the mainline Linux kernel. The LLVMLinux project that's been trying to advance the Clang-building-Linux effort has been stalled the past few years and there isn't much to report at this time on those prospects. With these releases tested on Ubuntu 17.04, the kernel build times were 42~46 seconds (my earlier 36 second build time was on Ubuntu 16.04 LTS).
Next is a look at the time to compile LLVM (not including Clang) with the various compilers. The GCC8 boost in compilation time here is rather surprising, but from initial tests look like that it may have been generating faulty code, so take that result worth a grain of salt for now.
Lastly with the PHP build times we see LLVM Clang running the fastest on the AMD Threadripper Ubuntu system. Amazing to see that PHP can be built in almost 10 seconds thanks to Threadripper!
Overall it was quite fascinating to see how quickly AMD's Threadripper 1950X was able to build some of these popular open-source packages as well as seeing which compilers were performing the best. More interesting Linux (and BSD) tests of the AMD Threadripper coming on Phoronix in the days ahead, but be sure to see this morning's AMD 1950X review if you have not already done so; this $999 USD CPU has a lot to offer for heavily threaded Linux workloads.
| 10
| 1,760,719,710.874965
|
https://www.phoronix.com/news/Ryzen-3-CPUFreq-Benchmarks
|
AMD Ryzen 3 CPUFreq Governor Benchmarks On Linux 4.13
|
Michael Larabel
|
For those curious about the performance impact of the different CPUFreq governors on a low-end Ryzen 3 processor, here are some benchmarks.
Using the Linux 4.13 Git kernel atop Ubuntu 17.04 with the AMD Ryzen 3 1200, I tested CPUFreq's ondemand, performance, powersave, schedutil, and conservative governors. As a reminder, Ubuntu defaults to CPUFreq's "ondemand" governor for AMD processors while the Intel CPUs using the P-State driver use "powersave" as their default.
All of these benchmarks were facilitated in a fully-automated and reproducible manner using the open-source Phoronix Test Suite.
In light tests like SQLite, the governor can make a small impact on the overall performance.
While for Linux gaming it tends to depend upon the title for how much of an impact the governor makes on the overall performance.
Many more benchmarks can be found via this OpenBenchmarking.org result file.
| 24
| 1,760,719,711.466196
|
https://www.phoronix.com/news/MJPEG-Gallium3D-VAAPI
|
AMD Patches MJPEG Decoding For VA-API Gallium3D
|
Michael Larabel
|
Leo Liu of AMD is out today with another series if video/multimedia related patches for the open-source Radeon Linux graphics driver stack.
Leo's latest work in the Mesa multimedia space is on exposing MJPEG decoding support through the VA-API Gallium3D state tracker. Of course, it's also hooked up into the Radeon UVD code to support the MJPEG decoding.
These 20 MJPEG VA-API patches for Mesa allow for 420(NV12) and 422(YUYV) format coverage. This support comes in at just over 400 lines of new Mesa code.
Those interested in GPU-based MJPEG decoding via VA-API for Radeon hardware can find the patches on Mesa-dev.
| 14
| 1,760,719,712.907502
|
https://www.phoronix.com/news/AMDKFD-Linux-4.14
|
AMDKFD Code Updated For Linux 4.14, More Changes Being Upstreamed
|
Michael Larabel
|
AMD is upstreaming more of their changes to the AMDKFD HSA kernel driver with Linux 4.14.
Separate from the AMDGPU changes for Linux 4.14, the AMDKFD HSA kernel driver changes have now been pulled into DRM-Next for then merging into Linux 4.14 once that merge window opens.
This cycle marks AMD developers cleaning up more of their code and sending patches from their internal development to the upstream community, per the commit message. They are working on getting the patches for non-discrete GPUs updated and into the upstream tree and then will focus on the discrete GPU support.
The Kaveri and Carrizo patches for AMDKFD in Linux 4.14 include a new ioctl interface for scratch memory VA, supporting image tiling mode, and various code clean-ups.
Hopefully in the months ahead we'll see more of AMD's work upstream just not for AMDKFD/HSA but also in supporting ROCm on more mainline components too, just not the Linux kernel changes.
| 15
| 1,760,719,713.255415
|
https://www.phoronix.com/news/AMD-ThreadRipper-Lift
|
AMD Threadripper Is Looking Good, At Least Under Windows
|
Michael Larabel
|
AMD's embargo has just expired on the Threadripper performance figures. The Windows numbers at least are very positive.
Launching today is the AMD Ryzen Threadripper 1920X and 1950X. As a reminder, the AMD Ryzen Threadripper 1920X has twelve cores plus SMT (24 threads), 3.5GHz base frequency, 4.0GHz boost frequency, and 4.2GHz XFR frequency. The Ryzen Threadripper 1950X meanwhile has 16 cores, 32 threads, 3.4GHz base frequency, 4.0GHz boost, and 4.2GHz XFR. Both models have a 180 Watt TDP, 32MB L3 cache, quad-channel DDR4-2666 support, and 64 PCI-E lanes.
Certainly some thread-happy chips there and should be great for many multi-threaded Linux workloads. The Windows reviews out so far seem to be mostly praising these new chips, albeit they are expensive at $799 USD for the 1920X and $999 USD for the 1950X. They are expensive, but they are able to compete well -- and in some workloads exceed -- Intel's current Core i9 / Core X Series offerings.
As a reminder, the Ryzen Linux issue doesn't affect Threadripper or Epyc for that matter.
Unfortunately, no Linux results to share today... At first AMD wasn't planning on sending Phoronix any Threadripper samples but then on Monday following the Ryzen Linux issue incident I was told that we would indeed be getting Threadripper and Epyc samples. Still waiting on them, but as soon as they arrive, I will be going benchmark crazy. Stay tuned.
| 64
| 1,760,719,714.22117
|
https://www.phoronix.com/news/AMDGPU-DC-9Aug-Kernel-Spin
|
An Ubuntu Kernel Built With The Latest AMDGPU DC Support
|
Michael Larabel
|
For those running Ubuntu or one of its derivatives that have been wanting to play with AMDGPU's DC "display code" functionality but can't be bothered to build the branched code, here's a fresh kernel build.
This morning I spun up a fresh Ubuntu package of the latest AMDGPU staging code containing the DC display code and is based on their Linux 4.12 staging tree. This DC display code is what provides HDMI/DP audio for modern Radeon GPUs, HDMI 2.0, Radeon Vega support, atomic mode-setting, and other modern display capabilities. It's built from the amd-staging-4.12 branch updated a few days ago with the new DC code but not quite the 120 patches sent out earlier today.
Yes, this kernel is additionally configured with the CONFIGURE_DRM_AMD_DC and CONFIG_DRM_AMD_DC_DCN1_0 switches enabled for actually enabling the new code paths.
Test at your own risk, but if you want an easy Ubuntu x86_64 build, you can hit it up here. It goes great, of course, with the Padoka PPA too if wanting easily the Mesa 17.3-dev Git state atop Ubuntu.
| 38
| 1,760,719,714.666649
|
https://www.phoronix.com/news/AMDGPU-120-More-For-DC
|
With AMDGPU DC Not Aligning For Linux 4.14, 120 More Display Code Patches
|
Michael Larabel
|
It's looking almost certain that AMDGPU's display code (a.k.a. "DC" and "DAL") will not be merged for the next Linux 4.14 cycle, but work on this massive display code-base is progressing and 120 more patches were published today.
AMDGPU DC for Linux 4.14 is looking increasingly unlikely considering Linux 4.13-rc4 was released this past weekend and this display code is not yet queued in DRM-Next let alone AMD's -next tree. David Airlie cuts merging off new feature material for the next kernel cycle around ~RC6 of the previous series, thus new feature work will be cut off for Direct Rendering Manager drivers for Linux 4.14 within about two weeks. But with the display code not being queued yet and the massive patch-set not having been sent out for for a public review on dri-devel by other upstream kernel developers, given the size of the code-base it's almost impossible for it to be reviewed within two weeks.
AMDGPU DC is up to 600+ patches with around 160k lines of new code to show just how complex this new display code is for this modern AMD Linux kernel graphics driver. And now today it gets 120 more patches.
AMD developer Harry Wentland posted to amd-gfx a short time ago, "A whole lot of stuff this time around because it's Raven crunchtime and a bunch of us found time to keep massaging DC in DRM's image." Raven Ridge is the upcoming APU launch expected around year's end to feature a Zen CPU microarchitecture with Vega class graphics.
These 120 patches have more Raven Ridge work plus more DC code cleaning to better align the display code's concepts with those used by the DRM subsystem in the Linux kernel. There is more flattening (less abstractions) around core/DC objects, renaming of some code, and a ton of other code changes and more bug fixes for Raven. Reducing the abstractions in the code is certainly good as that was one of the original big complaints with the driver's code back when it was published as DAL.
They are certainly moving in the right direction, but it looks like we probably won't see this mainlined until at least Linux 4.15. If you want AMDGPU DC today for either HDMI/DP audio, Radeon RX Vega display support, or other sought-after modern display features, the easiest way is to install AMDGPU-PRO (assuming you are on a supported distribution) or build the current AMDGPU staging tree that is currently based against Linux 4.12.
| 45
| 1,760,719,715.852257
|
https://www.phoronix.com/news/Ryzen-Segv-Response
|
AMD Confirms Linux Performance Marginality Problem Affecting Some, Doesn't Affect Epyc / TR
|
Michael Larabel
|
This morning I was on a call with AMD and they are now able to confirm they have reproduced the Ryzen "segmentation fault issue" and are working with affected customers.
AMD engineers found the problem to be very complex and characterize it as a performance marginality problem exclusive to certain workloads on Linux. The problem may also affect other Unix-like operating systems such as FreeBSD, but testing is ongoing for this complex problem and is not related to the recently talked about FreeBSD guard page issue attributed to Ryzen. AMD's testing of this issue under Windows hasn't uncovered problematic behavior.With the Ryzen segmentation faults on Linux they are found to occur with many, parallel compilation workloads in particular -- certainly not the workloads most Linux users will be firing off on a frequent basis unless intentionally running scripts like ryzen-test/kill-ryzen. As I've previously written, my Ryzen Linux boxes have been working out great except in cases of intentional torture testing with these heavy parallel compilation tasks. Even when carrying out other heavy, non-compilation (GCC or Clang) parallel workloads in recent days, from server tasks to scientific processing, my Ryzen test boxes have been stable. I'm still using Ryzen 5 on my main desktop system without any faults in day-to-day use on Fedora 26 Linux.
AMD was also able to confirm this issue is not present with AMD Epyc or AMD ThreadRipper processors, but isolated to these early Ryzen processors under Linux. We will also now be receiving Threadripper and Epyc hardware for testing to confirm their Linux state. Their analysis has also found that these Ryzen segmentation faults aren't isolated to a particular motherboard vendor or the like, contrary to rumors/noise online due to the complexity of the problem.
Ryzen customers believe to be affected by the problem can contact AMD Customer Care. Some of those who have contacted customer care about the segmentation faults have in turn been affected by thermal, power, or other problems, but AMD says they are committed to working with those encountering this performance marginality issue under Linux. AMD will also be stepping up their Linux testing/QA for future consumer products.
| 287
| 1,760,719,716.174404
|
https://www.phoronix.com/news/Ryzen-Test-Stress-Run
|
Ryzen-Test & Stress-Run Make It Easy To Cause Segmentation Faults On Zen CPUs
|
Michael Larabel
|
With running a number of new Ryzen Linux tests lately, a number of readers requested I take a fresh look at the reported Ryzen segmentation fault issues / bugs affecting a number of many Linux users. I did and still am able to reproduce the problem.
For those that missed our earlier article on the matter from early June, heavy workloads can cause problems on Ryzen, in particular segmentation faults while there have also been reports of some stability problems.
This Google Doc remains among the resources trying to track this issue on Linux while on the Gentoo Forums, AMD Forums, and elsewhere are more reports of various problems encountered under extreme workloads -- like a ton of code compiling for hours on end, but can also happen in other scenarios.
AMD hasn't publicly commented on the problem and as of Linux 4.13 the issue is still happening. If carrying out the same tests on Intel CPUs, the segmentation faults do not occur. There is even ryzen-test to easily try reproducing the issue. The ryzen-test script will build GCC in parallel loops from a compressed ramdisk, in order to easily stress the CPU. In my day-to-day benchmarking of Ryzen CPUs, however, I haven't hit this problem or even on my main production desktop with using Ryzen 5. The problem really comes to light just under very heavy and continuous workloads it seems.
I have been running some tests with a Ryzen 7 1800X with the MSI X370 XPOWER TITANIUM GAMING motherboard with Linux 4.13 and Corsair DDR4-3200 memory. I have been using the latest BIOS, 7A31v17, which is built with AGESA 1.0.0.6.
Sure enough, in the stock configuration of the Ryzen 7 1800X and with the DDR4-3200 speed activated, the first segmentation fault with ryzen-test happened in just 88 seconds.
I also tried setting the memory to its defaults at DDR4-2133, but the issue occurred still in 83 seconds. When disabling SMT, it takes much longer for the problem to come to head, but it does eventually happen. Disabling SMT seems to be the closest workaround those experiencing this problem more often in the community have done to try to avoid problems, but then you lose half the threads of the CPU.
Besides using ryzen-test to hammer the CPU to easily reproduce the issue, I also decided to use phoronix-test-suite stress-run. The stress-run command within the Phoronix Test Suite has been used by enterprise customers for stress testing / burn-ins of hardware and checking for stability. Rather than benchmarking for performance, stress-run allows executing multiple test profiles in parallel for fully loading the system with whatever workloads you would like. Using PTS_CONCURRENT_TEST_RUNS=4 TOTAL_LOOP_TIME=60 phoronix-test-suite stress-run build-linux-kernel build-php build-apache pgbench apache redis will have the Phoronix Test Suite continually running four different benchmarks simultaneously for a period of 60 minutes. As soon as one test finishes, another is fired up. The stress-run algorithm randomly picks the tests of your set to run, but does look at the test profile to ensure if the tests stress multiple subsystems, it tries to ensure stress on all subsystems are always being stressed. The Phoronix Test Suite's stress-run functionality isn't advertised as much as its other features, but is very useful for loading up a system with plenty of real-world workloads concurrently.
In this case of multiple processes of code compilation, PostgreSQL, Apache, and Redis, Ryzen drops to its knees very quickly.
While with the ryzen-fail demo program when disabling SMT it could take up to a half hour to get a fail reported, with the Phoronix Test Suite stress-run for the Ryzen 7 1800X with eight cores and no SMT, I managed to get the first segmentation fault after the system was booted up for just 229 seconds... And the segmentation faults would continue every few minutes in this configuration under the immense workloads.
We'll see now if AMD will provide public comments or if they investigate further as they now have another reproducible test case to slam the Ryzen chips hard in just a few minutes even with SMT disabled and running at DDR4-2133. As far as whether this just affects Ryzen or also Threadripper and Epyc remains unclear. While there are many Windows reviewers out there now with Threadripper, it doesn't look like AMD will be sending any Threadripper samples to Phoronix, at least in the immediate days ahead but I have asked if at least can get SSH access to a TR system for a few hours to be able to run some Linux benchmarks. We'll see. For the Epyc server processors as well, no samples are available according to a motherboard vendor that has been trying to get them on my behalf.
Just to reiterate, while this problem is easy to cause under very heavy workloads, under normal Linux desktop workloads and even normal benchmarking, I haven't run into any Ryzen problems. I will be running some more Ryzen stress-tests today.Update [7 August]: AMD Confirms Linux Performance Marginality Problem Affecting Some, Doesn't Affect Epyc / TR
| 104
| 1,760,719,717.648731
|
https://www.phoronix.com/news/Open-Source-ProRender
|
Radeon's ProRender Renderer Now Available In Source Form
|
Michael Larabel
|
AMD's GPUOpen initiative has announced the open-source availability of their ProRender renderer.
Their open-source Prorender is an implementation of the Radeon ProRender API and demonstration for the AMD Radeon Rays intersection engine that has evolved into its own full-functioning render engine.
The ProRender renderer relies upon OpenCL 1.2 support.
More details on the open-source ProRender project via GPUOpen.com or by going to the GitHub project site.
| 0
| 1,760,719,717.663314
|
https://www.phoronix.com/news/AMD-Ryzen-3-Today
|
AMD Ryzen 3 Rolls Out, Linux Benchmarks Coming
|
Michael Larabel
|
AMD has completed their Ryzen desktop rollout today with the availability of Ryzen 3 CPUs on the low-end.
Ryzen 3 CPUs are quad-core without any SMT support. The Ryzen 3 1200 is clocked at 3.1GHz with a 3.4GHz boost clock and 65 Watt TDP. The Ryzen 3 1300X meanwhile has a 3.5GHz base clock and 3.7GHz boost clock, also at 65 Watts. The Ryzen 3 1200 will sell for about $109 USD and the Ryzen 3 1300X for about $129 USD.
These should make for some nice quad-core Linux boxes for those just wanting to spend a little more than $100 on the CPU.
Fortunately, I've heard from AMD today I will be receiving both CPUs in the days ahead for Linux testing, but in the mean time you can see my past Ryzen 5 and Ryzen 7 Linux benchmarks compared to various Intel CPUs on Ubuntu.
More Ryzen 3 details at AMD.com.
| 54
| 1,760,719,718.97405
|
https://www.phoronix.com/news/81-More-DC-Patches
|
81 Fresh Patches For AMDGPU's DC (DAL) Display Code
|
Michael Larabel
|
AMD has published another set of DC/DAL display patches today, this time amounting to 81 patches touching around four thousand lines of code.
This latest DC patch series begins cleaning up some of the AMDGPU DRM driver's midlayer abstractions around this display code, prep work towards FBC support (frame-buffer compression), future-proofing of DCN functions, Raven pipe splitting features, and various bug fixes around Raven Ridge support -- the upcoming Zen+Vega APU.
There's also a fix for S3 gamma corruption, some other display fixes, "DAL1.1" updates, Raven Ridge stereo support, and some more steps around FreeSync support but that appears to be broken at present.
It remains to be seen if the AMDGPU DC code will be ready for mainline with Linux 4.14 or if interested users will still need to wait longer before seeing it part of the mainline Linux kernel... It should be clear within the next few weeks as it will first need to go through another big patch review process on dri-devel. As David Airlie cuts off the DRM-Next new material merge window around rc6~rc7 of the previous cycle and given the size of the DC patch series, that review process would need to go well and be done within the next couple of weeks if there is any chance in seeing the support for Linux 4.14. DC is what's needed for Vega display support, HDMI/DP audio on modern GPUs, and many other modern display features as we've been covering about with DC/DAL over the past many months.
These 81 newest DC patches can be found on amd-gfx.
| 58
| 1,760,719,719.239082
|
https://www.phoronix.com/news/AMD-SEV-v3-Linux
|
AMD Secure Encrypted Virtualization Updated For Linux
|
Michael Larabel
|
While AMD's new Epyc processors have a new "Secure Encrypted Virtualization" feature, the support isn't yet mainlined in the Linux kernel but is getting closer.
Brijesh Singh of AMD today published the third revision to the patches implementing Secure Encrypted Virtualization for the Linux kernel. SEV allows for encrypting the memory contents of a guest VM using a unique key for each guest. As Singh further describes, "SEV guests have concept of private and shared memory. Private memory is encrypted with the guest-specific key, while shared memory may be encrypted with hypervisor key. Certain type of memory (namely insruction pages and guest page tables) are always treated as private. Due to security reasons all DMA operations inside the guest must be performed on shared memory."
Secure Encrypted Virtualization builds upon Secure Memory Encryption (SME), another new feature to AMD Epyc and another yet-to-be-mainlined feature. The latest SME patches can be found here.
Hopefully SME and SEV will be ready for merging come the Linux 4.14 cycle as it's now too late for 4.13. The latest patches for those fortunate to have their hands on Epyc can find them via this kernel mailing list post.
| 17
| 1,760,719,720.144192
|
https://www.phoronix.com/news/AMD-Grenada-Linux-2
|
Some AMD Grenada Cards Have Been Borked On The Open-Source Driver For 2 Years
|
Michael Larabel
|
A frustrated Phoronix reader pointed out a bug report that's been open nearly two years regarding AMD Grenada (basically, Hawaii cards in the R9 300 series) support on the open-source Linux driver being in a tough position for a subset of users.
There are multiple users reporting of Grenada Linux driver problems, but the percentage of Grenada users affected isn't clear, of performance issues as well as blank screen problems. The original bug report dates back to September 2015 and there has been more than 150 comments about these woes without any definitive resolution. Workarounds like forcing the DPM state help some users while switching over to the AMDGPU DRM driver doesn't make a difference for some.
The performance problems slightly remind me of my R9 290 Hawaii woes, which appear to be back with Linux 4.13 under some games, leading to very low performance on the Radeon DRM driver. At least in my case, switching to the AMDGPU DRM driver has always worked out pleasantly.
Do you have any Hawaii / Grenada Linux driver problems? Share with us in the forums. If you have any insight to add about the problems or curious about what is being encountered, see FreeDesktop bug 91880.
| 27
| 1,760,719,720.862351
|
https://www.phoronix.com/news/AMD-Secure-Memory-V10
|
AMD Secure Memory Encryption Patches Updated For Linux
|
Michael Larabel
|
Adding to the list of changes/features you will not find in Linux 4.13 is AMD's Secure Memory Encryption as supported by the new EPYC processors.
AMD has been posting Secure Memory Encryption patches for the Linux kernel going back to last year, but so far have not been merged to mainline. The code continues to be updated and published today was the tenth version of these patches.
The AMD Secure Memory Encryption (SME) patches in their v10 form disable this feature when built for 32-bit kernels and has other changes. Still though to address is Kdump support. Additionally, these patches just enable SME and not yet Secure Encrypted Virtualization (SEV) as another security/encryption feature found on EPYC processors.
Hopefully we'll see this work be ready for mainline with the Linux 4.14 cycle now that it's too late for Linux 4.13, but those wanting to test the support today can find the current patch series.
| 9
| 1,760,719,721.555001
|
https://www.phoronix.com/news/AMD-ThreadRipper-Prices
|
AMD Reveals First ThreadRipper Prices, Early August Launch
|
Michael Larabel
|
AMD has announced the first ThreadRipper SKUs and more.
Here's the details from a video posted by AMD today:
- AMD's X399 platform and ThreadRipper CPUs will be available in early August.
- Initial CPUs include 16-core / 32-thread and 12-core / 24-thread models.
- The Ryzen ThreadRipper 1950X is a 16-core / 32-thread CPU with 3.4GHz base frequency and 4.0GHz boost frequency. This CPU will cost $999 USD.
- The Ryzen ThreadRipper 1900X is 12-core / 24-thread with 3.5GHz base frequency and 4.0GHz boost frequency. This CPU will be $799 USD.
- Ryzen 3 SKUs of the Ryzen 3 1200 with 4-core/4-thread and 3.1GHz base frequency and 3.4GHz boost. The Ryzen 3 1300X is another 4-core/4-thread part but with 3.5GHz base frequency and 3.7GHz boost. Ryzen 3 hardware will be out 28 July.
Watch the video below:
| 41
| 1,760,719,722.302738
|
https://www.phoronix.com/news/AMD-ARM-In-My-Hands
|
After Years Of Waiting, Hands On With The AMD ARM Board
|
Michael Larabel
|
With Zen CPUs turning out very well in the marketplace, AMD appears to have divested some of their interest in ARM-based processors at least for the time being. But after waiting for years, I finally have my hands on an AMD Opteron A1100 ARM-based SBC for testing.
I was ecstatic when earlier this week an AMD ARM board turned up at my door thanks to our friends at one of the ARM-focused companies around, LoveRPI. The board that arrived is the LeMaker Cello.
The Cello was announced in early 2016 as a $299 USD developer board with the A1100 ARM 64-bit SoC with four Cortex-A57 cores @ 1.7GHz. The Cello features two DDR3 SO-DIMM slots capable of ECC RAM, two Serial ATA 3.0 ports, Gigabit Ethernet, two USB 3.0 ports, and one PCI Express x16 3.0 slot. There are no integrated graphics, but the PCI-E x16 slot can make for some interesting testing.
This board was to begin shipping in Q2'2016, but it and the AMD HuskyBoard never materialized in 2016 and remain largely unavailable to this day. Last month we heard the LeMaker Cello finally shipping to some pre-order customers but on their web-site is listed as "no stock." (Meanwhile, the HuskyBoard that is still yet to be available was first talked about for release back in 2015 but never materialized.)
So it's really not clear what the future holds for the AMD A100 and future AMD ARM CPUs, but anyhow, our curiosity over this SoC can finally be answered with now being able to run some real Linux benchmarks on it. Unfortunately, no results to share today as the power adapter and DDR3 ECC Unregistered SO-DIMMs are arriving in a later shipment.
We are told by the folks at LoveRPI.com, they will be selling their limited supply of LeMaker Cello boards paired with supported memory in August. Stay tuned for the test results in the days/weeks ahead.
| 27
| 1,760,719,723.157102
|
https://www.phoronix.com/news/Vulkan-Memory-Allocator-1.0
|
AMD/GPUOpen Vulkan Memory Allocator 1.0 Released
|
Michael Larabel
|
AMD's GPUOpen team has announced their v1.0 release of the Vulkan Memory Allocator.
Last month is when they initially debuted their Vulkan Memory Allocator project as an easy-to-use library by developers wishing to make use of the Vulkan graphics API but trying to reduce boilerplate code around memory management and making some of the memory handling easier. This C++ library has now reached version 1.0.
With the 1.0 marking, AMD believes this Vulkan Memory Allocator is now ready for wide-spread use. The 1.0 release supports easy allocation of buffer and image storage, various code samples, and more. With the Vulkan Memory Allocator 2.0 they hope to make this library more suitable for games with functionality like texture streaming.
More details on the Vulkan Memory Allocator 1.0 via GPUOpen.com.
| 1
| 1,760,719,723.911609
|
https://www.phoronix.com/news/ROCm-OpenCL-Background
|
AMD's Background On The ROCm OpenCL Stack
|
Michael Larabel
|
A ROCm (Radeon Open eCosystem) developer at AMD has shared some of their background work on their OpenCL compiler stack, including the LLVM focus, as well as some of their current performance focuses for this open-source compute offering.
Gregory Stoner has written in a GitHub comment about their past and ongoing work around ROCm OpenCL performance. A Phoronix reader pointed out to me this comment. It's interesting so I've copied it below. His comments came up in a thread about why the ROCm performance is currently slow for Ethereum mining.
Now the AMDGPUpro driver for Vega10 supports the new lighting compiler and ROCm stack as well
When we started the ROCm project, we made a decision to build out fully open source solution, which meant we need to move away from the traditional Shader Compiler used in our graphics stack since it was staying proprietary. The traditional flow was two-stage compiler; we would compile the code to an intermediate language, HSAIL, then it would be picked up finalized and compiled by our shader compiler. This same backend used by Graphics shaders.
This journey started in earnest a little over a year ago to look the best way forward to fully open source compiler. We began with the LLVM R600 codebase which needed a bit of work to get to be production class compiler. But it was the right foundation to meet our goal of a fully open stack,
With this transition, we know we will have performance gaps, which we are working to close. What we need help with from the community is assist us in testing a broader set of applications and reporting the and do some analysis potentially why. One thing we have seen as well sometimes you need to code differently for LLVM compiler then the SC based compiler to get the best performance out if it.
We are now active in the LLVM community, pushing upgrades to the code base to better enable GPU computing. Also, changes are also up-streamed into LLVM repository.
Note one significant changes the compiler now generate GCN ISA binary object directly. With this change, it makes it easier for the compiler supports Inline ASM support for all of our languages ( OpenCL, HCC, HIP) and also native assembler and disassembler support. It is also a critical foundation for our math library and MiOpen projects.
For the last year, we have spent more time focusing on FIJI and Vega10 with Deep Learning Frameworks, MIOpen, and GEMM solvers. We also have been filling in the gaps in LLVM for the optimization we need for GPU Computing, also improving the scheduler, register allocator, loop optimizer and lot more. It is a bit of work as you can imagine. But we already saw where the effort been worth it since it faster on a number of the codes.
We test thing like follow on the compiler
Benchmarks: Bablestream, SHOC, Mixbench, Lattice, ViennaCL, COMD, Lulesh, xsbench. Rodina, DeepBench
Libraries: clFFT, rocBLAS, rocFFT, MIOpen
Application:
OpenCL: Torch-CL, Gromacs;
HIP: Caffe Torch, Tensorflow,
HCC: NAMD
Internal test we built up for performance for OpenCL
Conformance tests for
OpenCL 1.2 and 2.0 Conformance tests
HCC conformance test
Not above is small sample of what we run on the compiler. We do A/B compares
New test recently added: Radeon Rays, SideFX Houdini Test, Blender, Radeon ProRender,
In the process of adding a number of currency mining apps
On Ray Tracer we are just starting our performance analysis and optimization that more specific to this class of work, What you see over the summer is we will be focusing on optimization for the compiler for currency mining and raytracing. I just have to stage this work in with the team. I saw you referenced Phoronix article, for ROCm 1.5 the new compiler was faster than LLVM/HSAIL/SC on FIJI for Blender, but for Luxmark we were slower. //www.phoronix.com/review/rocm-15-opencl#=2
One thing I will leave you with is we build standardized loader and linker and object format, with this it allows us to do some you never could do with AMGGPUpro driver, upgrades the compiler before we release a new driver. So we can now address issue independently of the base driver for OpenCL, HCC, and HIP and the base LLVM compiler foundation.And for those wondering, I will have some more ROCm 1.6 OpenCL benchmarks shortly, which was released last week.
| 27
| 1,760,719,724.493765
|
https://www.phoronix.com/news/Linux-4.13-Hwmon
|
Hwmon Updates Submitted For Linux 4.13, Still No Ryzen/Epyc Temp Support
|
Michael Larabel
|
The hwmon (hardware monitoring) subsystem updates have been submitted for Linux 4.13 and what's sad about the pull request is what isn't present.
While AMD Ryzen is supported by the mainline Linux kernel as well as the forthcoming Epyc CPUs, they are sadly missing a feature important to some: thermal monitoring. While it may seem rather trivial, as of Linux 4.13 Git there still isn't any supported driver for being able to read the Zen-based CPU temperatures under Linux.
The hwmon pull request was sent today and sadly doesn't contain any Zen support changes for either adding the Zen temperature reading support to the k10temp driver (what currently supports from the AMD Family 10h CPUs up through Bulldozer/Kaveri/Carrizo as well as APUs like Kabini and Mullins) or a new driver.
Also for that matter, the fam15h_power driver that is also within the hwmon subsystem and reports power measurements on newer AMD CPUs also hasn't received any Zen support.
| 10
| 1,760,719,725.950304
|
https://www.phoronix.com/news/Radeon-RX-SIGGRAPH-Firming
|
AMD Firms Up It's Releasing Radeon RX Vega At SIGGRAPH
|
Michael Larabel
|
Just in case you had any doubt of AMD launching the Radeon RX Vega graphics card later this month at SIGGRAPH, they've now made it publicly clear.
AMD's Radeon twitter account tweeted yesterday, "We can't wait to announce our new Vega products, including RX at this year's #SIGGRAPH - make sure to follow us for more details."
SIGGRAPH this year is running from 30 July to 3 August in Los Angeles. Should be a great SIGGRAPH this year with Vega and OpenGL/Vulkan news. Stay tuned.
| 11
| 1,760,719,726.130325
|
https://www.phoronix.com/news/AMD-Ryzen-PRO
|
AMD Announces Ryzen PRO, Mobile Parts In 2018
|
Michael Larabel
|
First there was Radeon Pro and now there is Ryzen PRO for CPUs catering towards business customers. Ryzen PRO desktop CPUs will be out around the end of summer while mobile PRO parts will come in H1'2018.
Ryzen PRO CPUs are designed for "reliability, security, and performance to enterprise desktops worldwide."
The Ryzen PRO desktop line-up will consist of:
More details at AMD.com.
| 18
| 1,760,719,727.418073
|
https://www.phoronix.com/news/AMD-SME-SEV-Epyc
|
AMD's SME/SEV Security Support For EPYC Not Yet Ready On Linux
|
Michael Larabel
|
While AMD announced their EPYC 7000 series CPUs last week, prominent new security features of these high-end processors aren't yet ready with support in the mainline Linux kernel.
New security features added to the Zen-based EPYC server processors is Secure Memory Encryption (SME) and Secure Encrypted Virtualization (SEV). Secure Memory Encryption provides memory encryption on a per-page-table basis using AMD's ARM-based security co-processor. AMD SME + SEV are designed against both user-access attacks and physical access attacks with a particular focus on VM / hypervisor security. Sadly, support for SME and SEV have yet to be mainlined in the Linux kernel, thus EPYC Linux servers don't yet benefit from this new technology.
AMD posted SME patches back in April of 2016 but as of Linux 4.12 the work has yet to be mainlined and it's looking like it might not be ready yet for Linux 4.13. SEV patches are still pending for public posting. (For those concerned about a free software system, Epyc's secure processor firmware remains a binary blob.)
Posted on Tuesday was the latest SME patches. These 38 patches implement Secure Memory Encryption for the Linux kernel, "SME can be used to mark individual pages of memory as encrypted through the page tables. A page of memory that is marked encrypted will be automatically decrypted when read from DRAM and will be automatically encrypted when written to DRAM."
The Secure Encrypted Virtualization work meanwhile has yet to be published, "This patch series is a pre-cursor to another AMD processor feature called Secure Encrypted Virtualization (SEV). The support for SEV will build upon the SME support and will be submitted later."
Since the earlier version of these patches, the latest SME code has a number of fixes and some other changes. Still left to do by the developers is adding Kdump support. Hopefully it won't be too many more kernel releases before seeing SME/SEV appear in the mainline tree.
| 9
| 1,760,719,727.4908
|
https://www.phoronix.com/news/AMDGPU-PRO-17.20-Vega
|
AMDGPU-PRO 17.20 Emerges With Vega, ROCm Compute Support
|
Michael Larabel
|
Thanks to today's Radeon Vega Frontier Edition launch, AMD has released an updated AMDGPU-PRO Linux hybrid driver.
AMDGPU-PRO 17.20 is the new series introduced for Radeon Vega Frontier Edition, which finally appears to be a measurable update over the previous 17.10 series. Besides Vega Frontier support, the 17.20 -PRO driver has now bundled the ROCm OpenCL compute component into this hybrid driver as a replacement to its older OpenCL driver.
The release notes mention, "Ubuntu 16.04.2 and RHEL 7.3/CentOS 7.3 installations incorporate the ROCm component that can be optionally installed for running Compute/OpenCL applications – any other version of ROCm is not currently supported." But it isn't clear that this ROCm component is the new ROCm release, which AMD has said will be released on 29 June.
Vega and the ROCm integration are the only mentioned changes for the 17.20 series. At least though the Vega support supplies an updated DC display stack for this driver's DKMS module.
To much disappointment, the AMDGPU-PRO 17.20 driver still doesn't appear to support the months-old Ubuntu 17.04... Ubuntu 16.04.2 is listed as supported along with RHEL/CentOS 7.3, RHEL/CentOS 6.9, and SLED/SLES 12 SP2. Thus this 17.20 driver probably won't work yet on newer Linux kernels / X.Org Servers if it doesn't even have 17.04 support yet.
Those wanting to try out the AMDGPU-PRO 17.20 driver for other hardware can find it via the Frontier page. I'll be running some benchmarks shortly (on non-Vega GPUs) to see if 17.20 happens to bring any OpenCL/OpenGL/Vulkan performance changes.
| 42
| 1,760,719,728.832976
|
https://www.phoronix.com/news/Stoney-More-Coreboot
|
More AMD Stoney Ridge Code Lands In Coreboot
|
Michael Larabel
|
It looks like the first AMD-powered Chromebook might be getting closer to reality.
Landing overnight in Coreboot was more code concerning AMD Stoney Ridge APUs, which is what's used by the Google "Kahlee" Chromebook. Coreboot consultant Marc Jones has added Stoney Ridge CPU files, Northbridge support, and more. The AMD Gardenia development board was then also updated for making use of this new Stoney Ridge code.
Stoney Ridge was AMD's 2016 APU architecture with Excavator CPU cores and GCN 1.2 graphics. More exciting down the pipe is the Raven Ridge APUs with Zen CPU cores plus Vega graphics, though it's yet to be clear if/when we'll see Coreboot support on that front.
| 8
| 1,760,719,729.032694
|
https://www.phoronix.com/news/RadeonSI-ARB_gl_spirv-Steps
|
AMD's Plans For ARB_gl_spirv Support In RadeonSI
|
Michael Larabel
|
Back in May we talked about ARB_gl_spirv / NIR Support Being Worked On For RadeonSI while now we have more details from AMD's Nicolai Hähnle regarding these plans.
Nicolai has written a blog post to share an outline for his plans on supporting ARB_gl_spirv and being able to re-use the SPIR-V/NIR code-paths within RadeonSI, which right now are used by the RADV Vulkan driver.
For going from SPIR-V to LLVM IR for the AMDGPU stack, he's planning on making use of some RADV code. But there's still plumbing in the main Mesa code that needs to be done as well for SPIR-V support. He's planning to re-use the GLSL linker for making this happen.
Those interested in more details on his plans can read this blog post.
The OpenGL ARB_gl_spirv extension is what allows a SPIR-V module to be used by OpenGL and for modifying GLSL to be a source language for creating SPIR-V modules. SPIR-V, of course, being the IR used by Vulkan and OpenCL 2.1+.
| 3
| 1,760,719,730.260371
|
https://www.phoronix.com/news/AMD-ProRender-Baikal
|
AMD's GPUOpen Releases Baikal Renderer
|
Michael Larabel
|
The Baikal renderer is a newly-released, open-source implementation of the AMD Radeon ProRender API. Baikal has evolved into a fully-functional rendering engine and its only hardware requirement is on OpenCL 1.2.
Baikal supports light transport, geometry, advanced materials, a number of lighting capabilities, various samplers, and more. This renderer is cross-platform and AMD says it's intended for use by researchers, educational institutes, and open-source enthusiasts.
More details via GPUOpen.com. The code is hosted on GitHub.
| 10
| 1,760,719,730.437274
|
https://www.phoronix.com/news/AMD-EPYC-7000-Series
|
AMD EPYC 7000 Series CPUs Launched
|
Michael Larabel
|
AMD has formally announced today their EPYC 7000 series line-up of processors, their server/workstation offerings based on Zen to finally battle Intel's multi-year dominance with Xeon and AMD's long-awaited successor to the Opteron family.
AMD's flagship EPYC 7601 processor has 32 cores, 64 threads, 128 PCI Express 3.0 lanes, eight channels of DDR4-2666MHz memory, and clocks up to 3.2GHz turbo (2.2GHz base) with only a 180 Watt TDP. At the bottom of the current EPYC line-up is the 7251 that has eight cores, 16 threads, and clocks up to 2.1GHz base / 2.9GHz turbo with a 120 Watt TDP.
EPYC can address up to 2TB of memory per CPU. Pricing on the EPYC 7000 series hardware hasn't yet been officially announced.
The line-up of products and more details can be found at AMD.com. No word on any review samples for being able to deliver Linux benchmarks.
| 25
| 1,760,719,731.66177
|
https://www.phoronix.com/news/GPUOpen-New-VLK-Allocator
|
AMD's GPUOpen Posts New Vulkan Memory Allocator
|
Michael Larabel
|
AMD's GPUOpen initiative has posted a number of Vulkan open-source projects over time from the Anvil Vulkan framework to a Vulkan-supported CodeXL and various code samples. Their latest open-source project is a Vulkan memory allocator.
Details are still light as GPUOpen doesn't appear to have officially announced it yet, I just stumbled upon it through my routine Vulkan crawling around GitHub. The initial public commit to GPUOpen's VulkanMemoryAllocator was back on Friday.
The VulkanMemoryAllocator is designed as a "easy to integrate Vulkan memory allocation library." This allocation library tries to remove some of the boilerplate code, remove additional indirection, make it easier to manage memory across drivers, out-of-memory handling, and more. The library is cross-platform and is a self-contained C++ library.
Right now the documentation is fairly light with the README just encouraging interested developers to read the code comments in this header file.
| 5
| 1,760,719,731.94352
|
https://www.phoronix.com/news/AMDGPU-PRO-17.10-Perf
|
Updated AMDGPU-PRO Driver Has Performance Fixes, Mad Max Works On Vulkan
|
Michael Larabel
|
AMD has just released a new AMDGPU-PRO 17.10 driver. While another 17.10 series driver release for this hybrid stack may not seem exciting without a large version bump, there are some noteworthy changes to this release.
The AMDGPU-PRO 17.10-429170 is the new release available today. It now has official support for the Radeon RX 560 series (though it's worked out fine with the earlier 17.10 series that had RX 550/570/580 support officially) and there are a number of fixes. In fact, it's the fixes that make this release exciting:
Fixed an issue with Mad Max on Vulkan
Updated marketing names for Radeon™ Pro Duo and AMD FirePro™ S9300 x2.
Some performance issues with AMD Radeon™ R9 2xx cards on glmark2, tessract and gpuplot3D addressed. Great to see Mad Max on Vulkan now working with AMDGPU-PRO. The performance fixes for Radeon R9 200 series hardware is also interesting. I'll be working on some fresh AMDGPU-PRO vs. RadeonSI benchmarks shortly.
Sadly, this driver doesn't yet have official support yet for Ubuntu 17.04 or newer distributions, like those on Linux 4.10+, X.Org Server 1.19+, etc. This release for supported distributions is available for download from AMD.com.
| 19
| 1,760,719,732.928865
|
https://www.phoronix.com/news/AMDGPU-Linux-4.13-Round-1
|
AMD Submits Radeon/AMDGPU DRM Updates Slated For Linux 4.13
|
Michael Larabel
|
AMD developers have submitted their first round of Radeon/AMDGPU feature updates to DRM-Next for in turn targeting the Linux 4.13 kernel.
The first thing most of you are probably wondering: no, DC (DAL) isn't part of this pull request nor is there any commentary in the message whether they hope to have this support ready for Linux 4.13. The DC/DAL display stack is what's needed for Vega display support, HDMI/DP audio on newer GPUs, FreeSync, HDMI 2.0 features, atomic mode-setting, and more, but it's a huge amount of code and it remains up in the air when it will land.
The material to be found in this initial round of updates include Vega bug fixes, preliminary support for Raven Ridge APUs, KIQ support for compute rings, MEC queue management rework, audio support for DCE6 hardware (original Southern Islands GPUs), SR-IOV improvements, bug fixes, and other clean-ups.
This pull also has improved module parameters for being able to control easily whether to load the Radeon or AMDGPU DRM drivers for GCN 1.0/1.1 hardware. The module parameters are si_support and cik_support for being able to choose at boot-time (if AMDGPU was built with SI/CIK support) to boot instead to the AMDGPU driver rather than Radeon DRM.
The list of changes for this initial pull request of mostly AMDGPU material for Linux 4.13 can be found via this pull request.
| 36
| 1,760,719,733.265744
|
https://www.phoronix.com/news/Ryzen-Compiler-Issues
|
Some Ryzen Linux Users Are Facing Issues With Heavy Compilation Loads
|
Michael Larabel
|
I haven't encountered this issue myself on any of my Ryzen Linux boxes, but it seems there are a number of Ryzen Linux users who are facing segmentation faults and sometimes crashes when running concurrent compilation loads on these Zen CPUs.
A Phoronix reader pointed out some of the resources for Ryzen Linux customers facing problems namely when running heavy compilation tasks, like on Arch and Gentoo. AMD hasn't yet found the root cause of this issue, but given the spread of users affected, appears to be related to the processor itself.
Those interested in learning more about these Ryzen compilation issues can find a number of open threads on the matter such as on the Gentoo forums, AMD Community, as well as some entries via this Google Doc tracking Gentoo users having the problem.
AMD is expected to update their community thread when a solution is found. Some workarounds include fiddling with Load Line Calibration (LLC) from the BIOS and some users have found success if disabling the SMT functionality while others are still encountering the problems even if they turn off SMT on their Ryzen 7 CPUs. The issue is happening on multiple versions of GCC but I haven't seen any reports when using LLVM/Clang or alternative compilers.
| 149
| 1,760,719,734.456687
|
https://www.phoronix.com/news/Ryzen-Price-Cuts
|
AMD Begins Cutting Prices On Ryzen CPUs
|
Michael Larabel
|
If you have been waiting to pick up an AMD Ryzen CPU until the prices drop, they are beginning to do so.
It's not clear if they are permanent price cuts, but likely so, given these drops come just days after Intel announced its new high-core competition with the Core-X Series that will be shipping in June.
The price cuts so far entail the Ryzen 7 1800X dropping from $499.99 USD to $459 (or $439 at Amazon), the Ryzen 7 1700X from $399 to $349 and the Ryzen 7 1700 from $329 to $314 (or at Amazon is now $299). These are some pretty nice discounts considering the Ryzen 7 CPUs still appear to be selling well and are still some competitive offerings. The Ryzen 5 CPUs haven't received much in the way of a price adjustment yet.
The marked-down Ryzen 7 CPUs can be found at Amazon.com and NewEgg. If now picking up a Ryzen 7 processor from either shop, we kindly ask you use our affiliate links above as a portion of the sale then goes on to support Phoronix and our Linux hardware testing operations.
See our dozens of articles on Ryzen Linux coverage and benchmarks if you have questions about using these Zen CPUs under open-source or swing by the forums if you have further questions.
| 30
| 1,760,719,734.593627
|
https://www.phoronix.com/news/AMD-Seattle-Cello-Rich
|
AMD LeMaker Cello Board Finally Ships To Some
|
Michael Larabel
|
With all of AMD's excitement these days about their Zen-based Ryzen/EPYC processors and forthcoming Vega GPUs, you probably forgot about their ARM efforts that they appear to have pretty much abandoned. But it looks like some of those who pre-ordered the AMD Seattle powered LeMaker Cello board are finally receiving their kits.
Back in March 2016 is when the LeMaker Cello board was announced with the AMD A1100 SoC, 2 x DDR3 SO-DIMM slots, two Serial ATA 3.0 ports, Gigabit Ethernet, two USB 3.0 ports, one PCI Express 3.0 x16, etc. But months passed and there have been no signs of the LeMaker Cello or the long-awaited Husky Board.
We then heard of delays for AMD's ARM efforts and most recently, their ARM-based efforts appear stalled in favor of focusing on their Zen x86 offerings.
Coming as much surprise today, Red Hat's Richard Jones who has long specialized in ARM Linux has mentioned his Cello board arrived. But he too acknowledges AMD appears to have given up on ARM so the board is now there for "historical curiosity" and he also notes the large heatsink on the board he receives, possibly indicating thermal issues.
The LeMaker Cello order page shows that there have been just 47 buyers but there is "no stock" available for purchase. 96Boards.org still shows this board as "coming soon", albeit linking to the Lenovator page showing no stock. Meanwhile, the 96Boards HuskyBoard now shows that this AMD ARM board is "not under development", further looking like AMD's ARM efforts have gone the way of Calxeda.
| 17
| 1,760,719,735.88159
|
https://www.phoronix.com/news/AMD-EPYC-Launch-Coming
|
AMD EPYC Launching 20 June, Are You Interested?
|
Michael Larabel
|
Besides confirming the RX Vega launch for SIGGRAPH, AMD also announced today from Computex Taipei that their AMD EPYC launch is happening on 20 June.
EPYC, for the forgetful, is AMD's Zen-based server offering and what was formerly codenamed Naples. AMD EPYC CPUs are launching on 20 June with hard availability expected that day.
EPYC will provide up to an impressive 32 cores / 64 threads and up to 32MB L3 cache and 128 PCI Express lanes, but AMD is still light on details until the launch day.
Unfortunately, no word if we will be receiving any EPYC review samples for Linux/BSD benchmarking at Phoronix... Presumably many of you are interested in EPYC if working in the server/workstation fields and silly not to seed such hardware to a Linux site, but I digress. If we don't receive any review samples, I'll be weighing the possibility of picking up some EPYC hardware if the price isn't too bad and there is enough reader interest on Phoronix. So if you are interested in seeing EPYC Linux benchmarks, be sure to comment in the forums, especially if you are a premium supporter.
It will be an interesting June and at least there will be plenty of Core X-Series coming up on Phoronix regardless of any hopeful EPYC fun.
| 41
| 1,760,719,736.050251
|
https://www.phoronix.com/news/AMD-15-More-RADV-Patches
|
AMD Developer Posts More Patches For RADV
|
Michael Larabel
|
The state of the open-source RADV Vulkan driver remains rather murky with it not officially being supported by AMD while the company continues to back its still-proprietary multi-platform Vulkan driver with no signs of when it may be open-sourced, but an AMD developer posted some fresh RADV patches today.
More than one year past the Vulkan launch and when AMD debuted its Vulkan driver, which at the time they mentioned the possibility of it being open-sourced within six months or so, we haven't seen that big code drop reached yet. Meanwhile, the RADV driver within Mesa continues getting better with developers from Red Hat, Google, and others contributing to it. The performance is still improving, but they recently crossed the milestone of effectively being Vulkan 1.0 compliant. While with AMD's Vulkan driver it's compliant but they haven't dropped the code yet, is not integrated with Mesa but expected to be its own standalone repository, and we don't know much more yet about their grand plans.
But it's interesting waking up today seeing 15 more patches from AMD's Nicolai Hähnle to RADV. These 15 patches to RADV range from fixes to making use of more shared/common code with RadeonSI Gallium3D. These latest patches can be found on Mesa-dev. The patches were signed-off using Nicolai's AMD email address as opposed to his personal email. It's not very common seeing AMD contributing to this unofficial Vulkan driver, but maybe they are changing their Vulkan Linux path? Hopefully we'll get some more clear communication or action soon. Just a pity how long it's taken AMD with their open-source Vulkan ambitions.
| 36
| 1,760,719,737.332227
|
https://www.phoronix.com/news/AMD-Financial-Day-2017
|
AMD Talks Up Vega Frontier Edition, Epyc, Zen 2, ThreadRipper
|
Michael Larabel
|
Today was AMD's annual Financial Analyst Day where they revealed Zen CPU and Vega GPU details.
- AMD confirmed "Zen 2" is in development and will be based on a 7nm process. There's also Zen 3 that will be at 7nm or better. But don't get too excited, Zen 2 won't be out until at least late 2018.
- Navi, the GPU successor to Vega, will also be on a 7nm manufacturing process.
- The Mobile Ryzen platform was announced for 2-in-1s, ultra-portables, and more. Exciting about Mobile Ryzen is that it will feature Vega integrated graphics.
- Following the rumors, AMD confirmed the "Ryzen ThreadRipper" branding for their parts to feature 16 cores plus SMT to allow for 32 threads. There will also be quad-channel DDR4, more PCI-E lanes, and more I/O. Ryzen ThreadRipper will be a competitor to Intel's upcoming Skylake-X processors.
- Vega Frontier Edition was announced as a Vega-based GPU for HPC computing, data scientists, and other GPGPU computing purposes. Vega Frontier Edition with 16GB of HBM2 memory is said to be 3x faster than a Fury X.
- AMD's Zen-based "Naples" chip for servers is going to be branded as AMD Epyc.
More details forthcoming.
| 31
| 1,760,719,737.540613
|
https://www.phoronix.com/news/AMDGPU-Raven-Enablement
|
AMD Posts Initial Patches For Raven APU Support: 117 Patches, 314k L.O.C.
|
Michael Larabel
|
Alex Deucher of AMD has posted the initial AMDGPU DRM/KMS kernel driver patches for bringing up graphics on the next-generation "Raven" APUs.
Raven Ridge is expected later this year as the first Zen-based APUs. These DDR4-supported Zen APUs are expected for release in Q4 but there have been some indications it might happen a bit earlier. Raven Ridge graphics make use of AMD's new Vega architecture, which an APU in the form of Zen+Vega should be quite exciting if the price and performance are on point.
While Vega has already been enabled in AMDGPU, the Raven (Ridge) support published today adds in 314,946 lines of code across 117 patches. The vast majority of the new code though (almost 200k LOC) is in the form of header files for the registers.
The patches do confirm there is a new ACP audio with Raven Ridge as well as a new media block (VCN) that appears to replace UVD and VCE. At least initially there is just one PCI device ID for Raven Ridge graphics: 0x1002. Like Vega, the Raven Ridge graphics support depend on the DC (DAL) code.
The initial Raven graphics patches can be found on the mailing list. Hopefully we'll see these patches primed for Linux 4.13 so that there is good open-source graphics driver support by the time these APUs ship late in the year -- and hopefully we'll see DC finally land in 4.13 too.
| 16
| 1,760,719,738.936171
|
https://www.phoronix.com/news/AMD-AOCC-1.0-Released
|
AMD Releases Optimizing C/C++ Compiler For Ryzen
|
Michael Larabel
|
Longtime Phoronix readers and AMD Linux enthusiasts probably remember the AMD Open64 compiler for past CPU launches with various compiler optimizations for AMD processors. With Open64 being dead and all the compiler rage these days about LLVM/Clang, AMD has announced the "AMD Optimizing C/C++ Compiler" (AOCC) that's based upon Clang and optimized for Ryzen/Zen processors.
AMD this week has rolled out AOCC 1.0 as their first compiler optimized for Zen. LLVM Clang as well as GCC have offered mainline support for Ryzen via znver1 but AMD hadn't contributed a scheduler model yet for this upstream compiler support. Now they have out AOCC 1.0 for those wanting further optimized Zen binaries.
AOCC 1.0 is focused on Zen/17h processor support and reportedly offers improved vectorization, higher-level optimizer, and better code generation. AOCC also supports Fortran codes via the DragonEgg LLVM plug-in for integration with GCC. These various AMD optimizations are patched atop LLVM Clang 4.0. AMD has also released an AOCC optimized version of the Gold linker.
Download links for this AMD Optimizing C/C++ Compiler can be found via developer.amd.com. I haven't yet tested it but will be doing some fresh Ryzen GCC/Clang/AOCC benchmarks now this week.
The support for DragonEgg for AMD-optimized Fortran is interesting since upstream DragonEgg has basically been unmaintained for a while. This AOCC DragonEgg will also only work with the (now rather older) GCC 4.8.2 compiler.
Hopefully AMD will be able to upstream their relevant AOCC patches to LLVM/Clang in the near future... As of writing I haven't seen a source tree yet of the AOCC changes. With the AOCC binaries, they are still named "clang" for an easy drop-in replacement on Linux systems.
| 61
| 1,760,719,739.095301
|
https://www.phoronix.com/news/AMD-CodeXL-2.3
|
GPUOpen's CodeXL 2.3 Brings Ryzen Support, AMDGPU-PRO Compatibility
|
Michael Larabel
|
AMD's CodeXL utility that's open-source under the GPUOpen umbrella for graphics profiling/debugging is up to version 2.3.
CodeXL 2.3 adds support on Linux systems for operating with the AMDGPU-PRO hybrid driver. Other prominent features include Radeon Polaris GPU support as well as support for AMD Ryzen processors with the addition of supporting its performance counters, etc.
CodeXL 2.3 also has a number of CPU, power, and GPU profiling improvements. On the GPU profiling front there is ROCm 1.5 support. The GPU debugger back-end in CodeXL has removed its HSA debugging support and it has also disabled OpenCL kernel debugging support for the new OpenCL stack with LLVM.
Red Hat Enterprise Linux 7, Ubuntu 16.04, and SUSE Linux Enterprise 11 are the officially supported Linux distributions with CodeXL. Those wishing to learn more about CodeXL 2.3 can visit the project's GitHub page.
| 0
| 1,760,719,740.296878
|
https://www.phoronix.com/news/Radeon-RX-550-First-Try
|
Radeon RX 550 Stumbles On Open-Source, Working Fine With AMDGPU-PRO
|
Michael Larabel
|
This week I decided to pick up the Gigabyte Radeon RX 550 2GB graphics card for Linux testing at Phoronix, a $90 USD graphics card that was recently launched as part of the "Polaris Evolved" line-up. It's not working on the upstream open-source code-base at the moment, but at least does function with the latest AMDGPU-PRO 17.10 driver for the RX 500 series.
I only received the Gigabyte RX 550 this morning, so I don't yet have any benchmark results to share, but when being surprised by 3D acceleration not working out-of-the-box with the latest upstream code, I decided to write this preliminary article first while the Linux OpenGL/Vulkan test results will follow.
The Gigabyte Radeon RX 550 Gaming OC 2GB (RX550GAMINGOC-2GD) card retails for $89.99 USD on Amazon where I had purchased this card. This RX 550 card runs up to 1219MHz in its OC mode or 1206MHz in its gaming mode.
The past day waiting for this card I had been testing other low to mid-range Radeon cards in my possession from the Radeon R7 260X, R7 370, RX 460, RX 480, and several other cards including the RX 580. They all played nicely with the Linux 4.12 DRM-Next + Mesa 17.2-dev stack via the Padoka PPA. Even with last month's Radeon RX 580 launch that played nicely on Linux out-of-the-box.
So after the RX 580 testing last month for the MSI card I bought, I wasn't expecting any headaches with the RX 550 with it being another Polaris part, albeit part of the Polaris 12 family. The Gigabyte Radeon RX 550 has a PCI ID of 0x699F with C7 revision.
When booting this card on the same driver stack as used with my other Radeon tests over the past day, the AMDGPU DRM driver initialized fine, the card booted fine with the 4K DP display (no worries, during this low-end testing it's happening at 1080p), but quickly realized hardware acceleration wasn't working and instead the LLVMpipe fallback driver.
Then falling back to Linux 4.10 and 4.11 kernels as well as then building libdrm 2.4.80 and Mesa 17.2 Git from source rather than the Padoka PPA and also ensuring the latest linux-firmware.git for safe measure, the RadeonSI driver was still not loading. (And in the process also noticing libdrm's AMDGPU PCI ID table isn't being kept up to date with the latest marketing names.)
A pity considering the Radeon RX 580 was playing nicely out-of-the-box (including on this same system / software stack) and other recent Radeon GPU launches relatively smooth for being able to jump straight to benchmarking. Anyhow, I did decide to try the latest AMDGPU-PRO driver release and at least it does work fine with this RX 550:
Due to the courier only delivering the card a short time ago, I haven't dug much deeper yet but will be doing so, stay tuned for follow-up articles.Update: There was a bug affecting Polaris 12 within the AMDGPU winsys code. There's now a fix in Mesa Git.
| 19
| 1,760,719,740.547648
|
https://www.phoronix.com/news/AMDGPU-Linux-4.12-Fixes-1
|
AMDGPU Gets More Fixes For Linux 4.12
|
Michael Larabel
|
While the DRM-Next merge window is over for new feature material ahead of the Linux 4.12 merge window opening tomorrow, some AMDGPU fixes have been sent out for this next kernel cycle.
The biggest addition to AMDGPU DRM-Next for Linux 4.12 is initial AMD Vega 10 support but for this initial support there is no physical display support until Linux 4.13 or later when DC lands. Aside from the initial Radeon RX Vega support there is also some possible performance improvements, PRT support for sparse buffers, job tracing improvements, SR-IOV improvements, GPU sensor work, and more.
With Friday's fixes pull there are "lots of Vega10 fixes" ahead of the Radeon RX Vega launch next month, some display fixes for both Radeon and AMDGPU, and a variety of other fixes.
There are dozens of changes/fixes in total with this fixes pull request touching nearly two thousand lines of code. More details on these changes via the dri-devel list.
| 1
| 1,760,719,741.410934
|
https://www.phoronix.com/news/AMD-More-Open-Devs
|
AMD Is Hiring More Developers For Their Open-Source Graphics Team
|
Michael Larabel
|
Following news of Red Hat hiring another developer to work on open-source graphics compute, AMD is now hiring at least two more developers too.
In response to the talk of Red Hat hiring for open-source GPGPU work, AMD's Tom Writer commented in our forums:
AMD is also hiring! I am actively interviewing candidates for two open positions for Mesa (GL and multimedia) work. If you have relevant experience, esp. a track record of contributions to Mesa or other open source graphics projects. More details via this job listing.
| 24
| 1,760,719,741.834037
|
https://www.phoronix.com/news/AMDGPU-PRO-RX500
|
AMDGPU-PRO Updated With Radeon RX 500 Series Support
|
Michael Larabel
|
AMD has posted an updated AMDGPU-PRO 17.10 driver with support for the RX 500 "Polaris Evolved" graphics cards.
This doesn't appear to be a big driver update with AMDGPU-PRO 17.10 having been released earlier this month with Ubuntu 16.04.2/16.10 as its primary feature. The new driver out now lists the same changes but also now mentions the RX 500/570/580 support.
The driver can be downloaded from AMD.com. My Radeon RX 580 Linux testing will begin in a few hours with the card being out now for delivery. Stay tuned.
| 10
| 1,760,719,743.088235
|
https://www.phoronix.com/news/AMD-Ryzen-5-Ships
|
AMD Ryzen 5 Begins Shipping
|
Michael Larabel
|
Today marks AMD formally launching the Ryzen 5 line-up with immediate availability.
Models include the Ryzen 5 1400 (quad-core + SMT, 3.2GHz base, 3.4GHz turbo), Ryzen 5 1500X (quad-core + SMT, 3.5GHz base, 3.7GHz turbo), Ryzen 5 1600 (six core + SMT, 3.2GHz base, 3.6GHz turbo), and Ryzen 5 1600X (six-core + SMT, 3.6GHz base, 4.0GHz turbo). The TDPs are 65 Watt except for the Ryzen 5 1600X at 95 Watts. Pricing is at $169 for the Ryzen 5 1400, $189 for the Ryzen 5 1500X, $219 for the Ryzen 5 1600, and $249 for the Ryzen 5 1600X.
AMD has not supplied any review samples to Phoronix of Ryzen 5 parts, but you can see our past Linux tests of Ryzen 7. I am currently contemplating buying a Ryzen 5 1400 or more likely the Ryzen 5 1500X... Share your interest if wanting to see some of those benchmarks on Phoronix. Will likely decide in the next few hours with there being availability this morning from NewEgg and Amazon.
More details on Ryzen 5 at AMD.com.
| 45
| 1,760,719,743.536644
|
https://www.phoronix.com/news/AMDGPU-PRO-17.10
|
AMDGPU-PRO 17.10 Released With Ubuntu 16.04.2 Support
|
Michael Larabel
|
AMD has released the AMDGPU-PRO 17.10 hybrid Linux graphics driver.
The only "highlight" listed for AMDGPU-PRO 17.10 is Ubuntu 16.04.2 LTS support. But that itself is significant since if they are referring to 16.04.2's hardware enablement stack, then AMDGPU-PRO is now working on at least the Linux 4.8 kernel without issues. I am in the process of setting up AMDGPU-PRO 17.10 on my Radeon test systems to see if Linux 4.8 is working nicely and if Linux 4.10~4.11 will work. This should also mean AMDGPU-PRO 17.10 should play nicely now on Ubuntu 16.10 due to sharing the same components of the 16.40.2 HWE stack.
The listed fixed issues are for the system failing to boot RHEL 7.3 with DisplayPort 1.2 enabled and intermittent screen corruption when manually switching to the AMD performance mode.
That's it for the listed features via the AMD.com driver page. I'll be running AMDGPU-PRO 17.10 Linux driver benchmarks this weekend.
| 73
| 1,760,719,744.433257
|
https://www.phoronix.com/news/AMDGPU-Fixes-Linux-4.12
|
AMDGPU Has More Code Prepped For Linux 4.12
|
Michael Larabel
|
AMD already sent in their major feature pull request to DRM-Next of new AMDGPU/Radeon DRM material slated for Linux 4.12 while another pull has now landed for -next.
The main feature pull of AMDGPU Linux 4.12 material included initial Vega 10 support (but no display capabilities), PRT support for sparse buffers, SR-IOV improvements, GPU sensor work, job tracing improvements, TTM memory management enhancements, and other work.
The latest work for AMDGPU DRM-Next is TTM and AMDGPU support for non-contiguous video RAM CPU mappings, many fixes/clean-ups for Vega 10, and various other bug fixes and code clean-ups. This latest pull request includes around five dozen changes affecting just under one thousand lines of code.
The complete list of changes for this pull request can be found on dri-devel. This is the last major feature work expected of AMDGPU/Radeon for Linxu 4.12 with the DRM-Next deadline being this weekend.
| 0
| 1,760,719,745.502426
|
https://www.phoronix.com/news/AMD-Chromebook-Kahlee
|
Google Kahlee: The First AMD-Powered Chromebook
|
Michael Larabel
|
After years of many Intel and ARM Chromebooks, the first AMD-powered Chromebook appears to be gearing up for release.
Continuing with tradition, the AMD Chromebook is using Coreboot. Thus we have the first signs of it via Coreboot code review with this new Google board being codenamed "Kahlee." The Coreboot code began appearing for review just minutes ago after other Kahlee references in the Chrome OS world have been found in recent weeks.
This AMD Chromebook features two M.2 slots and is based on an AMD Stoney Ridge APU. AMD Stoney Ridge for the forgetful is their 2016 platform based on a 28nm process and using Excavator CPU cores with 3rd gen GCN graphics.
More details as they become available.
| 35
| 1,760,719,746.064432
|
https://www.phoronix.com/news/Ryzen-7-1700-B350-v1.2-DDR4
|
AMD Ryzen 7 1700 + B350 DDR4 Memory Speed Tests
|
Michael Larabel
|
Earlier this week I posted some Ryzen 7 1800X DDR4 memory scaling Linux tests now that MSI pushed out an updated BIOS for that X370 motherboard that allows running the system at higher -- but still rather limited -- DDR4 memory frequencies. Here are some similar tests with my Ryzen 7 1700 and a B350 motherboard.
The AMD Ryzen 7 1700 is currently paired with a MSI B350 TOMAHAWK motherboard. MSI recently released a v1.2 BIOS update that improves memory compatibility and system stability. With that board update, the 2 x 8GB DDR4-3000 Corsair memory modules can now run higher than DDR4-2133, but they don't yet work in their AMP mode to push the modules to 3000MHz -- the system is unbootable until clearing the CMOS. Even when trying for DDR4-2933, which was successful with the Ryzen 7 1800X + MSI X370 board, wasn't possible with this current BIOS. The highest I could boot with this Ryzen 7 1700 configuration on this latest BIOS was DDR4-2667.
So for these Friday benchmarks are just some Ryzen 7 1700 tests on Ubuntu Linux when running at DDR4-2133. DDR4-2400, and DDR4-2667.
The Linux gaming tests are basically a wash.
The Stream synthetic RAM test results are clear.
The other results are largely tight due to the limited RAM frequency selection available. All the data via this OpenBenchmarking.org result file. Hopefully MSI will release an improved BIOS soon for finally hitting DDR4-3000+ with Ryzen.
| 11
| 1,760,719,747.012536
|
https://www.phoronix.com/news/AMD-17-More-DC-Patches
|
17 Fresh AMDGPU DC Patches Posted Today
|
Michael Larabel
|
Seventeen more "DC" display code patches were published today for the AMDGPU DRM driver, but it's still not clear if it will be ready -- or accepted -- for Linux 4.12.
AMD developers posted 17 new DC (formerly known as DAL) patches today to provide small fixes for Vega10/GFX9 hardware, various internal code changes, CP2520 DisplayPort compliance, and various small fixes.
These latest patches can be found for review via amd-gfx list. What remains to be seen though is if AMDGPU DC will be ready for Linux 4.12, which is now more pressing since Radeon Vega support needs it. This weekend we'll see Linux 4.11-rc4 released and so that probably leaves two weeks or less (especially with Linus Torvalds being upset by late DRM-Next merges, especially after the 4.11 merge window controversy) for AMD developers to try to push the DC code through another final review and see if a pull request into DRM-Next would be honored by David Airlie.
Aside from being needed for Vega support, AMDGPU DC finally lays the framework for other modern display features like HDMI 2.0, FreeSync (although that's not fully wired up yet on the open code), HDMI/DP audio support, and atomic mode-setting. Stay tuned to Phoronix to find out if it gets lined up for Linux 4.12 or will be held off for yet another cycle.
| 16
| 1,760,719,747.401345
|
https://www.phoronix.com/news/AMDGPU-Vega-10-Support
|
AMD Sends Out 100 Patches, Enabling Vega Support In AMDGPU DRM
|
Michael Larabel
|
100 patches amounting to over fourty thousand lines of code was sent out today for review in order to provide "Vega 10" support within the AMDGPU DRM driver.
Adding Vega support to AMDGPU is a big task due to all of the changes over Polaris and other recent GPUs. Vega rolls out a new video BIOS interface, lots of new hardware intellectual property, support for video decode using UVD (UVD 7.0), support for video encode using VCE (VCE 4.0), support for 3D via RadeonSI, power management, full display support using DC, and support for SR-IOV virtualization.
This 40k+ lines of code is on top of the tens of thousands of lines of code needed for DC/DAL. AMDGPU Vega's display support is built on top of DC and thus that will need to land in order to allow this new hardware support to happen.
Seven device PCI IDs are currently listed for Vega. AMD is currently rumored to be launching AMD Radeon RX Vega in late May.
The 100 patches up for review can be found on amd-gfx. I'm still digging through the patches and will update if I find any other interesting remarks. The next opportunity to merge this code will be Linux 4.12, assuming its review goes well and AMDGPU DC gets accepted to mainline.
Update: Patches are now available for the Vega support in Radeon Gallium3D as the main user-space component.
| 24
| 1,760,719,748.680826
|
https://www.phoronix.com/news/Dota-2-Ryzen-Optimizations
|
Dota 2 Receives Optimization For AMD Ryzen CPUs
|
Michael Larabel
|
If you were an early buyer of AMD Ryzen hardware, Valve has pushed out a Dota 2 game update with some Ryzen optimizations.
Today's Dota 2 update from Valve mentions, "Improved threading configuration for AMD Ryzen processors." Presumably this is with better dealing of Ryzen's new SMT capabilities for AMD processors.
The list of changes with today's Dota 2 game update can be found via Dota2.com.Unfortunately I don't believe I have any old installations of Dota 2 on my Ryzen Linux boxes so probably won't be able to deliver any before vs. after data points, but will see if my archived test results on OpenBenchmarking.org have any comparable metrics to use for some fresh Ryzen Linux gaming tests.
| 18
| 1,760,719,748.907921
|
https://www.phoronix.com/news/RADV-Conformance-Tests-54.5
|
David Airlie Tackling RADV Vulkan Conformance
|
Michael Larabel
|
RADV co-founder David Airlie at Red Hat has begun focusing on the Vulkan conformance test suite for furthering along this open-source Radeon driver's conformance.
With the latest RADV code in Mesa Git, RADV is up to 54.7% of the tests passing in The Khronos Group's Vulkan Conformance Test Suite via VK-GL-CTS. 45.3% of the tests are not supported in the current stage of RADV.
Airlie commented on his blog post about the RADV conformance matter, "That is pretty conformant (in fact it would pass as-is). However I need to clean up the patches in the branch and maybe figure out how to do some bits properly without hacks (particularly some semaphore wait tweaks), but that is most of the work done."
AMDGPU-PRO has passed its conformance testing and is conformant for Vulkan but great to see RADV making so much progress without the official support of AMD.
| 120
| 1,760,719,750.006092
|
https://www.phoronix.com/news/AMD-Ryzen-5-April
|
AMD Ryzen 5 CPUs Launching Next Month, 4 & 6 Core Options
|
Michael Larabel
|
AMD has confirmed that Ryzen 5 CPUs will begin shipping on 11 April.
For those that were intrigued by the Ryzen 7 line-up but looking for something cheaper and with less cores/threads, the Ryzen 5 will be available in four and six core options in less than one month. On the low-end is the 4-core Ryzen 5 1400 @ 3.2/3.6GHz for $169 USD while the Ryzen 5 1600X is 6-cores at 3.6/4.0GHz for $249 USD. The Ryzen 5 CPUs still feature SMT.
I haven't heard from AMD yet whether they will be sending out review samples, but otherwise I plan to at least pick up one of the lower-end Ryzen CPUs myself for some interesting Linux benchmarks. By April hopefully we'll see out some Ryzen microcode updates, new motherboard BIOSes, and better supplies of Ryzen motherboards. Stay tuned for more Ryzen 7 1700 / 1800X benchmarks though still coming up in the days ahead.
| 51
| 1,760,719,751.351802
|
https://www.phoronix.com/news/Ryzen-7-1800X-OPM
|
Hammering The AMD Ryzen 7 1800X With An Intense, Threaded Workload
|
Michael Larabel
|
Today I got around to running a very heavy/demanding, very real-world workload on the AMD Ryzen 7 1800X that I've been meaning to test with this Zen CPU.
The workload I've been running on the Ryzen 7 1800X the past few hours is that of Open Porous Media, the open-source OPM project is a growing initiative around research and simulators for modeling and simulation of porous media processes, including a reservoir simulator and permeability upscaling. This sort of workload has relevance in areas like oil and natural gas industries.
While it's certainly a real-world workload, running it can take quite some time depending upon the thread configuration, and thus isn't tested as part of all our Linux hardware reviews at Phoronix (as well as having a long dependency chain), but figured it would be fun seeing how it performs with Ryzen.
I compared the AMD Ryzen 7 1800X performance to a few ouf our systems in the test farm including the Xeon E3-1270 v5, E3-1231 v3, and Core i7 5960X.
First up was trying OPM's Flow simulator, "Flow is a reservoir simulator for three-phase black-oil problems using a fully-implicit formulation. There are also specialized variants for solvent and polymer problems."
With just one thread, the Ryzen 7 1800X was running behind the Xeon E3-1270 v5. The E3-1270 v5 has a 3.6GHz base frequency with 4.0GHz turbo boost and has four cores plus Hyper Threading. These clock frequencies match that of the Ryzen 7 1800X with its base and turbo frequencies, but the E3-1270 v5 sells for around $300 USD while the 1800X is around $500. At least clock-for-clock, the Ryzen 7 1800X is right near the E3-1270 v5 Skylake.
When running the Flow MPI Norne test with two threads, the Ryzen 7 1800X was able to come out ahead of the E3-1270 v5 as well as the Core i7 5960X (and obviously, the E3-1231 v5 too).
But when hitting four threads is when the Core i7 5960X began pulling out ahead.
And at eight threads, the Core i7 5960X was remaining faster than the Ryzen 7 1800X by a noticeable margin. Unfortunately I don't have Open Porous Media results to share today from more systems due to the lengthy dependency requirements for these simulators as well as the significant amount of time needed to execute.
Lastly for this simulator was the 16 thread run, between the Core i7 5960X and Ryzen 7 1800X for both sporting 16 threads. The i7-5960X was significantly faster than the Ryzen 7 CPU tested.
Next was OPM's Upscaling test, "The Upscaling module contains programs that can do flow-based permeability upscaling as well as upscaling of relative permeability and capillary curves, using a steady-state approach."
At one thread, the upscaling test showed similar single-core performance between the Ryzen 7 1800X and Xeon E3 1270 v5 that are clocked the same, but with this Xeon E3 part retailing for almost $200 less for this quad-core + HT workstation CPU.
And with two threads is where the Ryzen 7 strikes a win over the tested Intel CPUs.
But with more than two threads is when the Core i7 5960X showed its strength.
And 16 threads showed the Core i7 5960X continuing to lead.
Those wanting to try running the OPM benchmark on their systems can install the Phoronix Test Suite and run phoronix-test-suite benchmark 1703160-RI-AMDRYZENO41.
| 38
| 1,760,719,751.56901
|
https://www.phoronix.com/news/Tom-Stellard-At-RedHat
|
AMDGPU LLVM Expert Has Left AMD, Now Working For Red Hat
|
Michael Larabel
|
A Phoronix reader pointed out an interesting reference on a Mesa patch today... Tom Stellard, the former GSoC student who was instrumental in developing the AMDGPU LLVM compiler back-end, is now working for Red Hat.
We haven't been able to find any announcement by Tom, but this RADV patch shows him reviewing a patch with a RedHat.com email address.
Tom Stellard has long been involved with the Radeon LLVM efforts and he also has served as the LLVM point release manager for getting out official bug-fix updates to LLVM.
He's been working on the open-source Radeon stack going back to the R300 Gallium3D days back in 2010 and has been heavily involved ever since.
It's interesting to see him shift his place of employment now from AMD to Red Hat. It's good to see at least he's continuing to work on the open-source Radeon stack given today's patch review. This also increases curiosity over Red Hat's work on the open Radeon driver stack with David Airlie continuing to invest significant effort into developing the RADV driver as well as Valve's continued open-source driver involvement on RadeonSI and RADV outside of AMD.
| 24
| 1,760,719,752.639107
|
https://www.phoronix.com/news/AMDGPU-Prep-Patches-For-Vega
|
AMD Sends Out Prep AMDGPU Patches For New GPUs
|
Michael Larabel
|
In the early hours of today AMD posted a set of 23 AMDGPU patches as "prep patches for new ASICs", which given the timing, is presumably prepping for the Radeon RX VEGA.
But before getting too excited, there isn't any new GPU support code as part of these 23 patches that touch several hundred lines of code. These patches are just prepping the driver infrastructure for being able to handle AMD's new GPUs but without actually adding in any new support at this time.
There are a fair amount of UVD/VCE changes along with other low-level alterations in prepping for new hardware support. Hopefully this is a sign it won't be too much longer before seeing Vega support patches published for the AMDGPU+RadeonSI open-source driver stack.
The earliest we'll see AMD Vega GPU support land would be for Linux 4.12, with the merge window for that likely being in late April or perhaps early May. Linux 4.12.0 stable would then be released in late June or early July. The big contingency though on the AMD Vega support is needing to land the AMDGPU DC/DAL display code, as AMD developers have previously indicated their new hardware will only be making use of the new display code paths. Latest rumors are that AMD Radeon RX Vega may be officially launching on 30 May.
The 23 prep patches for new ASICs can be found via this patch series.
| 18
| 1,760,719,752.913738
|
https://www.phoronix.com/news/Ubuntu-AMDGPU-DC-Kernel
|
An Experimental Ubuntu Kernel Build With AMDGPU DC/DAL
|
Michael Larabel
|
A Phoronix reader has written in about his independent work to make it easier trying out the latest AMDGPU DC/DAL code on Ubuntu.
Martin Babutzka has taken to maintaining a Linux kernel tree on GitHub with the latest AMDGPU DC/DAL patches pair with the latest upstream Linux kernel security patches, based upon the upstream Linux 4.9 kernel. He's carrying out this effort in order to make use of HDMI audio with his Radeon R9 380 graphics card.
Besides offering the source tree, he's also offering pre-built x86_64 Debian/Ubuntu packages of the said kernel.
Of course, try at your own risk, but an easy way to try out AMDGPU DC/DAL support for HDMI/DP audio on newer GPUs, HDMI 2.0 features, the start on atomic support, and other modern display features are sought out by many Phoronix readers.
His kernel tree can be found at M-Bab/linux-kernel-amdgpu and there is also the AMDGPU kernel binaries.
The AMDGPU DC/DAL code isn't being mainlined until the Linux 4.12 kernel at the earliest, thus the Ubuntu 17.04 release in April, which is using Linux 4.10, will still be without this display code.
| 17
| 1,760,719,754.117477
|
https://www.phoronix.com/news/Ryzen-7-Linux-4.11
|
Linux 4.11 Doesn't Change The Game For AMD's Ryzen
|
Michael Larabel
|
Linux 4.11 is worthwhile in that it's bringing ALC1220 audio support, the codec used by many Ryzen (and Intel Kabylake) motherboards, but this next kernel version doesn't appear to change Ryzen's performance.
I didn't see anything notable this Linux 4.11 merge window with regard to Ryzen for potentially affecting its performance, but I ran some benchmarks this weekend just to confirm.
With the Ryzen 7 1700 configuration on the MSI X370 XPOWER GAMING TITANIUM configuration I ran some benchmarks between Linux 4.10 and then Linux 4.11 Git as of yesterday.
There are potential gaming performance improvements to find on Linux 4.11, that's when using the AMDGPU DRM driver and in particular AMDGPU changes in 4.11 appear to mostly benefit the RADV Vulkan driver.
In most OpenGL games, the performance is the same.
Nothing really to see on the Ryzen performance side for Linux 4.11. Those that were hoping for any magical scheduler improvements or other refinements to Ryzen's Linux support will need to wait another couple months.
| 23
| 1,760,719,754.258724
|
https://www.phoronix.com/news/AMD-Ryzen-No-Coreboot-ATM
|
AMD Ryzen/Zen Currently Doesn't Support Coreboot Today
|
Michael Larabel
|
Back in 2011 was the glorious announcement that AMD would support Coreboot with its future CPUs. Sadly, a lot has changed at AMD over the past half-decade, and there isn't any Coreboot support to find today for Zen/Ryzen.
It was looking like AMD Zen wouldn't see Coreboot support and that is indeed the case, at least for now, with there being no support for AMD Ryzen processors or the new AMD chipsets within Coreboot. This though doesn't come as too much of a surprise as AMD stopped providing their AGESA code back in 2014 and there's been no recent communication out of AMD regarding Coreboot. The company AMD used to work with on Coreboot, Sage Engineering, is also no longer in business.
That's not to say things couldn't change in the future, but today there is no Ryzen Coreboot support and have confirmed there are no imminent plans, at least from developers in the know.
Perhaps we will see a change of course once low-power Ryzen APUs begin shipping if they will go for a Chromebook/Chromebox play with Google using Coreboot for their devices... But then again, this could also serve as a blocker otherwise.
In somewhat related news, Edward Snowden today called for AMD to open-source their PSP and firmware. On reddit they were also asked about opening up their firmware to which there is nothing on their books today.
| 37
| 1,760,719,755.447544
|
https://www.phoronix.com/news/RADV-SteamVR-Public-Code
|
RADV SteamVR Source Changes Now Public
|
Michael Larabel
|
Now that Vulkan 1.0.42 is public and it contains the extensions needed for SteamVR on Linux, the RADV changes are now public.
David Airlie had made the modifications to RADV for supporting the needed Vulkan extensions for SteamVR. Valve had made the RADV binary build available when SteamVR went into public beta until last week, but the source changes weren't disclosed since the extensions weren't yet made public. Now they are with v1.0.42 so Airlie has published his Git branch with VK_NV_dedicated_allocation, VK_KHX_external_memory_capabilities, VK_KHX_external_memory, and VK_KHX_external_memory_fd.
The RADV code changes are available via the radv-wip-steamvr Git branch. Dave wrote on his blog the external memory code will be going upstream once it's been cleaned up, the semaphore work is waiting on proper kernel changes, and that he's able to run Serious Sam The First Encounter in VR mode with this now public Radeon Vulkan driver code.
So now today we've seen an updated NVIDIA Vulkan beta driver, the start of Intel's ANV Vulkan patches, and now the RADV driver patches for the new Vulkan material. Sadly, we haven't yet seen any AMDGPU-PRO beta driver or word on when there might be an update to AMD's official Vulkan Linux driver.
| 6
| 1,760,719,756.082836
|
https://www.phoronix.com/news/AMD-Ryzen-Newer-Kernel
|
AMD's Ryzen Will Really Like A Newer Linux Kernel
|
Michael Larabel
|
AMD's Ryzen CPU is finally shipping in a few days! If you are planning to be an early adopter of AMD Ryzen processors, you will really want to be running a newer Linux kernel release for proper support and performance.
I haven't received any communication from AMD about any review samples or the like, so I am not under NDA for the upcoming AMD Ryzen launch on 2 March... And thankfully received some information from a reliable source this weekend regarding Ryzen support Linux requirements. So here they are.
I am told that if using Ubuntu 16.10 / Ubuntu 16.04.2 as a base state for AMD Ryzen, users should generally be okay. In other words, you won't get a kernel oops on boot or anything dramatic like that but could be missing some functionality. However, you will really be better off with a newer Linux kernel.
As I wrote about back in December, Linux 4.10 landed a lot of Zen/Ryzen code. So with Linux 4.10 looks to be -- and reaffirmed by this trusted confidant -- a good point for AMD Ryzen testing and usage. So far in the Linux 4.11 cycle we haven't seen anything Ryzen-specific appear to come through.
But if you don't feel comfortable moving to the recently-released Linux 4.10 or your distribution hasn't yet offered you an easy upgrade path, an alternative is to be running at least Linux 4.9.10.
I was pointed out in particular to this kernel commit that only landed in early February: "x86/CPU/AMD: Fix Zen SMT topology." It notes in the commit message, "After: a33d331761bc ("x86/CPU/AMD: Fix Bulldozer topology") our SMT scheduling topology for Fam17h systems is broken, because the ThreadId is included in the ApicId when SMT is enabled. So, without further decoding cpu_core_id is unique for each thread rather than the same for threads on the same core. This didn't affect systems with SMT disabled. Make cpu_core_id be what it is defined to be." That mentioned regression was introduced only in January but had been pulled back into stable kernel point releases from Linux 4.6 and newer. This fix for the (Ry)zen SMT topology is in Linux 4.10 and was back-ported to Linux 4.9.10+ for those still riding the 4.9 kernel.
Without that fix, you could be affected by broken SMT scheduling topology. Simultaneous Multi-Threading is one of the new features for AMD Ryzen/Zen and will play a big role. For instance, with the Ryzen 7 CPUs they are 8 cores but 16 threads via SMT. If you care about multi-threaded performance, just make sure you are riding a newer kernel release.
I am told that besides that, recent Linux kernels should be in good shape, but as generally is the case: the newer the better. One of the bigger focuses could be on the AMD Ryzen motherboard support. Some Ryzen launch motherboards are using the Realtek ALC1220 audio codec, for example, and that is support is only coming to Linux 4.11. Among the AMD motherboards using the ALC1220 codec are the Biostar X370GT7, Gigabyte GA-AX370-Gaming 5, Gigabyte GA-AB350-Gaming 3, and others.
| 77
| 1,760,719,757.006254
|
https://www.phoronix.com/news/GLSL-TGSI-Shader-Cache-Lands
|
GLSL/TGSI On-Disk Shader Cache Lands In Mesa For R600g/RadeonSI
|
Michael Larabel
|
Timothy Arceri who has been working on the Mesa on-disk shader cache for months and most recently began working for Valve on the AMD Linux driver stack has landed support in Mesa 17.1-devel for the GLSL/TGSI on-disk shader cache for the R600g and RadeonSI Gallium3D drivers.
With the latest Git activity today, the GLSL/TGSI on-disk shader cache has landed after being signed off by upstream AMD Linux developers. This will help in reducing load times for shader-heavy games and in making the Radeon gaming process more fluid for games that dynamically load OpenGL shaders when needed. Shadow of Mordor should be among the Linux games benefiting from this support.
This shader cache support makes use of previously-landed Mesa GLSL shader cache support and is also wired in through the Mesa state tracker so could be utilized too by other Gallium3D drivers that choose to.
I'll work on some benchmarks shortly to see if it ends up making much of a difference in our performance tests.
| 37
| 1,760,719,757.872295
|
https://www.phoronix.com/news/AMD-Ryzen-7-Pre-Order
|
AMD Ryzen 7 CPUs Shipping 2 March, Pre-Order Today
|
Michael Larabel
|
AMD Ryzen 7 CPUs are available for pre-ordering today and these long-awaited "Zen" CPUs will be shipping on 2 March.
The first three Ryzen CPUs to ship are the Ryzen 7 1700, 7 1700X, and 7 1800X. The Ryzen 7 1700 has 8 cores, 16 threads, 16MB L3 cache, 65 Watt TDP, 3.0GHz base frequency, and 3.7GHz turbo frequency. That 1700 model will set you back $329 USD. The Ryzen 7 1700X has eight cores and 16 threads and 16MB cache but a 95 Watt TDP as it has a base frequency of 3.4GHz with a turbo of 3.8GHz and an XFR frequency of 3.8GHz+. That CPU will set you back $399 USD. The high-end CPU for $499 USD is the Ryzen 7 1800X with 8 cores, 16 threads, 95 Watt TDP, 3.6GHz base, 4.0GHz turbo, and 4.0GHz+ XFR.
AMD claims that Ryzen has 52% more instructions per clock than their older Excavator cores. The Ryzen 7 1800X is said to compete with the Intel Core i7 6900K, at least under Windows. The Ryzen 7 1700 should beat out the Core i7 6800K, Core i7 7700K, and other Intel CPUs in that range.
All of the press material today has been about AMD Ryzen 7 CPUs on Windows. Unfortunately, no Linux details yet. AMD also hasn't contacted me yet about any review samples so I doubt that's going to happen for the Ryzen 7 launch day. But, given all the interest from Phoronix readers, I will be buying Ryzen 7 hardware myself then to be able to deliver near-launch-day results.
I am planning to pick up at least an AMD Ryzen 7 1700, but hopefully will be able to test the higher-end models too if there is enough support from Phoronix Premium members or PayPal tips. If you'll be pre-ordering any Ryzen hardware (or making any other online purchases) please use our Amazon and NewEgg affiliate links to help support the site. Thanks for your support and look forward to AMD Ryzen Linux benchmarks in early March!
| 137
| 1,760,719,758.35565
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.