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-SoC-PMC-Driver-Linux-5.11
|
AMD SoC PMC Driver Slated To Come With Linux 5.11
|
Michael Larabel
|
In addition to the AMD Sensor Fusion Hub (SFH) driver coming with Linux 5.11 for improving Ryzen laptop support, the AMD SoC PMC driver is also under review for landing in this next kernel release.
The AMD SoC PMC driver is for the power management controller found so far with Raven Ridge, Picasso, Renoir, and Cezanne SoCs. While Raven through Renoir has been available for a while, AMD only now is contributing this power management controller for the mainline Linux kernel -- presumably due to AMD Chromebooks and the like that have motivated the other recent AMD Linux mobile improvements.
This AMD power management controller driver is responsible for S2Idle transactions that are driven by the platform firmware on the system management unit. This is about improving the power management abilities for Ryzen notebook/embedded hardware but no impact on Ryzen desktops.
The AMD PMC driver is currently under review with the platform-drivers-x86.git code and if all goes well it will be there with the Linux 5.11 kernel and its merge window opening in December.
| 2
| 1,760,719,428.465878
|
https://www.phoronix.com/news/AMD-Renoir-Linux-5.10
|
AMD Renoir Running Smooth On Linux 5.10
|
Michael Larabel
|
After last week sharing some Intel Tiger Lake benchmarks on Linux 5.10, the tables have turned and here are some similar tests when running Linux 5.10 on an AMD Ryzen 4000 series "Renoir" notebook.
Using a Lenovo IdeaPad with Ryzen 5 4500U with Ubuntu 20.10, I ran some benchmarks of Linux 5.9 stable against the Linux 5.10 development snapshot of the time.
Via the Phoronix Test Suite dozens of benchmarks were run in comparing these latest kernel versions.
Notable in switching to Linux 5.10 on this laptop meant CPUFreq now deafaulting to the Schedutil governor by default rather than Ondemand as was the default on prior kernel versions.
With Schedutil being used by default on Linux 5.10, there were some changes in performance. The NCNN neural network Vulkan performance was mixed and rather noisy. Ignoring NCNN, GIMP seemes to have performed better on the Linux 5.9 kernel configuration. The other tests showed a few percent swings here and there.
Of 125 tests run, Linux 5.10 was the first place finisher 60% of the time regardless of the margin of the win.
But if taking the geometric mean of all 125 benchmark results, you can see the performance overall is quite close to that of Linux 5.9 even with Schedutil by default. There weren't CPU power measurements for this article since the amd_energy driver isn't currently working for Renoir but will work on such tests with the WattsUp Pro soon.
Long story short it was a smooth experience for this AMD Ryzen 5 4500U notebook on Linux 5.10. All 125 test results can be seen on OpenBenchmarking.org. There are many Linux 5.10 features and improvements but not much specifically for benefiting Renoir. Linux 5.11 should be a bit more interesting for AMD notebook users with the AMD SFH driver and AMD SoC PMC driver for mainline.
| 9
| 1,760,719,428.885882
|
https://www.phoronix.com/news/Project-X-AMD-Zen-Coreboot
|
"Project X" - Pure Open-Source Coreboot Support On AMD Zen
|
Michael Larabel
|
Not only are AMD Ryzen 5000 series completely dominating in performance but they could soon see open-source Coreboot support as an alternative to the proprietary firmware/BIOS. Project X is an interesting effort around blob-free Coreboot/Oreboot support on AMD Zen.
Ron Minnich of Google who has been one of the original developers to Coreboot/LinuxBIOS has been engaged in "Project X" for providing "pure open-source" support on AMD Zen platforms.
Project X is about "eXcising binary blobs from the x86 part of Zen CPUs." This open-source AMD Zen support is being worked on both for Coreboot and the Oreboot downstream that is written in Rust.
Minnich will be talking about Project X in more detail next month during the Open-Source Firmware Conference (OSFC 2020) taking place at the start of the month.
But that's not all at OSFC about open-source AMD support either... AMD system management architect Supreeth Venkatesh will be talking about the company's work on open-source OpenBMC firmware support for their platforms. They have been working on open-source OpenBMC support as well as upstream support around AMD system interfaces.
Stay tuned for more information next month during OSFC. Also to be discussed at the December event is AMD platform support for TrenchBoot.
| 32
| 1,760,719,429.8814
|
https://www.phoronix.com/news/RadeonSI-EGL-Protected-TMZ
|
RadeonSI Gallium3D Adds Support for EGL Protected Surfaces Using AMDGPU TMZ
|
Michael Larabel
|
Landing in Mesa 20.3 during this final week of feature development is support in RadeonSI Gallium3D for EGL_EXT_protected_surface. This long-standing EGL extension allows surfaces/windows to beset as protected and in which case the contents are only accessible to secure accesses. Outside/insecure accesses to the window (surface) contents are blocked.
EGL_EXT_protected_surface is common in the mobile/embedded world and now is supported by the RadeonSI Gallium3D driver. This support though does depend upon AMDGPU TMZ with the kernel driver.
AMD Trusted Memory Zone support merged earlier this year for allowing encrypted video memory support with capable hardware.
This year AMD has been working on the likes of TMZ as well as HDCP support for handling DRM'ed/encrypted content on Linux. This work appears to be mainly driven by AMD APUs beginning to appear in Google Chromebooks and ensuring copy-protected content can play properly, etc. For most users EGL_EXT_protected_surface will go unused but at least the capability is there for those who want it.
The code merged this morning and will be in Mesa 20.3 due for its stable release in December.
| 26
| 1,760,719,430.587147
|
https://www.phoronix.com/news/AMD-Acquiring-Xilinx
|
AMD To Acquire Xilinx In $35 Billion Stock Deal
|
Michael Larabel
|
Following the rumors earlier this month that AMD was in talks to acquire Xilinx, a deal has been announced this morning.
AMD is acquiring Xilinx in an all-stock transaction worth $35 billion USD. The combined enterprise value is approximately $135 billion of the two companies. They expect the deal to close by end of calendar year 2021.
Details on the deal via the AMD press release that just hit the wire.
This comes as this morning AMD reported their Q3'2020 financial results with revenue growth by 56% and net income and EPS doubled year-over-year. AMD pulled in $2.80 billion for Q3. Their computing and graphics segment saw revenue up by 31% year-over-year and 22% for quarter-over-quarter. AMD expects to hit $3.0 billion in revenue for Q4.
| 37
| 1,760,719,431.276939
|
https://www.phoronix.com/news/LLVM-Lands-Basic-Znver3
|
LLVM Lands Very Basic Support For AMD Zen 3 CPUs
|
Michael Larabel
|
While AMD has landed Znver3 support in GNU Binutils, the company hasn't yet sent out patches for either the GCC or LLVM/Clang compilers in setting up the Zen 3 target with its new instructions or optimized scheduling model / cost table. But a basic implementation has been merged to LLVM for allowing "-march=znver3" based on the limited public details thus far.
Merged to mainline LLVM 12 yesterday was a basic implementation allowing for -march=znver3 targeting that basically flips on the new instructions known to be supported by Zen 3. Beyond Zen 2, it flips on INVPCID, PKU, VAES, and VPCLMULQDQ. There are also a few other instructions supported by Zen 3 as outlined in this earlier article.
Besides not having all of the instructions wired up yet, there is no optimized scheduling model yet for Zen 3 and just re-using the existing Zen assets. Also, auto-detection for Zen 3 isn't yet wired up.
So this patch now in LLVM 12 will allow compiling code with -march=znver3 but it's incomplete.
The last I've heard is that AMD should be publishing their compiler patches in November, presumably around the time the Ryzen 5000 series processors begin shipping. That support should work its way into GCC 11 for release as stable in March~April while LLVM 12 should release in February~March.
| 7
| 1,760,719,432.287511
|
https://www.phoronix.com/news/i7-1165G7-vs-Ryzen-7-PRO-4750U
|
Intel Core i7 1165G7 Tiger Lake vs. AMD Ryzen 7 PRO 4750U Linux Performance
|
Michael Larabel
|
For the Intel Tiger Lake Linux benchmarking thus far with the Core i7 1165G7 on the Dell XPS 13 9310 it's primarily been compared against the Ryzen 5 4500U and Ryzen 7 4700U on the AMD side since those are the only Renoir units within my possession. But a Phoronix reader recently provided me with remote access to his Lenovo ThinkPad X13 with Ryzen 7 PRO 4750U (8 cores / 16 threads) for seeing how the Tiger Lake performance compares against that higher-end SKU.
Phoronix reader Tomas kindly provided SSH access to his ThinkPad X13 with Ryzen 7 PRO 4750U and 16GB of RAM. The Ryzen 7 PRO 4750U is quite close to the Ryzen 7 4800U with 8 cores / 16 threads but graphics capabilities in line with the 4700U. He's been quite happy with the ThinkPad X13 as a replacement to the Dell XPS 13 for business usage and has been running it with Ubuntu 18.04 LTS on the Linux 5.8 kernel.
For getting an idea how the Core i7 1165G7 compares to the 4750U, I ran tests with the Phoronix Test Suite on the ThinkPad X13 compared to the recent i7-1165G7 results both on Ubuntu 20.04 LTS and then again with 20.10 given that Ubuntu 20.10 really helps out with Tiger Lake. Those results and the other laptops tested remain unchanged from the earlier article, just now with the Ryen 7 PRO 4750U results for reference purposes.
Given that it was just remote access and limited time, it's not a full review like usual under properly controlled conditions. But hopefully these numbers will be helpful since I don't have any Ryzen 4800/4900 series laptops. Most laptops tested at Phoronix have to be purchased retail due to most laptop vendors generally not interested / concerned enough about Linux and its small marketshare.
Interestingly the context switching performance on the Ryzen 7 PRO was faster than the other Renoir non-PRO laptops.
The Ryzen 7 4700U does have a higher base clock frequency than the Ryzen 7 PRO 4750U (2.0GHz vs. 1.7GHz) while both having a 4.1GHz peak boost clock. So in cases where the Ryzen 7 PRO 4750U can't shine with its eight cores / sixteen threads, in some cases the 4700U does still lead the race.
But for properly threaded workloads, the Ryzen 7 PRO 4750U obviously excels well past the Ryzen 7 4700U and in turn well beyond the performance afforded by the Core i7 1165G7 in just being a quad-core part with Hyper Threading.
The Ryzen 7 PRO 4750U within the Lenovo ThinkPad X13 looks like it can serve quite well as a very capable and powerful mobile system. Thanks to Tomas for providing remote access for being able to provide these Renoir 8c/16t tests while hopefully we will find our hands on some other Ryzen 4800/4900 series mobile hardware in the future; readers can show their support by viewing this site without any ad-blockers, letting laptop vendors know your interest in seeing Linux benchmarks on sites such as Phoronix, or by joining Phoronix Premium or tips to help offset the hardware costs.
With continuing to run more benchmarks each day on the Core i7-1165G7 Tiger Lake, you can also begin to find i7-1165G7 benchmarks on OpenBenchmarking.org including the ability to easily compare the i7-1165G7 against other CPUs on Linux. From there thanks to an independent user are also a few common points for comparing against the 4800U, among other CPUs.
| 17
| 1,760,719,433.155793
|
https://www.phoronix.com/news/Linux-5.10-Zen-3-CPUFreq-Quirk
|
Linux Gets Fix For AMD Zen 3 CPU Frequency Handling Stemming From 8 Year Old Workaround
|
Michael Larabel
|
Since 2012 there has been a quirk in the Linux kernel to disable/override using ACPI _PSD data on all AMD processors as a workaround in turn for Windows-specific behavior that clashes with the semantics of the Linux ACPI CPUFreq driver for CPU frequency scaling. With AMD Zen 3 this quirk is no longer needed to behave correctly and thus Linux 5.10 is going to drop this eight year old quirk on Zen 3 and newer.
The change since 2012 in the Linux kernel for AMD CPUs has overrode the ACPI _PSD table supplied by the BIOS. But now for Family 19h / Zen 3, the table accurately reports the P-state dependency of CPU cores. That correct table is needed for proper CPU frequency control with the new processors and thus the new kernel will stop overriding it so it can be used by ACPI CPUfreq for its frequency handling on the shiny new CPUs.
This change was sent in as part of additional power management updates for the Linux 5.10 kernel.
Given that it's a "fix" and amounts to just changing around one line of code (only applying the override for CPUs less than Family 19h), it will hopefully be back-ported quickly to stable kernel series ahead of the AMD Ryzen 5000 series beginning to ship in early November.
| 8
| 1,760,719,434.056135
|
https://www.phoronix.com/news/AMD-SFH-For-Linux-5.11
|
AMD SFH Driver To Land With Linux 5.11 For Better Ryzen Laptop Handling In 2021
|
Michael Larabel
|
It was sadly too late for squeezing into the current Linux 5.10 merge window but it looks like for Linux 5.11 in early 2021 the AMD Sensor Fusion Hub "SFH" driver will make its long awaited debut.
The AMD SFH driver is similar to the long-standing Intel ISH driver for supporting the sensor hub on modern laptops. The AMD SFH support is needed for laptops bearing gyroscopic sensors and other capabilities.
Back in January AMD finally published the Sensor Fusion Hub Linux driver for supporting the Ryzen laptops of recent years. With time the driver was revised to address various feedback but wasn't quick to get picked up for mainline and at times several weeks passing between code revisions.
But earlier this month we saw AMD SFH up to its eighth version with hopes of it being mainlined soon. It looks like now with the latest code the support is in good standing for mainline inclusion.
While before it was talked about for introducing it via the IIO subsystem pull, created yesterday via the HID subsystem was for-5.11/amd-sfh-hid. The thousand plus lines of new code is now staging in the HID area until the Linux 5.11 cycle kicks off at the end of the year for finally offering this sensor functionality with Ryzen laptops. Functionality varies based on the laptop but the driver supports the SFH accelerometer, gyro, magnetometer, and ambient light sensor.
| 4
| 1,760,719,434.800961
|
https://www.phoronix.com/news/Znver3-Binutils-Support
|
AMD Sends Out Patches Adding "Znver3" Support To GNU Binutils With New Instructions
|
Michael Larabel
|
One of AMD's compiler experts this week sent out a patch wiring up Zen 3 support in the important GNU Binutils collection for Linux systems.
The patch adds Znver3 to Binutils and was sent out at the start of the week. Unfortunately though the GNU Compiler Collection (GCC) patches for Znver3 have yet to be posted by AMD but hopefully will be done with enough time still for reaching the early next year GCC 11 compiler release.
In any case the public Binutils patch for Znver3 confirms that the forthcoming AMD Zen 3 processors support a number of new instructions.
First up are instructions new to Zen 3 not found currently with Intel CPUs:
INVLPGB - An instruction for invalidating TLB entries with broadcast.
TLBSYNC - This is used as a synchronizing instruction for TLB invalidations.
SNP - The SNP instruction set is comprised of the PSMASH, PVALIDATE, RMPUPDATE, and RMPADJUST instructions. PSMASH expands a 2MB RMP entry into a corresponding set of 4KB page RMP entries. PVALIDATE is used for validating or rescinding validation of a guest page's RMP entry, RMPUPDATE writes a new RMP entry, and RMPADJUST modifies RMP permissions for a guest page.
The instructions aren't particularly exciting to most end-users but notable in that Intel currently doesn't expose these instructions where as in past Zen generations it was mostly an AMD catch up game for employing instructions already supported by Intel CPUs.
Seeing INVLPGB / TLBSYNC / SNP support though isn't too surprising as AMD's Programming Reference Manual has noted these instructions since April. But now with the Binutils patch is the first official confirmation that these instructions are supported by AMD Zen 3 processors.
But besides those original instructions, the Znver3 target also supports some additional instructions too that have been supported by Intel CPUs:
VAES - Vector AES instructions!
VPCLMULQDQ - This instruction is to perform a carry-less multiplication of two quadwords. VPCLMULQDQ allows VEX-encoded 256-bit version but part of the path to the AVX-512 extensions supported by Intel.
INVPCID - PCID support with Zen 3. Though we previously reported on AMD indicating work on PCID support and AMD Linux patches for using PCID/INVPCID with KVM guests.
OSPKE - PKE support albeit isn't too surprising and reported on back in April for memory protection keys with AMD CPUs.
That's it so far in terms of the Zen 3 patches for the GNU (or LLVM) open-source toolchain plumbing around Zen 3 (znver3) but as more lands we'll certainly let you know on Phoronix.
| 18
| 1,760,719,435.657017
|
https://www.phoronix.com/news/AMD-SEV-SNP-IOMMU-Linux-5.10
|
AMD Secure Nested Paging IOMMU For SEV-SNP Lands In Linux 5.10
|
Michael Larabel
|
In addition to Linux 5.10 supporting SEV-ES as the "encrypted state" for AMD EPYC's Secure Encrypted Virtualization, this kernel is also adding Secure Nested Paging (SNP) support to the AMD IOMMU driver as part of their next-generation SEV-SNP security.
AMD SEV-SNP is an effort to further boost virtual machine isolation and appears to likely be supported with upcoming AMD EPYC 7003 "Milan" processors based on the timing of their original SEV-SNP whitepaper earlier this year and now the timing of this SNP Linux kernel support. SEV-SNP builds on the original AMD SEV and SEV-ES to offer additional hardware-based memory integrity protections for fending off hypervisor-based attacks.
"The basic principle of SEV-SNP integrity is that if a VM is able to read a private (encrypted) page of memory, it must always read the value it last wrote.This means that if the VM wrote a value A to memory location X, whenever it later reads X it must either see the value A or it must get an exception indicating the memory could not be read. SEV-SNP is designed so that the VM should not be able to see a different value from memory location X," explains the SEV-SNP whitepaper from January 2020.
With Linux 5.10, the IOMMU driver changes add the Secure Nested Paging support to the AMD IOMMU code. The Linux kernel code will fault when a device tries DMA on memory owned by a guest. That came with the IOMMU PR for the ongoing Linux 5.10 merge window.
Further kernel changes for fully supporting AMD SEV-SNP are needed for this functionality that won't all be included in Linux 5.10 but at least the initial IOMMU side changes are now mainlined.
| 3
| 1,760,719,436.366634
|
https://www.phoronix.com/news/Linux-5.10-Does-AMD-SEV-ES
|
AMD SEV-ES Sent In For Linux 5.10 To Further Secure Guest VMs
|
Michael Larabel
|
While the mainline Linux kernel for quite a while now has supported AMD Secure Encrypted Virtualization for EPYC processors as a means of better securing guest virtual machines (VMs) and public clouds with hardware memory encryption and using one key per VM to not only protect between guests but also the hypervisor, with Linux 5.10 comes AMD SEV-ES as another step forward for secure virtualization on AMD EPYC.
AMD SEV-ES takes the security a step further by encrypting all the CPU register contents when exiting a VM to ensure there is no leakage of register information to the hypervisor. SEV-ES is also reportedly able to detect malicious modifications to the CPU register state. SEV-ES is particularly suited for protecting against control flow and rollback attacks and other scenarios of needing to know or manipulate the register state.
The Linux 5.10 implementation of AMD SEV-ES is ready to go and ensures the registers are encrypted/decrypted on world switches. There have been Linux kernel patches floating around for SEV-ES since early 2020 while now Linux 5.10 as the last full kernel cycle of the calendar year will see this support land.
The AMD SEV-ES support was sent in as part of its own pull request on Tuesday. This current implementation is KVM-focused with other hypervisors also needing to be adapted to handle the SEV-ES support.
| 0
| 1,760,719,437.76295
|
https://www.phoronix.com/news/AMD-SFH-Linux-Driver-V8
|
AMD Sends Out Latest SFH Linux Driver Patches
|
Michael Larabel
|
It was in January that AMD finally published an open-source Linux driver for their Sensor Fusion Hub used by AMD Ryzen laptops for various sensor functionality. As we approach the end of the year this driver still hasn't been mainlined yet but a new revision was sent out on Friday.
Friday marked the eighth revision to the AMD "SFH" Linux kernel driver patches for supporting the sensor functionality on many Ryzen laptops. This was the first revision to the driver since the prior patches were sent out in August. This new version has cosmetic code changes, adds the initialization code within the probe function, and other code changes but seemingly nothing too major.
AMD SFH provides similar functionality to Intel's longstanding ISH sensor hub driver for Linux. The AMD SFH v8 driver patches can be found on the kernel mailing list. So far no review comments so we'll monitor and see if this driver is now deemed in an acceptable state for mainlining via the IIO area and whether it can still squeeze into the imminent Linux 5.10 cycle.
| 6
| 1,760,719,438.141971
|
https://www.phoronix.com/news/AMD-Zen-3-Linux-Expectations
|
AMD Ryzen 5000 Series (Zen 3) Linux Expectations - Should Be Good But No "Znver3" Compiler Yet
|
Michael Larabel
|
Today perhaps will be the most interesting day since the start of the pandemic... It's finally the day where AMD Zen 3 desktop CPUs are expected to be revealed in just about one hour's time! Stay tuned, but before that virtual event, here is a word on the Linux prospects and support for these upcoming AMD CPUs.
Long story short, barring any issues like we've seen in the past around flakey RdRand or other problems, the AMD Zen 3 desktop CPU support under Linux should be in fine shape. On prior kernels we've seen new IDs added and other bits around Family 19h like preparing for larger microcode files. Especially with no new chipsets expected for this initial Zen 3 launch, I don't foresee many (or any?) Linux support issues at launch. But not everything is squared away quite yet.
Linux 5.10 will bring Zen 3 CPU temperature monitoring. Linux 5.10 won't be out as stable until year's end so if you are really concerned about CPU temperature monitoring support that will be coming late albeit at least this time it was contributed by AMD engineers ahead of launch rather than left up to the community and only tackled post-launch. There is also the RAPL PowerCap patches for Zen 3 that haven't yet been queued for introduction to mainline for power monitoring/control support. But in terms of core functionality that should all be in place with features like power and thermal monitoring again being tardy for Linux.
The one unfortunate area that isn't yet addressed for the Zen 3 support is on the compiler side. AMD has yet to provide any public patches for bringing up Zen 3 with the new znver3 target for the GCC or LLVM Clang compilers. That has yet to be published as presumably they don't want to reveal all their new instruction set extensions ahead of launch, even though Intel often reveals this many months or years ahead of time in wanting to ensure good GCC/Clang compiler support at launch and in released versions. Intel has already been hitting the compilers with Alder Lake, Sapphire Rapids, and other bits for forthcoming CPUs - their punctual open-source software support for years is much appreciated by early adopters. Even if AMD posts the compiler patches today for the znver3 target, it won't appear until GCC 11 in March~April in released form or until LLVM Clang 12 also around the March timeframe next year. And then for Linux distributions like Ubuntu it won't be shipping with GCC 11 until Ubuntu 21.10 this time next year.
The lack of Znver3 support in advance to line up with the major open-source compiler releases is unfortunate although AMD presumably will have out their LLVM/Clang-based AMD Optimizing C/C++ Compiler (AOCC) around the time these new CPUs ship. We do know PCID and other new extensions may be coming with Zen 3. A tuned scheduler model is also ideal in ensuring optimal binaries for Zen 3.
Granted, the lack of a Znver3 compiler target for GCC/Clang isn't an issue for gamers or most other desktop users for that matter, but still for Linux users often compiling their own packages from source and tinkering with their trying to obtain optimal open-source performance, it's too bad Znver3 support isn't yet out there. Hopefully AMD gets their Znver3 compiler code out there soon and no issues getting the support merged for those early 2021 compiler releases.
But at least the rest of the Ryzen 5000 series / Zen 3 Linux support will hopefully be in good standing for launch -- of course, once the CPUs begin to ship there will be plenty of Linux benchmarks and coverage on Phoronix.Continue on now to our AMD Ryzen 5000 series launch coverage.
| 18
| 1,760,719,439.881071
|
https://www.phoronix.com/news/AMD-Reportedly-Talking-To-XLNX
|
AMD Reportedly In Talks To Acquire Xilinx
|
Michael Larabel
|
While AMD is providing great pressure against Intel in the CPU space, it looks like AMD could be soon going up against them in the FPGA space too.
The latest M&A chip talk is that AMD is reportedly in advanced talks to buy Xilinx. According to the Wall Street Journal, AMD has been looking to court Xilinx for more than $30 billion USD.
Xilinx is known for their FPGAs and AMD acquiring them could help them in more areas compete against Intel. It's also at a time the chip industry has been consolidating especially with NVIDIA courting Arm.
The WSJ notes a deal between AMD and Xilinx could be announced as soon as next week. Interesting times ahead...
| 28
| 1,760,719,440.107302
|
https://www.phoronix.com/news/AMD-Green-Sardine
|
AMD Sends Out Initial Linux Graphics Driver Support For The "Green Sardine"
|
Michael Larabel
|
AMD has been sending out a lot of new Linux graphics driver enablement code recently for the Linux with the newest being the "Green Sardine" platform.
Green Sardine is for bringing up a new APU platform with their Linux graphics driver code. The new code treats the Green Sardine platform under existing Renoir family code paths but the principal code differences are for loading different firmware files on the Green Sardine versus Renoir. There isn't any other major alterations with the new Green Sardine code sent out today and still Vega-based.
Given that, it's quite possible Green Sardine is the "Cezanne" Ryzen 5000 APU series.
The straight-forward patches can be found on the amd-gfx mailing list.
In any case, it's great to see AMD continuing to deliver timely open-source Linux graphics driver support pre-launch and that punctual support has been improving quite nicely this year in particular.
| 30
| 1,760,719,441.513556
|
https://www.phoronix.com/news/AMD-RAPL-Linux-Now-19h
|
AMD RAPL PowerCap Patches Updated For Linux - Now Include Family 19h (Zen 3)
|
Michael Larabel
|
Patches from a Google engineer allow run-time average power limiting (RAPL) support for AMD Zen processors within the Linux PowerCap driver.
Earlier this year was AMD Zen RAPL support in the Perf subsystem while this more recent activity for AMD RAPL has been about the PowerCap code. The Linux PowerCap framework was originally conceived by Intel but AMD Zen CPUs have similar MSRs available for supporting limiting the CPU TDP and also reading the current energy usage. With these patches /sys/class/powercap/intel-rapl/intel-rapl:0/energy_uj is the latest way of exposing the AMD CPU energy usage under Linux.
Back in July we covered the AMD Zen RAPL PowerCap patches by Google's Victor Ding. Now though he sent out a new series. With this latest code for exposing the AMD RAPL counters the notable change is adding a new patch for Family 19h support -- Zen 3.
The Family 19h MSRs around the RAPL counters turn out to be identical to Family 17h (Zen / Zen 2) CPUs and so is extended to work on these forthcoming processors. These latest patches are now out on the kernel mailing list.
It's rather close to seeing the code land for Linux 5.10 given its merge window is opening up in a few days but otherwise this straight-forward code will hopefully be in Linux 5.11.
| 1
| 1,760,719,441.652196
|
https://www.phoronix.com/news/Linux-5.10-AMD-Zen3-k10temp
|
Linux 5.10 Adding Support For AMD Zen 3 CPU Temperature Monitoring
|
Michael Larabel
|
The next version of the Linux kernel will allow monitoring temperatures of the upcoming AMD Zen 3 processors.
While CPU temperature monitoring support may seem mundane and not newsworthy, what makes this Zen 3 support genuinely interesting is that it's coming pre-launch... This is the first time in the AMD Zen era we are seeing CPU temperature reporting added to the Linux driver pre-launch. Not only is it coming ahead of the CPUs hitting retail channels but the support was added by AMD engineers.
Contrast that to Zen 1 / Zen+ / Zen 2 and various quirks/differences within generations, previously the support hasn't come until after launch and figured out by the community/non-AMD developers. Due to the Linux kernel cadence, at times the support hasn't been part of a released kernel until months after a CPU shipped. There has been a few times that it hasn't come until I received my review samples and then dug into figuring out the IDs or testing changes or checking for any Tctl offset in getting new CPU support added.
Thus to see Zen 3 support patches sent out by an AMD engineer prior to the launch is pleasant to see and hopefully a sign of more punctual support moving forward. But at least the code is out there pre-launch. The patch adds the new bits needed for Family 19h (Zen 3) in the k10temp driver. That patch is part of hwmon-next ahead of the Linux 5.10 merge window opening in October. Granted, due to the kernel cadence, the Linux 5.10 merge window isn't opening until early to mid October and the Linux 5.10 stable release won't be out until the end of the calendar year or so, in which case AMD Zen 3 retail CPUs will still have been shipping for weeks... So far no Zen 3 patches for the AMD_Energy driver that was recently mainlined but hopefully that isn't far behind.
Aside from thermal reporting bits and other non-critical pieces, recent mainline Linux kernel builds appear to be in good shape for AMD Ryzen and EPYC Zen 3 CPUs after various additions since Linux 5.6. Once Zen 3 CPUs are released we'll be having our usual full reports on Linux compatibility and performance. AMD recently confirmed more details on Zen 3 will be announced on 8 October.
| 15
| 1,760,719,443.179848
|
https://www.phoronix.com/news/AMD-3000-C-Series-Chromebooks
|
AMD Announces Ryzen/Athlon 3000 C-Series For Chromebooks
|
Michael Larabel
|
AMD today announced the Ryzen 3000 and Athlon 3000 C-Series processors for use in Google Chromebooks from multiple vendors.
AMD announced these 3000 C-Series mobile processors as the first Zen optimized Chromebook processors with Acer, ASUS, HP, and Lenovo all committing to releasing AMD Chromebooks in Q4'2020.
Compared to the previous-generation AMD A-Series "Excavator" APUs in Chromebooks, AMD is promoting up to 251% better graphics performance, up to 104% faster productivity, and up to 152% better photo editing with these new Zen C-Series processors.
At the top-end is the AMD Ryzen 7 3700C as a 4c/8t 15 Watt processor with 2.3GHz base frequency and 4.0GHz boost frequency while having 10 Vega GPU cores and 6MB cache. There is also the Ryzen 5 3500C and Ryzen 3 3250C while at the bottom of the stack is an Athlon Silver 3050C and Athlon Gold 3150C.
More details on the AMD Athlon/Ryzen 3000 C-Series via AMD.com.
These 3000 C-Series Chromebooks will be going up against forthcoming Intel Tigerlake Chromebooks. Yesterday Intel posted their own Tigerlake Chromebook performance details in this blog post for those interested.
| 24
| 1,760,719,443.236915
|
https://www.phoronix.com/news/AMD-SME-SEV-Cache-Coherency
|
Linux 5.10 To Support AMD SME Hardware-Enforced Cache Coherency
|
Michael Larabel
|
Linux 5.10 is set to support a new feature of AMD Secure Memory Encryption (SME) as part of the Secure Encrypted Virtualization (SEV).
The new feature is supporting hardware-enforced cache coherency and is coming to Linux 5.10 via patches that have been worked on by an Oracle engineer. It's not clear what AMD EPYC CPUs support this hardware-enforced cache coherency as it's checking on a CPUID bit but given the timing it wouldn't be surprising if this is for the upcoming AMD EPYC Zen 3 processors.
On processors supporting this CPUID feature, there is guaranteed coherency between the encrypted/unencrypted mappings of the same physical page. Thus with this AMD hardware-enforced cache coherency there is no need to flush the page from all CPU caches in the system before changing the value of the C-bit for a page. Avoiding unnecessary cache flushes should help with performance albeit the patches don't detail any specific advantages and I haven't been able to find any public documentation this specific SME feature.
This patch adds the bits for "SME_COHERENT" feature while this followup patch ensures not to flush the cache if there is the hardware-enforced cache coherency across encryption domains. Both patches are queued as part of "x86/cpu" changes ahead of the Linux 5.10 merge window opening in October.
| 0
| 1,760,719,444.588085
|
https://www.phoronix.com/news/AMD-Zen-3-RDNA-2-October
|
AMD Begins Teasing Zen 3 + RDNA 2 With Dates In October
|
Michael Larabel
|
It's been widely expected AMD will launch their next-generation RDNA 2 graphics cards and Zen 3 processors in Q4 as they previously reported as well as leaks pointing to October reveal dates. Today the company is sharing actual dates for said announcements.
Via Twitter, AMD shared that "a new journey begins" for Zen 3 on October 8 while for RDNA 2 the date listed is October 28.
That's all for now but at least we have some dates to look forward to in learning more about Zen 3 and RDNA 2. This will likely be the dates for their virtual announcements while the ship dates for the new AMD products will likely be later on.
| 40
| 1,760,719,444.996465
|
https://www.phoronix.com/news/TUXEDO-Pulse-14-Ryzen
|
TUXEDO Introduces New Linux Laptop With Ryzen 7 4800H / Ryzen 5 4600H
|
Michael Larabel
|
Last month the German Linux PC vendor TUXEDO Computers launched the PULSE 15 with AMD Ryzen "Renoir" processors. Today they launched a new model also featuring the very popular AMD Renoir parts.
Last month's Renoir introduction from TUXEDO Computers was with the Pulse 15 while today the Bavarian company is introducing a new Pulse 14 for those wanting something a bit smaller and lightweight.
The new Pulse 14 is just 1.1 kg and comes in at 320 x 214 x 16.8 mm. The Pulse 14 has options for the AMD Ryzen 5 4600H and Ryzen 7 4800H processors with Radeon RX Vega onboard graphics being used. The laptop has a 48 Wh battery for around a six hour battery life under normal usage, options for up to 64GB of RAM, and varying storage options as well. The 14-inch display is using a 1080p display with IPS 60Hz panel.
More details on the new TUXEDO Pulse 14 with AMD Ryzen 4000 series mobile processors via TUXEDOComputers.com.
| 36
| 1,760,719,446.537032
|
https://www.phoronix.com/news/AMD-Going-To-64-MCE-MCA-Banks
|
Machine Check Banks To Double With Future AMD CPUs
|
Michael Larabel
|
In preparation for "future AMD systems", which is likely AMD EPYC Zen 3 "Milan" server processors, there will be more banks for the Machine Check Exception handling.
Banks for the Machine Check Architecture are a means of organizing subsystems/subevents with their relevant MSRs for detecting/reporting machine errors. Older processor families had only a few banks of registers for machine check support while the current limit in the AMD MCE driver has been 32 banks. With a patch queued in ras/core ahead of Linux 5.10, that maximum number of banks is being increased from 32 to 64.
The patch by one of AMD's engineers simply notes, "Increase maximum number of banks to 64 because future AMD systems will support up to 64 MCA banks per CPU."
It will be interesting to see what's in store with the machine check improvements for future AMD CPUs, which given the timing will likely be with Zen 3 server CPUs.
| 0
| 1,760,719,447.373557
|
https://www.phoronix.com/news/Linux-5.9-PCI-P2PDMA-Zen-Newer
|
Linux 5.9 Enables P2PDMA For All AMD CPUs Zen + Newer
|
Michael Larabel
|
The PCI subsystem updates have been sent in for the Linux 5.9 kernel. Peer-to-peer DMA support is now solid for all AMD CPUs of the Zen family or newer.
Support for peer-to-peer DMA (P2PDMA) on AMD Zen and newer CPUs is now set. This is for the whitelist being maintained around this feature. There previously was Zen bits in Linux 5.2 while now for Linux 5.9 appears ironed out.
This P2PDMA handling is now allowing all AMD CPU host bridges for AMD newer than Zen, to also cut-down on the maintenance burden moving forward with future generations. The kernel docs cover the support for PCI peer-to-peer DMA support for performing direct memory access transfers between two devices on the bus.
The PCI pull also has new drivers for Versal CPM Root Port and TI J721E PCIe host and endpoint. There are also various improvements to the likes of the Qualcomm, NVIDIA Tegra, Marvell, and Intel drivers.
The full list of PCI changes for Linux 5.9 via this pull request.
| 7
| 1,760,719,448.262674
|
https://www.phoronix.com/news/AMD-Sensor-Fusion-Hub-V6
|
AMD Sensor Fusion Hub Driver Under Review A Sixth Time For Linux
|
Michael Larabel
|
While a lot of interesting changes are coming for the in-development Linux 5.9 kernel, sadly a long overdue change isn't going to make the merge window and that is the AMD Sensor Fusion Hub driver.
The AMD Sensor Fusion Hub is utilized by some AMD Zen laptops for accelerometer and gyroscopic sensors on the devices, akin to the Intel Sensor Hub (ISH) that has long been supported under Linux. While the Sensor Fusion Hub (SFH) is used by laptops going back to Zen 1 hardware, it was only earlier this year that the AMD SFH Linux driver was posted.
Since the driver posting at the start of the year, the AMD Sensor Fusion Hub driver has gone through multiple rounds of review though not yet pulled into the mainline kernel tree. This morning the sixth revision to these driver patches were posted.
With the AMD SFH Linux driver v6 there are various code clean-ups and other items pointed out in prior rounds of review. That previous v5 driver was posted back in May. The v6 kernel patches are now up on the kernel mailing list. Some minor items have already been pointed out by the kernel build test bot as well as an upstream developer on improving the Kconfig bits for the driver, but further review will likely take some time.
With already being half-way through the Linux 5.9 merge window, we're looking at Linux 5.10 as likely being the earliest the AMD SFH driver could finally land. Linux 5.10 in turn should be out around the start of the new year though not appear on the likes of Ubuntu until Ubuntu 21.04 next April. It's unfortunate this driver has taken so long to come but at least the AMD developers are still working on getting it mainlined for improving the AMD Zen Linux laptop support. Aside from the SFH driver lacking, Zen and especially Zen 2 laptops are in generally great shape on Linux.
| 13
| 1,760,719,448.850481
|
https://www.phoronix.com/news/GCC-Drops-AMD-HSA
|
AMD HSA Offloading Support Dropped From The GCC Compiler
|
Michael Larabel
|
There didn't appear to be much usage ever out of the AMD HSA (Heterogeneous System Architecture) support within the GCC compiler and hadn't been maintained in a while so now has been wiped out of the GNU Compiler Collection.
AMD HSA offloading support was removed from the GCC compiler and libgomp library. This removal was done by a SUSE compiler engineer with SUSE having done much of the HSA bring-up for GCC under contract for AMD back in the day.
Dropping of the HSA offloading support from the GCC tree means lightening up the open-source compiler by some 21k lines of code. This HSA support in GCC hasn't seen much activity in a number of years/releases but is now no more.
AMD for their part have basically folded HSA functionality into the Radeon Open eCosystem (ROCm) umbrella as a separate software stack. On the Radeon Open eCosystem side they are mostly focused on their LLVM-based toolchain. However, in the GCC space since GCC 9 there has also been a Radeon GCN back-end being worked on by Mentor Graphics / CodeSourcery as part of work for AMD. That Radeon GCN back-end for GCC is still being matured for OpenACC/OpenMP offloading to Radeon GPUs. Perhaps for GCC 11 next year we will see that Radeon GCN back-end in good shape for OpenMP and OpenACC.
| 28
| 1,760,719,450.005304
|
https://www.phoronix.com/news/Google-Zen-RAPL-PowerCap
|
Google Sends Patches For AMD Zen / Zen 2 RAPL PowerCap Support
|
Michael Larabel
|
Building off the work sent out by Google engineers in recent months and merged for Linux 5.8 around RAPL support for AMD Zen / Zen 2 CPUs with supporting the "runtime average power limiting" counters on Linux similar to Intel's longstanding support, that work has continued now with Zen RAPL support in the PowerCap driver.
Google engineer Victor Ding sent out a set of patches this morning for AMD Zen / Zen 2 RAPL support within the PowerCap Linux driver that allows power capping of the CPU(s) if so desired and some new interfaces via sysfs.
The support amounts to less than 50 lines of new code as the RAPL counters exposed by AMD Zen/Zen2 CPUs operate in a similar manner to the Intel RAPL support that has long offered working Linux support. More details on the kernel's power capping framework can be found via the kernel documentation.
With re-using all of the existing Intel code paths, existing Linux power capping tools and other programs tapping the /sys/class/powercap/intel-rapl/intel-rapl:0/energy_uj and /sys/class/powercap/intel-rapl/intel-rapl:0/intel-rapl:0:0/energy_uj interfaces will "just work" with the AMD Ryzen / EPYC processors.
It's a pity it took this long for the support to materialize, but at least it's finally coming thanks to the work by Google engineers. The Linux 5.9 merge window is kicking off potentially next week already, so it may be too close for getting it into this next version of the kernel, but we'll see if it squeezes in at the last minute given the rather simplicity of the patches.
| 3
| 1,760,719,450.412758
|
https://www.phoronix.com/news/KDE-Slimbook-Ryzen-7-4800H
|
New KDE Slimbook Released - Powered By AMD Ryzen 7 4800H
|
Michael Larabel
|
The KDE Slimbook is getting a big upgrade in the form of the ProX and ProX 15 that are powered by AMD's Ryzen 7 4800H "Renoir" processor for offering much better performance and all-around better specs.
The KDE Slimbook ProX is a 14-inch ultrabook that weighs just 1.1 kg while the ProX 15 is a 15.6-inch version. Both models are powered by the AMD Ryzen 7 4800H, IPS LED displays, magnesium-aluminum construction, 47Whr battery on the ProX and 92Whr on the 15.6-inch model, support for up to 64GB of memory, three USB ports, USB-C, and HDMI outputs.
That's all for now due to having short notice before the embargo launch time, but more details should begin appearing at Slimbook.es. We will have one of these new KDE Slimbook ProX models at Phoronix for our review and benchmarking shortly, so look forward to much more in the weeks ahead with benchmarking this notable AMD Ryzen 7 4800H laptop catering towards a nice KDE experience.
| 34
| 1,760,719,451.54038
|
https://www.phoronix.com/news/AMD-Ryzen-4000-APUs
|
AMD Launches Ryzen 4000 APUs - But Only For Pre-Built PCs / OEMs
|
Michael Larabel
|
AMD today officially revealed their Renoir-based Ryzen 4000 APUs. Unfortunately though for enthusiasts, at least for now these APUs are just available for pre-built systems and OEMs.
At a later date these new APUs will be available for retail/DIY. The top-end model announced today is the Ryzen 7 4700G with 8 cores / 16 threads and a boost frequency up to 4.4GHz and 3.6GHz base clock. The Ryzen 7 4700G has just a 12MB L3 cache and a 65 Watt TDP. The line-up ranges from the Ryzen 3 4300GE at the bottom end with 4 cores / 8 threads and 6 graphics cores on a 35 Watt combined TDP up through the Ryzen 7 4700GE as another 35 Watt part while having 8c/16t on a 3.1GHz base frequency and 4.3GHz boost. Vega graphics are still used with these APUs.
The Linux support should be similar to the existing AMD Renoir mobile support, which generally works with modern Linux distributions but on distributions like Ubuntu 20.04 will mean upgrading the kernel beyond what is found out-of-the-box. Generally for AMD Renoir I recommend at least Linux 5.7.
More details will be up shortly on AMD.com.
| 73
| 1,760,719,452.151068
|
https://www.phoronix.com/news/Ryzen-7-4700U-Daily-Driver
|
I've Been Running The AMD Ryzen 7 4700U + Ubuntu 20.04 As My Main System
|
Michael Larabel
|
For about one and a half months now I have been using the AMD Ryzen 7 4700U as my main laptop paired with Ubuntu 20.04 LTS. It's been working out very well for not even being the top-of-the-line AMD Renoir SKU. Here is some additional commentary for those thinking about one of the new AMD laptops with Linux use.
Back in May I picked up a Lenovo IdeaPad 5 in order to deliver AMD Ryzen 7 4700U Linux benchmarks. This laptop for just over $800 USD came with a Ryzen 7 4700U, 16GB of RAM, 1080p display, 512GB NVMe storage. The performance of the Ryzen 7 4700U as an 8-core part with 2.0GHz base clock and 4.1GHz boost has been quite good and better than the Intel Whiskeylake Core i7 Dell XPS laptop I had been using as my daily driver. The Renoir graphics are also quite good for desktop use-cases.
While originally buying the laptop for AMD Renoir Linux testing due to a good deal, I was (and remained) decently impressed with the build quality of this Lenovo IdeaPad with never owning an IdeaPad model before but many ThinkPads over the years. I was impressed with the build quality of the laptop enough and the Ryzen 7 4700U performance that after my initial Linux testing I decided to make it my main laptop to replace the Dell XPS.
The Lenovo IdeaPad 5 build quality has been fine albeit especially during these pandemic times the vast majority of the time this laptop is connected to my KVM setup in the office. But with that said, driving a 4K display over HDMI with the Ryzen 7 4700U with Radeon Vega graphics has been working out fine. At least for my workflow of using GNOME Shell and keeping open Firefox, Thunderbird, GNOME Terminal, Gedit, and other applications, the Vega performance has been fine at 4K. This should get even better come GNOME 3.38 this autumn given all the performance optimizations on the GNOME side, but already it's fine. The occasional YouTube video also works without issue. Obviously though the performance would come up short if expecting to game.
On the CPU side, the Ryzen 7 4700U eight-core mobile processor has proven to make a noticeable difference compared to an Intel Core i7 Whiskeylake and I have no complaints about the performance for heavy web browser usage with Firefox and Chrome, Thunderbird always running for mail and RSS, frequent GNOME Terminal usage, editing photos within GIMP for articles, writing articles in Gedit, coding in Gedit and GNOME Terminal, etc. During Phoronix Test Suite test profile development is also code compilation and other demanding tasks, though usually for the very heavy workloads I am doing that on the dozens of other systems around.
While originally I hoped to buy a Ryzen 4800/4900 series laptop, the budget has sadly not allowed, and the Ryzen 7 4700U has proven to be plenty capable out of this laptop I originally didn't even intend to use for more than just frequent benchmarking. Similarly, there is the Ryzen 5 4500U with the Lenovo Flex 5 where I continue to be working on new Linux tests about daily (more interesting tests there slated for publishing next week).
As my main system, I have been using Ubuntu 20.04 LTS. With Ubuntu 20.04, as outlined in the earlier Ryzen 7 4700U, the main change needed is to upgrade the kernel for working accelerated graphics and battery reporting. I am using Linux 5.7 with Ubuntu 20.04 LTS on this laptop and it's been working out very well. I will be upgrading to Linux 5.8 shortly as there are some performance benefits.
Previously on my main system I was using Intel's Clear Linux. But when it came to transitioning to a new laptop, given Clear Linux divesting sort of from the desktop (albeit still supported) and other internal reorganizations that have happened within Intel, didn't instill much confidence in continuing to use it as my daily driver. While prior to Clear Linux I've been a longtime Fedora Workstation user, I have been quite happy with how Ubuntu 20.04 shaped up and all the work Canonical has been pouring into GNOME, so I decided to return to that as my daily OS. It's been working out well and no complaints. We'll see if that keeps up or if I move to Fedora Workstation whenever it comes time to move to a Renoir+1 laptop.
So all in over one month after moving to a Lenovo IdeaPad with Ryzen 7 4700U running Ubuntu 20.04 LTS, I am quite happy with the laptop itself, the performance out of the Ryzen 7 4700U, and Ubuntu 20.04 for that matter as my daily OS these days. This IdeaPad has even been probably the cheapest laptop I've used as my daily system at least in many years if not ever yet the performance with the Ryzen 7 4700U has been great, the build quality of the laptop is good enough when being predominantly in the office attached to a keyboard and 4K display, and the Linux support is there if using a new enough kernel. This is also the first time in more than one decade my main laptop has been AMD powered. As for using Ubuntu again as the OS on my main production system, I am very happy with how Ubuntu 20.04 LTS turned out.
| 58
| 1,760,719,453.781431
|
https://www.phoronix.com/news/TUXEDO-Pulse-15-AMD-Ryzen
|
TUXEDO Computers Launches A Linux Laptop With Ryzen 7 4800H / Ryzen 5 4600H
|
Michael Larabel
|
Back in May the folks at TUXEDO Computers in Germany launched their first AMD Linux laptop. That device though was a letdown in being based on a previous-generation AMD Ryzen 3000 series mobile processor rather than the far better Ryzen 4000 "Renoir" processors. Fortunately, today they announced the Pulse 15 laptop that comes in Ryzen 5 4600H and Ryzen 7 4800H processor options.
TUXEDO's Pulse 15 laptop is the Renoir laptop we've been waiting on for those wanting Linux pre-loaded on a compatible laptop rather than loading Linux on your own on the many other Renoir laptops already available.
The TUXEDO Pulse 15 is a 16.8mm thick laptop with magnesium alloy chassis, a reported idle battery life of 20 hours, more than 10 hours during light to moderate workloads, up to 64GB of RAM with dual channel memory slots, a M.2 slot, USB 3.1, and all of the other usual characteristics we are used to seeing in AMD Renoir laptops.
Pricing on the Pulse 15 starts at 896 EUR configured with a 1080p display, 8GB of single channel memory, Ryzen 5 4600H, and 250GB of storage. The Ryzen 7 4800H model starts at 974 EUR or 1018 EUR for the base price of the 4800H with 16GB of RAM in dual channel mode.
More details on the Pulse 15 via TUXEDOComputers.com.
| 79
| 1,760,719,454.257275
|
https://www.phoronix.com/news/Ryzen-9-3900XT-DDR4-Memory
|
AMD Ryzen 9 3900XT Memory Scaling Performance Under 100 Different Tests
|
Michael Larabel
|
For those thinking of picking up one of the new AMD Ryzen 3000XT series processors and weighing whether it's worthwhile on your budget picking up DDR4-3600 memory or other higher frequency DDR4 modules, here are some fresh benchmark results with the Ryzen 9 3900XT looking at 100 different tests on Linux and showing how the performance changes from DDR4-2133 through DDR4-3800.
These tests are mainly intended for reference purposes in showing the Ryzen 9 3900XT's impact on Linux performance with a pair of 2 x 8GB GSKILL DDR4-4133 modules operating at different frequencies.
The only change during the course of the testing was the memory frequency. Via the Phoronix Test Suite a wide assortment of benchmarks were carried out for helping to analyze the performance impact in workloads of relevance to you.
Head on over to OpenBenchmarking.org to see all the individual benchmark results in full.
| 18
| 1,760,719,455.361749
|
https://www.phoronix.com/news/Linux-5.8-Ryzen-5-4500U-Tests
|
Linux 5.8 Bringing Some Performance Boosts For AMD Renoir Graphics
|
Michael Larabel
|
Over the weekend I began running some benchmarks of the Linux 5.8 development kernel on the Lenovo Flex 5 laptop with Ryzen 5 4500U. One of the standouts so far for from this Linux 5.8 testing compared to the stable 5.6/5.7 kernel series is better Radeon graphics performance with the Renoir laptop.
Tests were done from this quite capable six-core ~$600 AMD laptop.
When running nearly 100 benchmarks on Linux 5.6 / 5.7 / 5.8 Git from the Ryzen 5 4500U laptop, the performance overall was quite close:
But when digging into all of the results, where Linux 5.8 was providing a visible advantage was in many of the graphics tests in offering better performance:
In other areas of performance, the Linux 5.8 kernel didn't equate to any meaningful changes either way for this Ryzen 5 4500U. Those interested in all of the data anyhow can find them on OpenBenchmarking.org. To the negative the only issue encountered on Linux 5.8 has been Google Chrome crashing with this new kernel, but presumably will be worked out in time for the release and an issue others have been encountering too and not specific to Renoir.
Additional Linux 5.8 tests including on more Radeon graphics configurations will be coming up soon to see if this is part of broad AMD Radeon graphics gains or specific to these latest Vega-based mobile graphics.
| 11
| 1,760,719,455.712388
|
https://www.phoronix.com/news/AMD-SMM-Callout-Privilege
|
AMD SMM Callout Privilege Escalation Bug Disclosed For APUs
|
Michael Larabel
|
AMD has made public "SMM Callout Privilege Escalation" or more formally CVE-2020-12890 as an AGESA vulnerability that could lead to arbitrary code execution on APUs.
The SMM Callout Privilege Escalation impacts mobile/embedded AMD APUs that could lead to arbitrary code execution undetected by the operating system. Affected hardware was launched between 2016 and 2019. AMD is shipping updated AGESA to motherboard vendors for mitigating the issue. Updated AGESA should be out for most systems before the end of the month.
More details on the AMD SMM Callout Privilege Escalation bug via AMD product security.
| 23
| 1,760,719,456.960022
|
https://www.phoronix.com/news/AMD-Ryzen-3000XT-Series
|
AMD Announces The Ryzen 3000XT Series
|
Michael Larabel
|
After weeks of rumors, AMD today officially confirmed the existence of the Ryzen 3000XT series.
The Ryzen 3000XT remains Zen 2 based desktop processor but with higher base/boost clock frequencies in posing more fierce competition against Intel's recent Comet Lake processors. The XT Series is comprised of the Ryzen 9 3900XT, Ryzen 7 3800XT, and Ryzen 5 3600XT processors.
AMD says the Ryzen 9 3900XT offers about 4% better single-threaded performance over the Ryzen 3000 series. The Ryzen 9 3900XT has a 3.8GHz base frequency and 4.7GHz boost frequency while will retail for $499 USD.
These AMD Ryzen 3000XT processors will begin shipping on 7 July. Stay tuned to Phoronix for Linux benchmarks on these new CPUs at a later date.
All of the new XT CPU specs can be found at AMD.com.
| 54
| 1,760,719,457.192071
|
https://www.phoronix.com/news/AMD-Zen3-INVPCID-KVM-Guests
|
AMD Prepping PCID/INVPCID Support For KVM Guests On Zen 3 EPYC
|
Michael Larabel
|
The latest Linux kernel patch work we are seeing out of AMD in preparations for Zen 3 processors coming later this year is INVPCID instruction support for KVM virtualization guests.
One of the new capabilities we are looking forward to on the Zen 3 instruction set side is adding Process Context Identifiers (PCID) support. As part of the PCID support is the INVPCID instruction for Invalidate Process-Context Identifier to invalidate TLB mappings and caches based on the process context identifier.
Sent out on Thursday were 3 patches in allowing INVPCID to be supported by AMD guests for the Kernel-based Virtual Machine (KVM). This allows PCID/INVPCID to be supported by AMD guests on future Zen 3 CPUs, namely next-generation EPYC, and important for performance in the cloud.
Given the timing of these patches, the PCID/INVPCID support for AMD KVM guests won't be merged for the current Linux 5.8 merge window but will have to wait now until at least 5.9. But at least we are seeing more AMD Linux kernel patches coming out now ahead of launch rather than after the product launch. The turnaround time for new Zen 3 features to the Linux kernel so far is shaping up better than Zen / Zen 2 while there still is room for improvement especially with Intel tending to send out their hardware enablement patches to Linux much earlier.
| 0
| 1,760,719,458.646991
|
https://www.phoronix.com/news/Linux-5.8-RAS-Core
|
AMD MCE Improvements, Renoir Temperature + EDAC Support Sent In For Linux 5. 8
|
Michael Larabel
|
Linux kernel developer Thomas Gleixner sent in the RAS/core changes on Friday night for the Linux 5.8 kernel merge window that is wrapping up this weekend.
Notable to this pull request is including the new AMD Family 17h Model 60h PCI IDs in the amd_nb code. And in turn this pull request ends up adding the 17h 60h support to the hwmon k10temp driver. The 60h series is for the recently launched AMD Ryzen 4000 "Renoir" processors. So now finally with Linux 5.8 is the ability to offer working AMD Renoir CPU temperature monitoring.
The 60h PCI IDs are also added to the AMD64 EDAC driver (Error Detection And Correction).
Meanwhile in the formal hwmon subsystem pull last week was the new AMD Energy driver as another step forward for AMD Zen/Zen2 CPUs on Linux 5.8.
This RAS core pull request also has AMD MCE (Machine Check Exception) improvements and other general x86 MCE improvements by Intel.
The list of changes in full for this last minute pull request via this kernel mailing list thread.
| 0
| 1,760,719,458.752038
|
https://www.phoronix.com/news/AMD-Video-What-Is-ROCm
|
AMD Publishes Video To Explain The Radeon Open eCosystem Stack (ROCm)
|
Michael Larabel
|
AMD has shared with us that they have published a video to explain in basic terms for the audience at large "What is ROCm?", a.k.a. the Radeon Open eCosystem stack.
The video is arguably long overdue with ROCm being several years old, but it has been evolving a lot lately with new features and capabilities for better taking on the likes of NVIDIA CUDA and Intel oneAPI. With AMD increasing securing super-computing wins, they have also been ramping up their efforts on this standards-based GPU compute stack.
For those wanting this nice high-level overview of the ROCm stack for AMD Linux systems, here is their new video:
For the question many have been asking: while passing along this video, I did ask them any status update on Navi/RDNA support for ROCm to which I am still waiting on hearing back. Obviously though their main focus for the time being is based on Vega given the upcoming Arcturus accelerator and principally tooling ROCm for their workstation/SC needs.
| 0
| 1,760,719,460.11598
|
https://www.phoronix.com/news/System76-Serval-WS-AMD-Ryzen
|
System76 Begins Offering Serval WS Laptop With AMD Ryzen
|
Michael Larabel
|
Linux enthusiasts have long been after System76 to offer an AMD Ryzen powered laptop and today they announced such in the form of the new Serval WS.
The newest System76 Serval WS is now based on AMD Zen 2! But before getting too excited, this is a workstation laptop and actually is using Ryzen desktop CPUs with either the Ryzen 5 3600, Ryzen 7 3700X, or Ryzen 9 PRO 3900. Unfortunately, System76 is not yet offering any lightweight AMD Ryzen 4000 "Renoir" laptop but just this workstation model.
The other aspect that will rub some Linux users the wrong way is that the laptop only comes with a NVIDIA GeForce graphics option. The base model starts at $1299 USD with a Ryzen 5 3600, GeForce GTX 1660 Ti, and 8GB of RAM.
The rest of the Serval WS specs include a 15.6-inch 1080p display, up to 64GB of RAM, dual NVMe/SATA slots, and all the other usual features expected out of a workstation laptop. Given the cooling requirements and workstation focus, this laptop is quite large at 32.5 x 360.9 x 258 mm and weights 2.7 kg / 5.95 lbs.
More details on the new Serval WS via System76.com.
| 49
| 1,760,719,461.455452
|
https://www.phoronix.com/news/AMD-Zen-RAPL-Linux-5.8
|
AMD Zen/Zen2 RAPL Support Merged In Linux 5.8
|
Michael Larabel
|
Complementing the new AMD Energy Driver in the hwmon subsystem for Linux 5.8 to provide per-socket/core reporting, the Linux perf subsystem in this new kernel version has run-time average power limiting (RAPL) framework integration for AMD Zen/Zen2 CPUs.
This Zen RAPL integration is what a Google engineer posted last month for integrating in the current-gen AMD processor power data into the RAPL framework, including exposing this information via the PowerCap sysfs interface and perf tool.
This RAPL code within the Linux kernel was led by Intel years ago when they added it for their processors. An example usage of the AMD RAPL usage with Linux 5.8 is perf stat -a --per-socket -I 1000 -e power/energy-pkg/ with energy-pkg being the only supported AMD event at this point.
This AMD Family 17h RAPL support was sent in via the perf updates for Linux 5.8 that have now been merged. This perf update also includes Zhaoxin CPU support for the perf subsystem and various perf tool updates.
| 0
| 1,760,719,461.485594
|
https://www.phoronix.com/news/Linux-5.8-SPI-Updates
|
AMD SPI Driver Sent In For Linux 5.8
|
Michael Larabel
|
Adding to the multiple new AMD drivers coming with Linux 5.8 is their new SPI controller driver.
The AMD SPI controller driver (spi-amd) was mailed out in April and for supporting the SPI controller within newer AMD SoCs. This 300+ lines of code driver was previously outlined in this earlier article.
Other SPI updates sent in for Linux 5.8 include performance improvements for the DesignWare driver, slave mode support for the Rockchip drivers, Raspberry Pi 4 SPI support, and Intel Elkhart Lake support is also available.
See the SPI pull request for the full list of Linux 5.8 SPI subsystem changes.
| 1
| 1,760,719,462.847547
|
https://www.phoronix.com/news/AMD-Sienna-Cichlid
|
AMD Radeon Linux Driver Sees Patches For New "Sienna Cichlid" GPU
|
Michael Larabel
|
Hitting the mailing list just minutes ago were a set of more than 200 patches bringing up support for the previously unheard of Sienna Cichlid GPU.
The patches indicate Sienna Cichlid is a Navi-based GPU with new VCN 3.0 capabilities for video encoding and DCN3 on the display front and a variety of other alterations compared to the existing Navi support.
It's quite possible Sienna Cichlid is the "big Navi" / RDNA2 GPU. AMD developers have talked before of using alternative codenames when volleying patches early for their open-source Linux driver stack as to not reveal the product/marketing codenames, which could be the case here. This is the first time we are hearing of Sienna Cichlid or seeing any references on the web of it related to AMD.
Still digging through the 207 patches for the AMD Radeon Sienna Cichlid, but will update if seeing anything else of note. For the most part it's leveraging the existing Navi code paths but the usual churn surrounding firmware, clock-gating / power management differences, and other modifications in the usual spots for bringing up new hardware. The main code additions primarily pertain to the new DCN3 and VCN3 blocks.
Given the timing of these patches, the AMD Sienna Cichlid won't be mainlined until the Linux 5.9 merge window opening in August and then releasing in stable around October. That timeframe at least does point to Sienna Cichlid likely being the "RDNA 2" graphics card launch coming later in the calendar year.
| 40
| 1,760,719,462.855372
|
https://www.phoronix.com/news/AMD-Energy-Driver-Linux
|
AMD Energy Driver Queued Ahead Of Linux 5.8 For Core/Package Power Sensors
|
Michael Larabel
|
Landing this weekend in hwmon-next ahead of the upcoming Linux 5.8 kernel cycle is the recently reported on "amd_energy" driver for supporting AMD Zen/Zen2 core and package energy sensors.
This is the recently reported on work of a Google engineer allowing AMD Zen CPUs to expose power usage on Linux via the Runtime Average Power Limiting (RAPL) framework. The amd_energy driver is making it to the Linux 5.8 kernel by way of the hardware monitoring "hwmon" subsystem thanks to work by AMD engineers following the prior Google patch series.
Having this reporting for modern AMD CPUs on Linux is something we have long wanted to see but something that hadn't been worked on by AMD for the Linux kernel while under Windows there is similar reporting via Ryzen Master, etc.
This amd_energy driver addition plus the recent community work on the k10temp driver now allows for good CPU thermal monitoring and core/package energy reporting under Linux. There still are some gaps on the thermal side, but at least in fairly good shape for those wanting to read those numbers under Linux.
The amd_energy driver is just over 400 lines of new code. Google engineers wrote this driver based on public AMD data sheets concerning the relevant MSRs. The numbers are reported via sysfs for those interested.
| 8
| 1,760,719,464.353126
|
https://www.phoronix.com/news/AMD-Energy-Driver-Working-Well
|
The New AMD Energy Driver Is Working Out Well On Linux For Per-Socket/Core Reporting
|
Michael Larabel
|
Of the many features coming for Linux 5.8 one of the new drivers we are very much looking forward to is the AMD energy driver for finally exposing per-core and per-socket/package energy reporting of Zen/Zen2 CPUs under Linux. It's working out well so far in my evaluation.
CPU energy/power reporting is something that I and many other Linux users have long wanted to see under Linux for Zen CPUs, since it's exposed after all on Windows with Ryzen Master and other software. In the past AMD also maintained the "fam15h_power" driver for power reporting back on Bulldozer CPUs. But until Google sent out RAPL Zen patches recently and this "amd_energy" driver was then sent out by AMD engineers, there wasn't much public activity on getting this capability for existing Zen processors. There has also been the out-of-tree "Zenpower" driver for offering this based on public MSR data for Zen, albeit that driver isn't mainline, not maintained by AMD, and conflicts with k10temp when loading.
Now this amd_energy driver is queued in hwmon-next ahead of the Linux 5.8 merge window, I've been tested out this driver and it has worked out fairly well. For those wanting to give it a whirl but not feeling like spinning your own kernel from source using the patches, here is an Ubuntu x86_64 kernel build as of the latest hwmon-next branch that I built for the amd_energy testing.
This amd_energy driver has been working out well so far for the per-core and per-socket energy reporting, finally! While Intel CPUs expose their energy usage under /sys/class/powercap/intel-rapl, the amd_energy metrics are exposed as a conventional hwmon driver.
A test run with the Threadripper 3970X above... The latest Phoronix Test Suite Git code also has support for reading the amd_energy data now as well for calculating and exposing total CPU power consumption reporting. That addition is in Git for next quarter's Phoronix Test Suite 9.8 release.
So far so good on trying the Linux amd_energy driver with different CPUs. This driver will debut official with Linux 5.8, which should be out as stable in August, but if you are comfortable doing so feel free to try hwmon-next to help ensure this driver is in great shape for finally exposing AMD Zen CPU energy data on the mainline kernel. This new driver plus the recent k10temp thermal reporting improvements (Linux 5.6+) have the recent AMD processors now in good shape on the sensor front under Linux.
| 14
| 1,760,719,464.662807
|
https://www.phoronix.com/news/Torvalds-Threadripper
|
Linus Torvalds Switches To AMD Ryzen Threadripper After 15 Years Of Intel Systems
|
Michael Larabel
|
An interesting anecdote shared in today's Linux 5.7-rc7 announcement is word that Linux and Git creator Linus Torvalds switched his main rig over to an AMD Ryzen Threadripper.
At least for what he has said in the past, Linus has long been using Intel boxes given his close relationship with the company (and even close proximity to many of the Intel Portland open-source crew). In fact, he commented this is the first time in about fifteen years not using an Intel system as his primary machine. He made this interesting remark in the RC7 announcement:
In fact, the biggest excitement this week for me was just that I upgraded my main machine, and for the first time in about 15 years, my desktop isn't Intel-based. No, I didn't switch to ARM yet, but I'm now rocking an AMD Threadripper 3970x. My 'allmodconfig' test builds are now three times faster than they used to be, which doesn't matter so much right now during the calming down period, but I will most definitely notice the upgrade during the next merge window.
The Threadripper 3970X and the rest of the 3900 series line-up are incredibly great options for kernel developers and those frequently compiling large code-bases. He didn't mention the CPU in his prior Intel box, but he is seeing 3x faster builds.
With the upcoming Linux 5.8 merge window in early June, his Threadripper system is sure to have a great workout.
This in turn is actually good news as well for AMD Ryzen Linux users: as Torvalds is constantly building the latest kernel code for mainline, he tends to shout quite publicly and loudly when any code breaks on his systems stemming from botched/poorly-tested pull requests... Thus with the extra and immediate exposure on Threadripper, he will hopefully be spotting any kernel-breaking regressions more quickly and who knows whatever other improvements he may be able to wrangle up as he's burning in his new system.
| 205
| 1,760,719,466.080397
|
https://www.phoronix.com/news/AMD-No-SFH-For-Linux-5.8
|
AMD Sensor Fusion Hub Support Is Not Coming With Linux 5.8
|
Michael Larabel
|
For those AMD Ryzen laptop users eager to see the Sensor Fusion Hub driver for supporting the different hardware sensors on these AMD Zen laptops, that driver still isn't going to be merged for the upcoming Linux 5.8 cycle even after the patches were first published months ago.
AMD SFH missed the mark for Linux 5.7 due to concerns raised over the new code at the time. Since then, there hasn't been any new patch revisions out from AMD for their SFH driver. So it's not really a surprise over it not being queued for the upcoming Linux 5.8 merge window.
The only patches we have seen on the AMD SFH front since that prior article was extending SFH support into the existing i2c-amd-mp2 driver and done by an independent community developer rather than AMD. However, an AMD engineer commented they were not interested in that duplicated effort. As another follow-up comment, it also looks like in part the SFH driver work may be delayed due to at least one of the engineers involved not having access to a SFH-based system during their lockdown / work-from-home period. But that comment was at the end of April and no other updates since on the Linux SFH support front per this thread.
But this weekend a third and final IIO pull was sent in to Greg KH for queuing ahead of the upcoming Linux 5.8 merge window. None of the IIO pulls for Linux 5.8 have the AMD SFH support, thus postponing it now until at least Linux 5.9 later in the year.
| 33
| 1,760,719,468.960225
|
https://www.phoronix.com/news/AMD-Renoir-k10temp-Linux-5.8
|
AMD Renoir Temperature Monitoring To Come With Linux 5.8
|
Michael Larabel
|
When it comes to the support for AMD Ryzen 4000 "Renoir" laptop support under Linux, as outlined in my testing so far this month the main caveat is needing Linux 5.6~5.7 for good graphics support but on the likes of Ubuntu 20.04 LTS with Linux 5.4 you will not have GPU acceleration. At least in the case of the Lenovo IdeaPad 5 I have been using to test, you also need Linux 5.7 Git for battery sensor support. Another item that in turn is coming with Linux 5.8 is CPU temperature reporting for the Renoir processors.
Currently on Linux the k10temp driver for Zen/Zen2 CPU core temperature reporting doesn't work under Linux, but the necessary patches were queued up a few minutes ago in the ras/core branch ahead of the upcoming Linux 5.8 merge window. The changes are fairly trivial but unfortunately coming by the community after Renoir laptops are already shipping.
amd_nb.c needed the PCI IDs for Family 17h Model 60h (Renoir). In turn k10temp added the new PCI IDs to match in order to begin reporting Tdie and Tctl values. (The Renoir IDs were also added to the AMD EDAC driver.)
With the PCI IDs added, Renoir CPU temperature reporting will now work under Linux once these patches are merged to mainline (Linux 5.8).
| 20
| 1,760,719,470.712298
|
https://www.phoronix.com/news/TUXEDO-Book-BA15-AMD-Laptop
|
TUXEDO Computers Launches Their First AMD Linux Laptop
|
Michael Larabel
|
TUXEDO Computers has launched their first AMD-powered Linux laptop! The excitement quickly faded though when seeing it's not a Renoir design.
While announcing the TUXEDO Book BA15 this weekend as their first AMD laptop, sadly it's based on a Ryzen 5 3500U and not the current Ryzen 4000 series with its mighty impressive performance and power efficiency gains thanks to the Zen 2 CPU cores. This though still should be a decent showing with the Ryzen 5 3500U with Vega graphics, metal chassis, 91 Wh battery, up to two SSDs, up to 32GB RAM, and 15.6-inch 1080p display. But given the timing just a shame it's not a current generation Ryzen 4000 laptop given the enticing performance from the latest hardware. Also a bummer with this design is that it seems to be relying upon single channel memory.
More details on the TUXEDO Book BA15 can be found at TUXEDOComputers.com. Pricing starts at 859 EUR for the configuration with 8GB RAM and 256GB SSD storage.
For those looking for a Ryzen 4000 series Linux laptop, I continue to be quite happy with the Ryzen 7 4700U with the Lenovo IdeaPad 5 from its build quality to performance. Renoir is running well on Linux when say using Ubuntu 20.04 LTS and then upgrading to Linux 5.6~5.7 for the graphics support and battery sensors while the last big holdout is the k10temp support hitting Linux 5.8. The Ryzen 4000 mobile performance is quite incredible.
| 37
| 1,760,719,470.926147
|
https://www.phoronix.com/news/Google-RAPLs-AMD-Zen
|
Google Posts Patches Allowing AMD Zen/Zen2 CPUs To Expose Power Usage On Linux Via RAPL
|
Michael Larabel
|
One of the long sought after features for AMD Zen (and Zen 2) processors on Linux has been the ability to monitor the CPU package power consumption on Linux, similar to what's long been available for Intel CPUs on Linux and similarly for older AMD Bulldozer era CPUs with a power monitoring driver. Now on Friday evening a patch series was posted by a Google engineer to provide this long sought after functionality.
Google engineer Stephane Eranian posted a patch series to expose the necessary hardware counters for reading the power usage on Linux and tie into the kernel's RAPL framework. The Running Average Power Limit "RAPL" interface was devised by Intel years ago and supported by their hardware for being able to read (and, in some cases, limit) the power consumption. The power usage is exposed via a PowerCap sysfs interface and via a perf interface.
With the patches posted by Google that come in at less than 60 lines of code, it allows finally being able to read the CPU power consumption based upon the data exposed via its counters. With the current patches, the energy package level consumption is exposed. Both Zen 1 and Zen 2 CPUs are supported.
It's great to see these patches finally materialize but long overdue especially with the AMD Windows software (Ryzen Master, et al) long exposing the power metrics. It's been a topic I've long brought up with AMD as an area to improve, similar to the AMD Zen thermal monitoring Linux driver support being another item left up to the community to tackle.
| 29
| 1,760,719,472.749322
|
https://www.phoronix.com/news/Wraith-Master-1.0
|
Wraith Master 1.0 Released For Controlling AMD RGB Fans On Linux
|
Michael Larabel
|
Wraith Master 1.0 has been released as the "feature complete" version of this Linux GUI application for providing RGB lighting controls for the AMD Wraith Prism heatsink under Linux.
The open-source Wraith Prism was written via reverse-engineering in analyzing the USB protocol via Wireshark. With Wraith Prism 1.0, there should be similar functionality via the CLI and GUI to the official Windows application.
Wraith Prism is written against the GTK3 tool-kit and interestingly uses the Kotlin programming language.
More details on Wraith Prism 1.0 via the release announcement. The code for this AMD independent RGB lighting solution is available via GitLab.
Another similar solution we covered last year is the open-source CM-RGB.
| 45
| 1,760,719,472.760867
|
https://www.phoronix.com/news/AMD-GPUOpen-Relaunched
|
AMD Relaunches GPUOpen For Open-Source Game Development Resources
|
Michael Larabel
|
AMD announced this morning the relaunch of GPUOpen, the company's effort on providing open-source development resources primarily for game developers in making use of various Radeon GPU technologies and other open standards across Windows and Linux.
AMD is marking the relaunch with "releasing new GPUOpen tools and technologies every day this week." Announced for Monday is an expansion of FidelityFX as their open-source toolkit for high quality post-process game effects. There is now support for screen space reflections, ambient occlusion, HDR mapper, and a downsampler as part of FidelityFX.
More details on the GPUOpen relaunch via this AMD blog post. Head on over to GPUOpen.com for their new and improved open-source game / content creation development portal.
| 29
| 1,760,719,474.039735
|
https://www.phoronix.com/news/Next-Gen-EPYC-Does-MPK-Linux
|
New Linux Patches Confirm Next-Gen EPYC "Milan" CPUs To Support MPK
|
Michael Larabel
|
Recently we noted the latest revision to AMD's Programmer Reference Manual points to PCID and MPK/PKEY support coming on future CPUs. New Linux patches now confirm that the MPK support is on the way with next-gen EPYC processors.
The Memory Protection Key (PKE) support has been supported by Intel processors since Xeon Skylake and allow for page-based memory protections with the RDPKRU and WRPKRU instructions. As noted in the PRM, the feature "provides a way for applications to impose page-based data access protections (read/write, read-only or no access), without requiring modification of page tables and subsequent TLB invalidations when the application changes protection domains."
Sent out today were a set of patches reusing the Linux kernel's X86_INTEL_MEMORY_PROTECTION_KEYS code path and supporting MPK within the AMD SVM KVM virtualization code.
The patch comment spells it out quite clearly that this is for next-gen EPYC CPUs, "AMD's next generation of EPYC processors support the MPK (Memory Protection Keys) feature."
The patches are quite straight-forward in re-using the kernel's existing MPK code plumbed in previously by Intel and just two dozen or so lines of code for the AMD SVM code within the Kernel-based Virtual Machine (KVM).
It's possible this AMD MPK support will be mainlined for Linux 5.8. The patches may be revised as an upstream Intel developer has asked not to have the Kconfig X86_INTEL_MEMORY_PROTECTION_KEYS option to the proposed X86_MEMORY_PROTECTION_KEYS to avoid configuration file woes when transitioning to new kernel versions, so we'll see, but in any case it's a small enough change that there still is time to get this material queued for the forthcoming 5.8 cycle kicking off in June and should be out as stable in August.
This goes along with a string of other open-source work we have seen in recent weeks/months around Zen 3. AMD EPYC "Milan" Zen 3 based processors are expected to be released by the end of this calendar year.
| 6
| 1,760,719,474.988327
|
https://www.phoronix.com/news/AMD-GNU-Discussing-Better-RT
|
AMD Working With GNU Developers To Provide More Robust Runtime Detection For Better Performance
|
Michael Larabel
|
Back in March we reported how AMD developers were looking at GNU C Library platform optimizations for Zen and in part could be leveraging some of the capabilities currently employed by Intel for Haswell and newer. It's looking like some solid progress is being made in that direction.
The patches from AMD in March provided better run-time detection for CPU features like AVX2 and could allow making use of more optimized code-paths at run-time when run on such hardware, similar to Intel Glibc optimizations for Haswell and newer. This would be a big win not only for AMD Zen but also current Zen 2 CPUs and future Zen 3 parts, etc in basically providing more reasonable optimized code-paths at run-time for prominent CPU instruction set extensions.
Issues were pointed out in the early AMD proposed code, but it looks like AMD engineers are better communicating with the GNU toolchain folks and a solution could be coming.
Florian Weimer of Red Hat commented back in February that this bug report over AMD EPYC/Zen CPUs should default to the Haswell platform directory as a better default has a new comment. While in February was the comment by Weimer that he was blocked until getting feedback from AMD, at the end of April he added:
I'm happy to report that I've been in contact with the right people at AMD for a while.
I do not know yet what the exact outcome will be (if the “haswell” directory will be used), but there will be a way to automatically load AVX2-optimized libraries on AMD CPUs as well. Finally! Let's hope this will be settled soon for seeing greater AMD performance on Linux with greater AVX2, delivering even greater performance than what AMD's desktop and server CPUs have already been delivering out-of-the-box on Linux.
| 16
| 1,760,719,475.895061
|
https://www.phoronix.com/news/AMD-PRM-PCID-PKEY
|
AMD Programmer Manual Update Points To PCID Support, Memory Protection Keys
|
Michael Larabel
|
It looks like AMD Zen 3 CPUs will finally be supporting PCID! And memory protection keys are coming too, at least according to AMD's latest programmer reference manual.
AMD has published a new revision of their Programmer's Reference Manual. The new registers detailed are for PCID and PKEY.
Yes, PCID as in Process Context Identifiers are finally being supported on AMD CPUs. PCID has been supported by Intel CPUs since 2010 while since 2018 have been talked about much more for helping to offset some of the performance overhead stemming from Meltdown mitigations as PCID allows avoiding an extra TLB flush. While Intel has long supported PCID, AMD CPUs to date have not supported Process Context Identifiers. The Linux kernel has supported making use of Process Context Identifiers and the INVPCID instruction since around Linux 4.14 with PCID optimized TLB flushing.
As outlined in the PRM, "The Process Context Identifier (PCID) feature allows a logical processor to cache TLB mappings concurrently for multiple virtual address spaces. When enabled (by setting CR4.PCIDE=1), the processor associates the current 12-bit PCID with each TLB mapping it creates. Only entries matching the current PCID are used when performing address translations. In this way, the processor may retain cached TLB mappings for multiple contexts."
The other notable addition with today's updated manual is Memory Protection Key (PKE) support. The AMD PKE support also enables the RDPKRU and WRPKRU instructions. Intel provided that Linux support for years and on the CPU side has been supported since Xeon Skylake. Memory protection keys allow for page-based memory protections via the RDPKRU/WRPKRU instructions. As explained in the AMD manual, "[MPK] provides a way for applications to impose page-based data access protections (read/write, read-only or no access), without requiring modification of page tables and subsequent TLB invalidations when the application changes protection domains."The Linux kernel, GCC and Clang, and other key libraries like Glibc already have memory protection key support in good order thanks to Intel's earlier work.
While not guaranteed, given the timing of the manual update, the PCID and PKE/MPK functionality is almost surely part of the AMD Zen 3 processors launching later this year.
The latest AMD manual can be found here.
| 6
| 1,760,719,476.524051
|
https://www.phoronix.com/news/RADV-ACO-Ubuntu-20.04-Test
|
For Radeon Gamers On Ubuntu 20.04 LTS, It's Generally Worthwhile Flipping On RADV's ACO
|
Michael Larabel
|
A premium supporter was asking this week whether for those newly-upgraded to Ubuntu 20.04 LTS if the graphics stack is in good enough shape or if I would recommend running Mesa 20.1-devel for better AMD Linux gaming performance... The short answer, sans any particular changes you are after in Mesa 20.1-devel, the bigger gain for running on this new Ubuntu release is to instead enable RADV+ACO as a much more pressing boost.
Following the Phoronix Premium member's request, I ran some Vega and Navi benchmarks to show the current difference of Ubuntu 20.04 LTS with its Mesa 20.0 build compared to Mesa 20.1-devel via the Oibaf PPA as of a few days ago. The default (Linux 5.4 based) kernel and other software packages were kept at the same versions.
But besides just running Mesa 20.0 vs. 20.1-devel, I also tossed in ACO benchmarks. For the RADV Vulkan driver, ACO can be enabled with the RADV_PERFTEST=aco environment variable.
Across various benchmarks, the Mesa 20.0 vs. 20.1-devel performance wasn't that much different for tests I ran on both OpenGL and Vulkan. But the more interesting takeaway remains the much better performance for the RADV+ACO combination for faster Vulkan gaming on Linux, including under Steam Play.
If you haven't tried it yet, I would certainly recommend giving RADV+ACO a go for achieving better AMD Radeon Linux gaming performance over simply switching to Mesa 20.1-devel. With Mesa 20.0, ACO is generally in great shape and getting even better for Mesa 20.1 that will be out as stable by the end of May. See more cross-AMD-driver configurations in the recent Radeon Software 20.10 vs. upstream comparison.
More benchmarks soon including a look at whether changing out the Linux kernel build for Ubuntu 20.04 LTS is worthwhile on the open-source Radeon graphics driver stack.
| 15
| 1,760,719,477.673221
|
https://www.phoronix.com/news/Navi-12-14-Displayable-DCC
|
Radeon Displayable DCC Gets Enabled For Navi 12 + Navi 14 GPUs
|
Michael Larabel
|
Adding to the last minute AMD Radeon additions for making the Mesa 20.1 feature cut-off is enabling displayable DCC support for Navi 12 and Navi 14 graphics processors.
GFX10/Navi has already supported delta color compression for saving video memory bandwidth while this change is about "displayable DCC", or DCC for surfaces being scanned out to the display for benefiting from color compression.
AMD added displayable DCC back with Raven Ridge Up to now the AMD Mesa code has just enabled display DCC for Raven/Raven 2 APUs along with Renoir. But now paired with the latest Linux kernel and Mesa Git, display DCC should be working for Navi 12 and Navi 14 GPUs for further bandwidth savings with this lossless compression means. There isn't any support in place for display DCC on Navi 10.
This displayable DCC for Navi 12/14 was merged this morning via this MR ahead of the upcoming Mesa 20.1 feature freeze.
| 9
| 1,760,719,478.097905
|
https://www.phoronix.com/news/AMD-SPI-SoC-Linux-Ready
|
AMD SPI SoC Linux Driver Appears Ready For Mainline In Linux 5.8
|
Michael Larabel
|
The new spi-amd driver looks like it could make its debut in Linux 5.8.
Published earlier this month was an SPI controller driver for AMD SoCs. This 333 lines of code driver enables Serial Peripheral Interface (SPI) support for AMD SoCs with the "AMDI0061" device.
Sent out this weekend was the latest AMD SPI driver code. Mark Brown who oversees the upstream SPI area already commented that the V2 driver now looks good. So assuming no further objections are raised, this new AMD driver will quite likely end up in the Linux 5.8 cycle this summer. Nothing too exciting about the driver itself: just supporting SPI with FIFO transfers on relevant SoCs.
Meanwhile on other AMD awaited Linux drivers there isn't anything new to report yet on the Sensor Fusion Hub driver in development for laptops nor any CPPC / Schedutil AMD update, but hopefully we'll still see work on these drivers in the near-term.
| 0
| 1,760,719,479.307271
|
https://www.phoronix.com/news/AMD-Academy-Software-OSL
|
AMD Joins The Academy Software Foundation, Open Shading Language Becomes Hosted Project
|
Michael Larabel
|
Some news that went seemingly unreported at large this month until stumbling across it... AMD has joined the Academy Software Foundation as a premier member. This joint initiative between the Linux Foundation and Academy of Motion Picture Arts is focused on pushing open-source software through the motion picture and content creation industries. Additionally, the Open Shading Language has become a new hosted project under the Academy Software Foundation.
AMD has joined the Academy Software Foundation as a premier member for pushing open-source in this industry. AMD joins the likes of The Walt Disney Studios, Sony Pictures, Red Hat, Pixar, NVIDIA, Netflix, Microsoft, Intel, Epic Games, Google Cloud, and others backing this foundation.
With AMD's interest, they are certainly interested in seeing more EPYC and Threadripper CPUs and Radeon Pro hardware used by content creators.
The academy also announced the same day that the Open Shading Language has become their sixth project. This Sony-backed shading language for VFX and animation will now be developed via this open-source foundation. Those unfamiliar with OSL can learn more -- including where it's currently supported and the numerous films already having made use of it -- via GitHub.
Besides OSL, the other projects currently under the Academy Software Foundation umbrella are OpenVDB for efficient storage and manipulation of sparse volumetric data discretized on three-dimensional grids, OpenTimelineIO as an API/interchange format for editorial timeline information, OpenEXR as the HDR raster file format, and OpenCue as a render management system.
More details on the foundation's latest work at ASWF.io.
| 2
| 1,760,719,479.634347
|
https://www.phoronix.com/news/AMD-April-21-2020-Announcement
|
AMD Announces The Ryzen 3 3100 + Ryzen 3 3300X Processors
|
Michael Larabel
|
AMD has expanded its highly successful Zen 2 family with some new low-end parts.
Announced today are the Ryzen 3 3100 and Ryzen 3 3300X processors as the long-awaited Zen 2 processors for those trying to assemble a tidy PC on a budget.
The Ryzen 3 3100 is a four core / eight thread processor with a 3.6GHz base frequency and 3.9GHz boost clock while having a 65 Watt TDP.
The Ryzen 3 3300X is a quad-core / eight thread processor with a 3.8GHz base clock and 4.3GHz boost clock while still having a 65 Watt TDP.
These new Ryzen 3 3000 series processors support the common Zen 2 features like PCIe 4.0, 7nm, DDR4-3200 support, etc. These new Ryzen 3 processors are expected to retail for around $100 USD while more firm pricing should come soon.
More information and benchmarks soon while the basic information on these new CPUs can be found via AMD.com.
AMD also announced with these new Zen 2 budget processors that the B550 chipset will begin appearing with new motherboards in June.
| 75
| 1,760,719,481.426912
|
https://www.phoronix.com/news/Radeon-Software-For-Linux-20.10
|
Radeon Software For Linux 20.10 Driver Released
|
Michael Larabel
|
AMD has finally released their first "Radeon Software for Linux" packaged driver release to succeed their Radeon Software for Linux 19.50 driver series that saw its last update in December. Radeon Software for Linux 20.10 is available today as their first packaged Linux driver update of 2020 for AMD Radeon Linux owners as the packaged solution intended for easy installation of their All-Open and "PRO" driver components.
While four months has passed since their last Radeon Software for Linux release, the official release notes are quite light on the changes:
Introducing full support for Ubuntu 18.04.4 HWE
Introducing full support for RHEL 7.8
Introducing support for the AMD Radeon™ RX 5600 family of cards
The Radeon RX 5600 series support was well overdue albeit has been supported since day zero with the open-source driver on the latest upstream components. The RHEL 7.8 and Ubuntu 18.04.4 LTS support is also overdue and sadly this driver doesn't provide support for Ubuntu 20.04 LTS due to be released next week.
The officially supported Linux distributions by Radeon Software for Linux 20.10 are Ubuntu 18.04.4, RHEL/CentOS 7.8, RHEL/CentOS 8.1, and SUSE Linux Enterprise Desktop/Server 15 SP1.
Besides the expanded Navi support and new RHEL/Ubuntu point release support, no other items are noted but there should be various improvements to their OpenGL and Vulkan drivers both as part of the All-Open and "PRO" driver components. I'll be working on some fresh AMD Radeon Linux driver comparison benchmarks over the weekend.
Those wanting to try out the Radeon Software for Linux 20.10 release can find it at AMD.com.
| 24
| 1,760,719,481.606164
|
https://www.phoronix.com/news/Linux-AMD-TRX40-Audio-Quirks
|
Linux Seeing Fixes For AMD TRX40 Motherboard Audio Issues
|
Michael Larabel
|
Various patches are pending for improving the Linux support for onboard audio with motherboard sporting the AMD TRX40 chipset for 3rd Gen Ryzen Threadripper systems.
There are a number of known issues at present affecting the integrated audio on numerous TRX40 motherboards including the likes of the MSI TRX40 Creator, ASUS ROG STRIX, ASUS ROG Zenith II, Gigabyte TRX40 Aorus Pro, and others.
Under Linux with recent kernels there are issues over no audio codecs being found and no sound output via SPDIF, among similar bug reports elsewhere.
Currently being queued in a usb-trx40 sound branch are the workarounds being prepared for better handling of the audio hardware on TRX40 motherboards. The issues appear to originate in USB audio firmware bugs. These workarounds for the Linux kernel code in turn will hopefully soon be mainlined soon and back-ported to recent kernels.
| 7
| 1,760,719,483.139817
|
https://www.phoronix.com/news/AMD-Larger-Microcode-Linux
|
The Linux Kernel Prepares For Larger AMD CPU Microcode Updates
|
Michael Larabel
|
Future AMD CPUs (more than likely, Zen 3) will be bearing larger CPU microcode sizes, resulting in the Linux kernel needing a change to load them.
Currently the AMD Linux CPU microcode handler has a static upper limit of the page size, which is generally 4K. But in preparation for "future AMD CPUs" that will exceed that size, that upper limit is being bumped.
With a change now pending as part of the x86/urgent work, that upper limit is being bumped to a factor of three times the page size. In other words, up to 12K microcode update size.
It will be interesting to see if they are tripling the limit for future-proofing and Zen 3 just crossing the existing threshold slightly or if some fundamental change/addition in Zen 3 is leading to much larger microcode updates for AMD CPUs moving forward.
That patch is also being called for back-porting back through to the Linux 4.14 LTS days with CPU microcode updating being important particularly these days, though generally for other aspects of AMD Zen 3 support it's like with all new hardware: the newer the Linux kernel, likely the better the support experience and performance. In recent months AMD has begun sending out a lot more patches for AMD Family 19h / Zen 3 plumbing to the Linux kernel ahead of the first of those CPUs expected later this calendar year.
| 44
| 1,760,719,483.15291
|
https://www.phoronix.com/news/AMD-PTDMA-Driver-No-Queue-V3
|
AMD PassThru DMA Engine Driver Still Pending For The Linux Kernel
|
Michael Larabel
|
In addition to the AMD Sensor Fusion Hub driver that we are hopeful could land in Linux 5.7 albeit not yet queued in the iio-next branch, another AMD driver that has been around for a few months in patch form but yet to be mainlined is the AMD PassThru DMA Engine driver.
The AMD PassThru DMA Engine driver was volleyed last year and saw two additional rounds of revisions but has been quiet since the end of January. As of writing, it hasn't yet made it into the DMA-next area ahead of the Linux 5.7 merge window expected to open in early April.
The AMD PassThru DMA (PTDMA) engine allows for high bandwidth memory-to-memory and IO copy operations and direct memory access transfers. Each AMD CPU has multiple PTDMA device instances.
The driver for AMD PTDMA in total comes in at just more than two thousand lines of code. Hopefully there still is chances of seeing the AMD_PTDMA driver queued for Linux 5.7 considering no other issues were raised in the latest patch round.
| 11
| 1,760,719,484.474666
|
https://www.phoronix.com/news/AMD-SHF-Driver-Needs-Work
|
AMD Sensor Fusion Hub Laptop Driver Unlikely To Land For Linux 5.7
|
Michael Larabel
|
While we were hoping to see the AMD Sensor Fusion Hub driver introduced in Linux 5.7 for improving the AMD Ryzen Linux laptop experience, that now looks quite unlikely.
This driver has been sought after by AMD Linux laptop customers since 2018 for supporting the accelerometer, gyroscopic sensors, and other functionality on modern AMD laptops, similar to the Intel Sensor Hub. Patches for the AMD Sensor Fusion Hub (AMD-SFH) driver for Linux were posted in January and underwent a few rounds of review.
After going through a few rounds of review, we've been hoping to see it for Linux 5.7 especially with the new AMD Renoir laptops beginning to ship in the near future. Sadly, however, Linux 5.7 looks out of scope for landing AMD-SFH.
The Linux 5.6 release potentially happening later today and that in turn opening the two-week-long Linux 5.7 merge window, but the AMD-SFH driver has yet to be pulled into the IIO subsystem's "-next" branch. But more pressing is that new issues with the code have now been raised.
On Friday upstream long-time kernel developer Andy Shevchenko reviewed the driver and found a number of faults with the code. The Intel engineer summed it up as "TL,DR; it requires a bit of work."
From missing header includes to seemingly bizarre to "completely useless" code that questions how much it was tested, the driver looks unlikely to see mainlined in its current form. The code review comments can be found via this mailing list post.
Given the changes/clean-ups needed, it looks almost certainly out of scope nope for seeing it potentially in the imminent Linux 5.7 merge window. That would punt it off to Linux 5.8 at least -- its merge window should be in June and see a stable release around August. At least if AMD-SFH gets into Linux 5.8, that should be the kernel used by the likes of Ubuntu 20.10 this summer.
For those wondering about the current state of AMD Ryzen laptop support on Linux, I do intend to buy an AMD Ryzen 4000 series laptop for Linux testing soon as I find an interesting model available at a decent price point. These laptops should start coming in the weeks ahead and given the successes I've had with AMD Zen 2 on desktops and servers, actually looking quite forward to seeing how well the Ryzen 4000 series will perform, albeit disappointed in cases like this of the much belated AMD-SFH support on the Linux side. If you too are interested in Ryzen 4000 Linux laptop testing, consider showing your support by joining Phoronix Premium or making a PayPal tip.
| 15
| 1,760,719,484.494531
|
https://www.phoronix.com/news/AMD-SB-TSI-Linux-Driver
|
Google Engineers Have Been Working On An AMD SB-TSI Temperature Driver
|
Michael Larabel
|
Google open-source engineers have been working on a temperature driver for AMD's SoC SB-TSI emulated temperature sensor for the Linux kernel.
The SB-TSI (Sideband Temperature Sensor Interface) is a SMBus compatible interface for reading the AMD SoC temperature interface connected to a BMC. The driver was posted this week by Google's Kun Yi who is part of the Google Platforms Infrastructure Server Software team, which doesn't come as much surprise considering the AMD EPYC successes in the data center.
The Sideband Temperature Sensor Interface is publicly documented but has lacked a public Linux driver to this point.
The SBI temperature sensor interface (SB-TSI) is an emulation of the software and physical interface of a typical 8-pin remote temperature sensor (RTS) on AMD SoCs. It implements one temperature sensor with readings and limit registers encode the temperature in increments of 0.125 from 0 to 255.875. Limits can be set through the writable thresholds, and if reached will trigger corresponding alert signals. The driver amounts to just under 300 lines of code and jives with the Linux kernel's hardware monitoring "hwmon" subsystem. We'll see if it manages to land still for the upcoming Linux 5.7 merge window or gets pushed off for a later cycle.
| 30
| 1,760,719,485.882351
|
https://www.phoronix.com/news/GNU-libc-Platform-Optimize-Zen
|
AMD Developers Looking At GNU C Library Platform Optimizations For Zen
|
Michael Larabel
|
It's long overdue but AMD engineers are now looking at refactoring the GNU C Library (Glibc) platform support to enhance the performance for AMD Zen processors.
Stemming from Glibc semantics that effectively "cripple AMD" in just checking for Intel CPUs while AMD CPUs with Glibc are not even taking advantage of Haswell era CPU features, AMD developers are now looking at properly plumbing AMD Zen platform support into this important C library for Linux users.
Under a "request for comments" flag, patches tentatively posted add AMD Zen and AVX/AVX2 platform support and refactor the platform support within the CPU features detection. This would at run-time allow CPU features like AVX2, FMA, BMI2, POPCNT, and other instructions to be enabled when detected to be running on an AMD Zen based processor. There are also fall-backs for "generic" AVX2 and then AVX CPUs for older CPUs.
The patches are currently under discussion with some upstream GNU toolchain developers providing their recommendations on how they would like to see this more diverse platform support handled.
Hopefully this Glibc support for AMD Zen can be sorted out in short order and mainlined. It's a pity that it has taken this long to get sorted out but it's nice to see the recent uptick in AMD's Linux/open-source patches particularly around these fundamental improvements/optimizations. It's also too bad this didn't start a few months earlier so it could have stood chances for seeing such an improved library within the likes of the upcoming Ubuntu 20.04 LTS release that will see many AMD EPYC server deployments.
It looks like the surge in AMD CPU sales, expanding EPYC cloud deployments, and continuing super-computer contract wins that are principally Linux based are at least in part helping to foster some of this open-source work. It's not that the GNU C Library has been intentionally thwarting AMD CPUs but rather Intel developers for years have invested heavily in toolchain improvements and other areas of GNU/Linux to take advantage of their latest CPU microarchitectures. Meanwhile AMD has traditionally contributed much less to these upstream projects directly but that is fortunately now changing. Going back years to when Intel was punctually hitting their CPU release schedules on time, even then they would generally have out their new compiler/toolchain optimizations upstreamed well ahead of those product launches for ensuring good launch-day customer support on Linux.
| 48
| 1,760,719,485.943551
|
https://www.phoronix.com/news/AMD-PPIN-Processor-ID-Linux
|
AMD Plumbing Linux Support For Reading The CPU's Protected Processor Identification Number (PPIN)
|
Michael Larabel
|
Going back to Ivy Bridge processors, Intel has supported "PPIN" as the Protected Processor Identification Number as a globally unique identification number set in the factory. It turns out recent AMD CPUs are also supporting PPIN and that reading their value is about to be supported on Linux.
The Protected Processor Identification Number (PPIN) is effectively a unique serial number for each processor. One of the intended use-cases for PPIN is in large data centers and multi-socket servers to be able to more easily identify a particular CPU, especially in case of problems. At least in Intel's case, Intel can also translate a customer's PPIN number back into the fab and production run of that particular CPU along with any other internal data in isolating any issues. Intel has supported reading the PPIN under Linux for years and plumbed it into the MCE (Machine Check Exception) code for allowing server administrators to potentially more easily identify a particular CPU in the event of problems as well as tracking CPU inventory.
AMD developers have been working on a patch for PPIN support within the AMD MCE code. The patch basically follows Intel's implementation given their implementations of PPIN are nearly identical. The patch only mentions "newer" AMD CPUs support this feature and not specifying if that means Zen 2 or forthcoming Zen 3 or potentially only implemented by EPYC CPUs (the actual code is simply checking a particular feature bit, waiting to hear any clarification from AMD).
For now the AMD PPIN number is just being reported as part of the MCE logs and not exposed by other means, for those concerned about this globally unique CPU identifier being used for tracking purposes.
| 21
| 1,760,719,487.435024
|
https://www.phoronix.com/news/AMD-SEV-ES-Guest-Support-2
|
AMD SEV-ES Guest Support Updated With More Improvements, Rebased
|
Michael Larabel
|
Back in February came patches for AMD SEV-ES "Encrypted State" support as building off the Linux kernel's existing support for Secure Encrypted Virtualization in conjunction with AMD EPYC processors. The SEV-ES enablement work has now been revised.
The SEV "Encrypted State" patches sent out this morning are for enabling Linux to run as a guest under an SEV-ES enabled hypervisor. The encrypted state portion of SEV is about protecting the guest register state from the hypervisor, beyond the memory encrypted afforded by SEV. The CPU register state becomes encrypted by SEV-ES and cannot be accessed or modified by the hypervisor in order to fend off control-flow attacks and other similar attacks.
With the v2 patches sent out today for SEV-ES guest support, the patches have been re-based against the latest Linux 5.6 Git code, emulation of REP/MOVS instructions is now in place, other instruction handling improvements, and some bug fixes.
It's cutting close though whether the updated patches could be reviewed in time for possible inclusion in the forthcoming Linux 5.7 merge window otherwise could be pushed off until at least Linux 5.8 later in the summer before seeing this functionality in place.
| 0
| 1,760,719,487.462954
|
https://www.phoronix.com/news/AMD-Ryzen-9-4900H
|
AMD Ryzen 9 4900H Mobile Processor Announced For Top-End Laptop Performance
|
Michael Larabel
|
AMD today announced the Ryzen 9 4900H as their new top-end Zen 2 mobile processor for notebooks.
The AMD Ryzen 9 4900H is an eight-core / sixteen thread part with a 3.3GHz base frequency and 4.4GHz boost clock. This Renoir processor comes with a 45 Watt TDP given its 8c/16t setup and high clock frequencies.
Coming in below the Ryzen 9 4900H in the H-series is the Ryzen 7 4800H as an 8c/16t 2.9/4.2GHz part as well as the Ryzen 5 4600H at 6c/12t 3.0/4.0GHz configuration. All of these H-series parts have a 45 Watt TDP and 12MB cache.
This mobile processor will begin appearing in devices this spring. More details on AMD.com.
Once AMD Renoir laptops are available in the US, I intend to pick up one for benchmarking given that the Ryzen 4000 series mobile processors are looking quite competitive against Intel. If you are interested in seeing the AMD mobile Linux benchmarks please consider joining Phoronix Premium or making a PayPal tip to help make this testing possible and to ideally allow for a more interesting Renoir laptop rather than just the cheapest unit available.
| 19
| 1,760,719,488.955355
|
https://www.phoronix.com/news/AMD-Sensor-Fusion-Hub-No-Queue
|
AMD Sensor Fusion Hub Driver Isn't Yet Queued For Linux 5.7
|
Michael Larabel
|
While a lot of feature work has been building up for Linux 5.7 in various subsystem development repositories ahead of the merge window in a few weeks, one of the big driver additions many users have been clamoring for isn't yet queued. The AMD Sensor Fusion Hub open-source driver for Linux appears stalled pending more reviews from upstream developers.
Back in January AMD published the long awaited Sensor Fusion Hub support for Linux that AMD Linux laptop customers have been after since 2018. The Sensor Fusion Hub (SFH) driver is needed for supporting various laptop sensors like the accelerometer and gyroscope. The amd-sfh-hid driver is similar to Intel's Sensor Hub that has seen Linux support from the start for supporting the sensors on various Intel-powered laptops.
A second revision to the AMD SFH Linux driver was posted at the end of January but didn't make it for the Linux 5.6 kernel cycle.
In February the third and latest version of the AMD SFH Linux driver was posted following upstream code review.
Sadly, while the Linux 5.7 merge window is just a few weeks away, this driver isn't queued yet via the IIO tree.
Last week the AMD developer involved reached out on the Linux-IIO mailing list to try to find out the status and/or encourage any remaining code review. So far though no response or queuing of this driver that is important to AMD Linux laptop users.
Hopefully we'll see this driver still get picked up by the IIO maintainers in time for Linux 5.7, especially with the AMD Ryzen 4000 "Renoir" laptops starting to appear in the days ahead and many of them likely employing Sensor Fusion Hub functionality.
| 11
| 1,760,719,489.122387
|
https://www.phoronix.com/news/Family-19h-Zen3-Init-Linux-5.7
|
More AMD Family 19h (Zen 3) Code Trickling Into Linux 5.7
|
Michael Larabel
|
We continue to see bits here and there of AMD Family 19h / Zen 3 support coming together for the mainline Linux kernel.
For Linux 5.6 has been some Family 19h PCI IDs being added to existing AMD Zen support code paths. Last month were a few 19h patches for the perf subsystem.
The latest is a patch now queued in the x86/cpu branch ahead of the Linux 5.7 merge window opening in a few weeks. That patch is extending the AMD kernel code to now also call the AMD Zen init function (init_amd_zn()) for Family 19h CPUs. "Family 19h CPUs are Zen-based and still share most architectural features with Family 17h CPUs, and therefore still need to call init_amd_zn() e.g., to set the RECLAIM_DISTANCE override."
The patch also confirms Family 19h "Zen 3" has the same mitigation handling around SSBD (Speculative Store Bypass Disable / Spectre V4). Also, the patch confirms Family 19h isn't impacted by the earlier Zen CPU erratum around the Core Performance Boost (CPB) bit potentially not being set. Neither of those bits of information come as a surprise and no juicy details revealed as part of these Family 19h patches so far.
We'll see what other AMD Family 19h support gets squared away in time for Linux 5.7. If this flow of Family 19h patches keep up, the support should get ironed out in time for offering nice out-of-the-box support by the Linux distribution releases this fall (Ubuntu 20.10, Fedora 33, etc). AMD Zen 3 processors are expected to begin shipping later this calendar year.
| 2
| 1,760,719,490.567977
|
https://www.phoronix.com/news/AMD-Linux-Hire-Lead-Kernel-Dev
|
AMD Is Hiring Another Lead Linux Kernel Developer To Work On Their Graphics Driver
|
Michael Larabel
|
Should you be experienced in upstream Linux kernel development, AMD is hiring a lead Linux kernel developer.
A recent AMD job posting puts AMD still looking for a lead Linux kernel developer after initially hearing of the position towards the start of the year. This Linux kernel developer position is on their GPU driver side for AMDGPU as well as focused on APU product support but appears to be less so about Zen/CPU efforts.
AMD is looking for someone who is already well experienced on Linux kernel development and ideally having GPU programming experience with the likes of CUDA/OpenCL/HIP and machine learning frameworks.
More details on this lead Linux kernel developer position here at their Austin campus. They also have out several other Linux positions throughout their various offices. It's great seeing AMD continuing to ramp up their open-source/Linux investments, not too surprising though considering their recent string of supercomputing contract wins and this position further reinforcing the GPU computing aspect of the work.
| 34
| 1,760,719,491.184928
|
https://www.phoronix.com/news/Threadripper-3990X-Linux-5.6
|
Linux 5.1 To Linux 5.6 Benchmarks On The AMD Ryzen Threadripper 3990X
|
Michael Larabel
|
For those wondering how the Threadripper 3990X is performing with the upcoming Linux 5.6 kernel, here are benchmarks of a recent development snapshot of Linux 5.6 as well as benchmarking all of the major kernel releases going back to Linux 5.1.
This Threadripper 3990X testing continues to be done from the mighty powerful and beautiful System76 Thelio Major workstation. A wide variety of workloads were benchmarked for Linux 5.1 through Linux 5.6 on this 64-core / 128-thread beast.
Though due to the Navi graphics not being supported back to Linux 5.1, there are not graphics tests as part of this comparison (but there is a Linux 5.6 AMDGPU comparison separately for those interested).
IO_uring is one of the areas that has improved nicely since its introduction in Linux 5.1... Mostly on the feature front but for some of the I/O tests, there have been performance improvements since 5.1 as well.
The pmbench paging/VM benchmark did see some minor improvements on newer kernels.
iPerf was reporting lower localhost TCP socket performance since Linux 5.3.
The Hackbench scheduler benchmark was also seeing worse performance since Linux 5.3.
For most of the heavy CPU workloads on the Threadripper 3990X, there wasn't much change to see out of the tested kernels.
PostgreSQL read performance appears to have regressed on Linux 5.5.
But other databases were looking good.
Thanks to the mighty power of the AMD Ryzen Threadripper 3990X, I'll be looking at some of these kernel regressions in more detail over the weekend.
| 2
| 1,760,719,492.12672
|
https://www.phoronix.com/news/AMD-Family-19h-Perf-Uncore
|
A Few More Linux Kernel Patches Floated This Week For AMD Family 19h (Zen 3)
|
Michael Larabel
|
Going back to the start of 2020 we've been seeing a few patches here and there around AMD Family 19h, almost certainly Zen 3. That patch work has continued with a few more bits out this week while hopefully more bring-up is on the horizon ahead of the Linux 5.7 merge window opening in just over one month's time.
Like the earlier Family 19h EDAC bring-up, this week's work isn't too juicy besides it being refreshing to see AMD punctually getting out Linux kernel patches for forthcoming hardware. The patches this week involve a few additions to AMD's perf subsystem code around the uncore bits. No enticing details of Family 19h are revealed but just shifting code around for supporting the L3 thread mask for the forthcoming CPUs and also the L3 PMU.
This is good to see the perf support bring-up but not important for end-users. With not seeing any radical patches yet for Family 19h, it gives us hope the Linux kernel is already in good shape for AMD Zen 3 CPUs coming later this year. Following the MCE issue with Threadripper 3, the RdRand issue with the initial Zen 2 CPUs, and other past launch gotchas, it would be quite refreshing to see the Linux kernel already fit and good to go for reliable Zen 3 support on launch day without any brown paper bag issues. In any case, we'll let you know when the launch comes how the Linux support stacks up but so far we're optimistic that AMD has formed a good Linux rhythm that will only be improved moving forward.
We will in particular be closely monitoring to see what other AMD Family 19h patches may be dropping over the next month with the Linux 5.7 merge window likely happening around the start of April and thus an important target for any enablement that needs to get out there early so it has time to work its way into various Linux distributions, etc.
| 37
| 1,760,719,492.679542
|
https://www.phoronix.com/news/Linux-4K-Fix-AMD-Stoney-Ridge
|
Linux Will Finally Stop Flickering With AMD Stoney Ridge On 4K Displays
|
Michael Larabel
|
For those still running the AMD "Stoney Ridge" mobile APUs from 2016 that were launched aside Bristol Ridge with Excavator-based CPU cores and GCN 1.2 graphics, the Linux kernel has a fix finally for flickering issues when driving a 4K display off the APU.
It turns out the fix for this 4K flickering issue on Stoney Ridge APUs is to disable IOMMU support for the hardware. Canonical's Kai-Heng Feng explained in the patch, "Serious screen flickering when Stoney Ridge outputs to a 4K monitor. Use identity-mapping and PCI ATS doesn't help this issue. According to Alex Deucher, IOMMU isn't enabled on Windows, so let's do the same here to avoid screen flickering on 4K monitor."
Going back to at least last year has been bug reports about screen flickering when running at 4K with Stoney Ridge hardware. Back then it was also noted when booting the kernel with iommu=off, screen flickering didn't occur.
The change is blanket disabling IOMMU for systems with Stoney Ridge graphics. It looks to be a case of Stoney Ridge's IOMMU capabilities being stoned.
| 15
| 1,760,719,493.602706
|
https://www.phoronix.com/news/AMD-EPYC-7532-EPYC-7662
|
AMD Announces EPYC 7532 + EPYC 7662 As Newest Rome Processors
|
Michael Larabel
|
AMD has expanded their 7002 series "Rome" family with the availability today of the EPYC 7662 as their latest 64-core / 128-thread offering and the EPYC 7532 as a new 32-core part but with a full 256MB cache to offer more per-core L3 cache than other 32-core processors.
The EPYC 7662 is AMD's fifth 64-core Rome processor and aims to come in at a lower cost. The EPYC 7662 64-core / 128-thread CPU has a 2.0GHz base frequency, 3.3GHz boost frequency, 256MB cache, and a 225 Watt TDP. Remaining at the top end of the Rome line-up is the EPYC 7742 64-core CPU with a 2.25GHz base clock and 3.4GHz maximum boost clock.
AMD hasn't yet confirmed the new "lower cost point" of the EPYC 7662, but at least from one Internet listing it looks like the part will be available in retail channels for about $6650 where as the EPYC 7742 is selling for about $7500 USD and the EPYC 7702 at about $7000.
The EPYC 7532 is just a 32-core / 64-thread part but making that 2.4GHz / 3.3GHz processor interesting is having the full 256MB L3 cache for benefiting cache sensitive workloads. It appears the EPYC 7532 will retail for about $3600.
More details at community.amd.com.
| 32
| 1,760,719,494.195579
|
https://www.phoronix.com/news/AMD-SEV-Live-Migration
|
New Patches: AMD Live Migration Support For VMs With Secure Encrypted Virtualization
|
Michael Larabel
|
Beyond the Linux kernel patches presented earlier this week for AMD SEV-ES "Encrypted State" support, another Linux patch series out overnight provides another improvement to Secure Encrypted Virtualization with AMD EPYC server processors.
The newest open-source SEV work to report on this week is live migration support when making use of AMD Secure Encrypted Virtualization. Currently VMs can't be live migrated when making use of this hardware-backed encryption support of virtual machines, but a new patch series enables QEMU/KVM live migration to now work in the presence of SEV.
A set of 12 patches adding more than one thousand lines of new kernel code put the necessary bits into place for handling live migration under SEV. The patches basically allow for passing an encryption context with the hypervisor as well as the commands for copying into a new SEV guest memory space. While the initial work is focused on the Kernel-based Virtual Machine (KVM), this SEV live migration could be extended to other hypervisors as well.
Longtime upstream Linux kernel developer Andy Lutomirski so far has commented on the code and raised some concerns over the increasing complexity of AMD SEV within the Linux kernel, "To be blunt: if I had noticed how the SEV code worked before it was merged, I would have NAKed it. It's too late now to retroactively remove it from the kernel, but perhaps we could try not to pile more complexity on top of the unfortunate foundation we have."
We'll see if this SEV code gets cleaned up in time for the Linux 5.7 cycle this spring.
| 1
| 1,760,719,495.074034
|
https://www.phoronix.com/news/AMD-Sensor-Fusion-Hub-V3
|
AMD Volleys Third Revision To Sensor Fusion Hub Linux Driver
|
Michael Larabel
|
One of the letdowns of the forthcoming Linux 5.6 kernel is that AMD's long-awaited Sensor Fusion Hub "SFH" open-source driver wasn't ready in time for merging this cycle, but it continues moving forward with hopes of seeing it in Linux 5.7.
The AMD Sensor Fusion Hub driver for supporting the accelerometer and gyroscopic sensors along with other possible sensors on modern AMD Ryzen laptops was published in early January and then revised again a few weeks later. Now after going through more code review, the third version of the AMD SFH Linux driver is available for review and testing.
Today's newest code drop of amd-sfh-hid by AMD's Sandeep Singh has more code fixes and splitting out of DebugFS support into a separate patch. This driver is important for improving the Linux experience on higher-end AMD Ryzen laptops.
Those wanting to test the AMD SFH driver ahead of it being mainlined in the months ahead can find the v3 patches here with the nearly twenty five hundred lines of new kernel code.
| 0
| 1,760,719,495.710762
|
https://www.phoronix.com/news/Noctua-NH-U9S-AMD-3950X
|
Noctua NH-U9S Performance For The AMD Ryzen 9 3950X + Ondemand vs. Performance Governors
|
Michael Larabel
|
For those that may be looking to run an air-cooled AMD Ryzen 9 3950X especially in a rack-mount 4U chassis, here are some recent results I did from some testing using a Noctua NH-U9S with two 92mm fan configuration. Additionally, these results contain performance metrics from both CPUFreq Ondemand vs. Performance governors as an additional point of interest.
These results are for reference purposes of the Noctua NH-U9S in a dual fan setup for this 16-core / 32-thread 3.5GHz (4.7GHz boost) CPU rated with a 105 Watt TDP.
The NH-U9S was tested as one of our go-to heatsinks given that it fits within 4U height requirements that we are often running within for our many rackmount benchmarking systems.
The Ryzen 9 3950X testing was done on Ubuntu 20.04 daily with its default Linux 5.4 kernel at the time. Tests were done out-of-the-box with the CPUFreq Ondemand governor and then when switching over to CPUFreq Performance governor.
On the thermal front, ondemand vs. performance governors for the Ryzen 9 3950X didn't yield much of a difference, similar to what we have seen in the past -- contrary to popular belief that keeping "performance" on will lead to much higher operating temperatures normally isn't the case for at least desktops. Under hours of load, the average temperature was 62 degrees with a peak of 84 degrees. The idle temperature was around 32 degrees.
As for the performance between governors, librsvg and dav1d had the most profound benefit from the performance governor while GIMP, vpxenc, DeepSpeech, and others saw less of a benefit. For the web browsing performance, using the CPUFreq performance governor only led to ~3% better performance. In numerous workloads, there isn't a measurable difference between ondemand and performance governors. Those wanting to dig through all these cooling and governor performance tests conducted on the Ryzen 9 3950X, you can find all the data via this OpenBenchmarking.org result file.
| 17
| 1,760,719,501.282877
|
https://www.phoronix.com/news/AMD-SEV-ES-Linux-2020-Patches
|
The Linux Kernel Now Seeing Patches For AMD SEV-ES "Encrypted State" Support
|
Michael Larabel
|
While since 2016~2017 AMD has been posting Linux kernel patches for Secure Memory Encryption (SME) and Secure Encrypted Virtualization, coming out this morning is finally the first public patch series wiring up the Linux kernel for SEV-ES as further enhancing virtualization encryption.
On top of the Secure Encrypted Virtualization support that's been plumbed into the mainline Linux kernel and related components for a while now, AMD and SUSE developers have sent out the patches today for SEV "Encrypted State" support.
AMD SEV-ES allows for protecting the guest register state from the hypervisor. CPU register state is encrypted that cannot be accessed or modified by the system hypervisor. The intent of SEV-ES is to help fend off control-flow attacks by modifying the VM state, unauthorized reading of the virtual machine state, and other similar attacks. SEV-ES does allow for selectively sharing certain information to the hypervisor about certain switches where needed.
SEV-ES is supported with EPYC CPUs but has taken some time to get the Linux kernel support plumbed. There are also patches still coming out for adding the AMD SEV-ES support to the Kernel-based Virtual Machine (KVM) code. Other hypervisors also need to be updated to handle SEV Encrypted State functionality as well.
There are 62 patches out today and go more into the technical details of SEV-ES. The intent now is for getting more public code review of the changes while there are some known issues with the current code that could lead to crashes and other shortcomings still being addressed.
Further technical information on SEV-ES can also be found via the APM. On a higher level, there is this late 2017 presentation by AMD's Thomas Lendacky that covers AMD SEV-ES for those interested in more information on this long work-in-progress feature for Linux hypervisors.
Given there still is a lot of work ahead and code review of these dozens of patches, it's not clear if the work will be settled in time for the Linux 5.7 cycle this spring otherwise will likely end up in a kernel release later in 2020.
| 10
| 1,760,719,501.715819
|
https://www.phoronix.com/news/2004-CPU-3990X-Plus-FX-9590
|
CPUs From 2004 Against AMD's New 64-Core Threadripper 3990X + Tests Against FX-9590
|
Michael Larabel
|
With having the initial AMD Ryzen Threadripper 3990X Linux benchmarks out of the way, I did some fun Friday night benchmarking with this 64-core HEDT CPU... Seeing how the 3990X performance compares to some CPUs used when starting out Phoronix back in 2004 as well as also for fun seeing how the Threadripper 3990X compares to the notorious AMD FX-9590.
The CPUs tested for comparison to what was trending when starting out Phoronix in 2004 were an Intel Pentium 4 2.80GHz processor, a 2.8GHz processor with Hyper Threading for offering up two threads. Then on the low-end a single-core with no HT Intel Celeron 2.4GHz processor. Both of those CPUs were running on the legendary Abit IC7-MAX3 with 1GB of dual channel memory, 160GB Seagate HDD, and ECS Radeon 9200 graphics card. That's the oldest historical data I still have from when Phoronix started out and in a manner via the Phoronix Test Suite testing and version-locked test profiles where I can still carry out the test in a repeatable and standardized manner.
So solely for some Friday night benchmarking fun in between other more serious Threadripper 3990X benchmarks, here is how these extremes compare:
With the Intel Pentium 4 2.8GHz CPU from ~2003 it took 347 seconds to run the C-Ray multi-threaded ray-tracing benchmark or 520 seconds for the Intel Celeron that is approaching two decades old. But the AMD Ryzen Threadripper 3990X can run the same exact test in the same configuration in just over one second.
Even on the single-threaded test front, there's a significant evolution from 2004 era hardware to 2020. The single-threaded FLAC audio encoding test on the old hardware took 45~53 seconds while the Threadripper 3990X could run the same exact test in 4.5 seconds.
That was fun, but for running some other tests a bit more relevant, I pulled up some historical OpenBenchmarking.org data from the AMD FX-9590 for comparing the Threadripper 3990X to that. The AMD FX-9590 as a reminder is the eight-core processor with 4.7GHz base frequency but could boost up to 5.0GHz and was known for running hot with its 220 Watt TDP. While such a TDP is no longer unheard of these days, the performance in going from an AMD FX-9590 to Threadripper 3990X is astounding.
This very parallel benchmark saw the Threadripper 3990X running 21x faster than the FX-9590 that was once at the top-end of AMD's portfolio about six years ago.
Code compilation with the Linux kernel is about four times faster due to I/O and other factors at play.
The C-Ray test is another multi-threaded case with about a 21x difference.
Or even for single-threaded MP3 encoding, the Threadripper 3990X with its Zen 2 CPU cores took half the time or so as the pre-Ryzen FX-9590.
Anyhow, hopefully you enjoyed these numbers for some casual weekend pondering, but back onto more interesting AMD Ryzen Threadripper 3990X benchmarking under Linux, BSDs, and Windows.
| 59
| 1,760,719,503.205597
|
https://www.phoronix.com/news/AMD-Adding-To-Microsoft-ONNX
|
AMD Contributing MIGraphX/ROCm Back-End To Microsoft's ONNX Runtime For Machine Learning
|
Michael Larabel
|
AMD is adding a MIGraphX/ROCm back-end to Microsoft's ONNX run-time for machine learning inferencing to allow for Radeon GPU acceleration.
Microsoft's open-source ONNX Runtime as a cross-platform, high performance scoring engine for machine learning models is finally seeing AMD GPU support. This project has long supported NVIDIA TensorRT and CUDA along with Intel's APIs around DNNL / nGraph / OpenVINO, and even the ARM Compute Library. But now this Linux / Windows / macOS machine learning run-time will be able to support Radeon Open eCosystem (ROCm) for Radeon GPU acceleration on Linux.
The ONNX Runtime code from AMD is specifically targeting ROCm's MIGraphX graph optimization engine.
This AMD ROCm/MIGraphX back-end for ONNX is being reviewed here. Hopefully it will be released soon along with a new Radeon Open eCosystem release that formally adds in MIGraphX.
Those wanting to learn about the Microsoft ONNX Runtime itself can visit the GitHub project site.
| 6
| 1,760,719,503.272935
|
https://www.phoronix.com/news/Linux-5.6-Expanded-AMD-APICv
|
Linux 5.6 KVM Expands AMD APIC Virtualization Support With Dynamic APICv
|
Michael Larabel
|
A second round of KVM virtualization updates were sent out today for the Linux 5.6 merge window that is still open through the weekend.
On the AMD side, this round of Linux KVM virtualization updates has expanded support for APIC virtualization. APICv/AVIC is now supported with AMD SVM in configurations previously not possible. AMD's Suravee Suthikulpanit previously explained:
The 'commit 67034bb9dd5e ("KVM: SVM: Add irqchip_split() checks before enabling AVIC")' was introduced to fix miscellaneous boot-hang issues when enable AVIC. This is mainly due to AVIC hardware doest not #vmexit on write to LAPIC EOI register resulting in-kernel PIC and IOAPIC to wait and do not inject new interrupts (e.g. PIT, RTC).
This limits AVIC to only work with kernel_irqchip=split mode, which is not currently enabled by default, and also required user-space to support split irqchip model, which might not be the case.
The goal of this series is to enable AVIC to work in both irqchip modes, by allowing AVIC to be deactivated temporarily during runtime, and fallback to legacy interrupt injection mode (w/ vINTR and interrupt windows) when needed, and then re-enabled subsequently (a.k.a Dynamic APICv). Advanced Virtual Interrupt Controller is one of the means of reducing overhead in virtualized environments and has been supported for years already in AMD CPUs (since Carrizo) and the Linux kernel but with some exceptions as noted. This dynamic APICv/AVIC activation will be part of Linux 5.6 that in turn should be released as the stable kernel around APril.
The KVM pull request also includes various other x86 KVM fixes, several s390 fixes, and other fixing.
| 2
| 1,760,719,504.886906
|
https://www.phoronix.com/news/AMD-Coreboot-In-Early-2020
|
The Current State of AMD Zen Coreboot Support: Basically Limited To Chromebooks
|
Michael Larabel
|
Firmware developer Michał Żygowski of embedded consulting firm 3mdeb has provided a convenient overview over the current state of AMD Coreboot support for booting with this open-source alternative to conventional proprietary BIOS.
As most Phoronix readers should know, at the moment the Coreboot support for AMD platforms comes down to either several generations old hardware or if talking about Zen-era hardware is limited to Google Chromebooks. The only support AMD has been providing for Coreboot in recent times revolves around their APUs beginning to surface within Chromebooks and Google mandating Coreboot support, among their other engineering requirements.
This current Chromebook-focused support is far different from when nearly a decade ago they would offer open-source AGESA support and ended up allowing for decent Coreboot open-source support on various desktop and server motherboards. Michael characterized the current Coreboot Zen possibilities as, "family 17h support (Ryzen/Zen) is rather unlikely for other products thanChromebooks in coreboot."
His slide deck can be found here from FOSDEM 2020 along with the WebM/VP9 session recording from Brussels.
Not talked about was the possibility of AMD going back to open-source AGESA / Coreboot support as I mentioned AMD was exploring last August around the time of the EPYC "Rome" launch. I haven't, however, heard any update from AMD since then if that is still being pursued or its current state/plans. We certainly hope AMD having better Coreboot support is on their road-map for 2020 -- after all, AMD had been hiring Coreboot developers in late 2019 but it's not known yet if the focus was just for AMD Chromebooks or not.
| 8
| 1,760,719,505.122787
|
https://www.phoronix.com/news/AMD-Zen-2-Scheduler-LLVM-Fixes
|
AMD's Zen 2 Scheduler Model Gets Partially Fixed Up In LLVM
|
Michael Larabel
|
Landing in the LLVM compiler infrastructure code-base in January was finally an AMD Zen 2 scheduler model optimized for the latest-generation AMD processors when compiling code with Clang using the -march=znver2 targeting. However, now some important fixes to this scheduler model have landed.
It turns out there were some misunderstandings / typos in the Znver2 scheduler model that was merged in January. A commit last week to LLVM has fixed the "obvious mistakes" to the AMD Zen 2 model including for latencies but other work remains to be addressed.
The AMD Zen 2 scheduler model defects may at least partially explain some of the LLVM Clang 10 performance differences we saw last month when testing the original scheduler model addition.
| 0
| 1,760,719,506.372922
|
https://www.phoronix.com/news/AMD-SFH-Linux-V2
|
AMD Sensor Fusion Hub Driver Revved But Not On Tap For Linux 5.6
|
Michael Larabel
|
Earlier this month AMD finally published their Sensor Fusion Hub driver for Linux to improve the Ryzen laptop support. That new "SFH" driver hasn't been queued as part of any Linux 5.6 pull request but a second version of the driver did make it out this week.
The AMD Sensor Fusion Hub support has been long awaited and is needed for supporting the accelerometer/gyroscopic sensors on Ryzen laptops among other functionality. There have been requests for supporting the Sensor Fusion Hub on Linux going back to 2018.
With the AMD-SFH-HID version two, various issues raised during the first round of review were addressed. Some warnings were also cleared up from build testing and the Smatch static analyzer. The v2 patches for testing can be found here.
Given the timing of the Sensor Fusion Hub driver publishing, it's an uphill battle if it's to get into Linux 5.6 but more than likely will be ready for Linux 5.7. We're approaching the midway point of the Linux 5.6 merge window and while it could be a late addition for 5.6 since AMD-SFH-HID is for supporting new hardware with low risk to breaking anything for existing users, kernel maintainers often opt to diver new drivers to the next cycle... But we'll see how it plays out soon enough, at least this driver is finally out there.
| 1
| 1,760,719,506.551334
|
https://www.phoronix.com/news/Raven-Fix-For-Clicking-Noises
|
AMD Prepares Fix To Address Clicking Issue With Audio Playback On Raven APUs
|
Michael Larabel
|
Unfortunately it wasn't a trouble-free experience at launch but with time Raven Ridge APUs have been getting cleaned up on Linux for a pleasant experience, thanks in part to the Google Chromebook play that has also seen these newer AMD APUs seeing HDCP content protection support and PSP / TEE trusted execution functionality.
The latest overdue improvement on the AMD Raven APU front is a fix for a pesky issue during audio playback. If playing audio streams immediately one after another, clicking noises can be heard. That is in the process of being resolved thanks to a new kernel patch.
This patch from an AMD engineer just shifts around a few lines of code in the AMD ACP3x Audio Co-Processor 3.x code. But that small patch is enough to address the clicking noise observed up to now if playing audio streams back-to-back.
The code hasn't yet been merged to mainline but hopefully will soon be and back-ported to the stable releases.
| 1
| 1,760,719,508.014826
|
https://www.phoronix.com/news/AMD-Znver2-Clang-10-Tests
|
AMD Zen 2 "Znver2" Optimizations With LLVM Clang 10 Bring Some Improvements
|
Michael Larabel
|
With LLVM Clang 10 having added a Zen 2 scheduler model tuned for the latest AMD CPUs over the existing "znver2" tuning that had just copied the Zen 1 scheduler, here are some benchmarks looking at the LLVM Clang 9 vs. 10 compiler performance on AMD EPYC when making use of "-march=znver2" optimizations.
On the AMD EPYC 7742 2P server running Ubuntu 19.10 with the Linux 5.5 kernel, I carried out benchmarks earlier this month comparing the LLVM Clang 9.0.1 performance to that of LLVM Clang 10.0 after the Zen 2 (znver2) improvements landed and around the time of the LLVM 10.0 branching.
Both Clang 9 and Clang 10 Git were built the same in their release modes. As usual with compiler optimizations/tuning when looking at the performance across dozens of workloads, the results are mixed:
GraphicsMagick and PostgreSQL saw some big wins when using LLVM Clang 10 and also some smaller improvements in some of the video encode and compression tests. To no surprise, Clang 9 was building faster than Clang 10 given that newer compilers add more optimization passes and tuning thus taking longer to compile code in an effort to produce faster binaries. If ignoring the timed compilation results and then the multiple libgav1 tests, Clang 10 overall is looking good with just a few losses.
Out of 73 C/C++ benchmarks tested between these Clang compiler builds, Clang 10 did lead nearly 60% of the time.
If taking the geometric mean of all those benchmark results, Clang 10.0 Git and Clang 9 were neck-and-neck with this AMD Zen 2 targeting.
See all of these benchmarks in full via this OpenBenchmarking.org result file.
| 6
| 1,760,719,508.038367
|
https://www.phoronix.com/news/AMD-Family-19h-EDAC-PR
|
The Initial AMD Family 19h Support Sent In For Linux 5.6 EDAC Driver
|
Michael Larabel
|
SUSE's Borislav Petkov sent in the (Reliability, Availability and Serviceability) updates for the Linux 5.6 kernel on this first day of the new merge window.
Notable to the RAS pull is the AMD Family 19h support in the AMD EDAC driver. Up to now Zen / Zen+ / Zen 2 has been Family 17h CPUs but for Zen 3 and beyond it's looking like it will be treated as Family 19h. With Linux 5.6 this early enablement is underway beginning with their memory EDAC (Error Detection And Correction) driver code.
This initial AMD F19h support is mostly adding in new PCI IDs and other changes for Family 19h 00h through 0Fh support. No juicy hardware details are revealed as part of these early patches. There isn't much more on the Linux 5.6 front in regards to Family 19h but we anticipate that to change this summer with Linux 5.7... Stay tuned as always to Phoronix for bleeding edge kernel development news.
The RAS pull also has various other MCE fixes and code clean-ups throughout but nothing else as exciting as seeing a new family of CPUs brought up.
| 0
| 1,760,719,509.512564
|
https://www.phoronix.com/news/Renoir-Graphics-Not-Experiment
|
AMD Ryzen 4000 Mobile Series "Renoir" Graphics No Longer Experimental With Linux 5.5
|
Michael Larabel
|
While the Linux 5.5 kernel is expected to be released as soon as this Sunday, a last minute change to the AMDGPU DRM driver makes the Renoir graphics no longer treated as experimental. With that, there is open-source support out-of-the-box rather than being hidden behind a kernel module flag.
AMD has been working on the Renoir support for Linux going back to the end of last summer. Renoir was sent in for the Linux 5.4 kernel but initially treated as "experimental" support while now at the end of the Linux 5.5 cycle it's no longer treated as experimental.
This is good due to the first batch of Renoir parts being announced back at CES earlier this month. The Ryzen 4000 mobile series is Renoir from the low-end Ryzen 4300U through the top-end Ryzen 4800H and 4800U models. Renoir is based on the Zen 2 CPU cores with a revised Vega GPU.
As Renoir is Vega-based, the enablement from the graphics driver side wasn't too big and at least is now promoted out of experimental so those running the upcoming Linux 5.5 kernel (which should be on the likes of Ubuntu 20.04 LTS) will find the support enabled by default. When AMD Radeon graphics are behind the experimental flag, it requires booting the kernel with amdgpu.exp_hw_support=1 to enable the non-production-ready support. So this kernel should be out and adopted before seeing more Renoir laptops hitting the market.
The promotion was the only change as part of the amd-drm-fixes sent in today for 5.5.
| 25
| 1,760,719,509.624256
|
https://www.phoronix.com/news/Linux-k10temp-debugfs
|
Linux k10temp Driver For AMD CPUs Updated To Better Handle Power/Temp Analysis
|
Michael Larabel
|
As we have been eagerly talking about for the past week, the Linux kernel's k10temp driver was updated for better AMD CPU CCD temperatures and voltage/current reporting. Those improvements have been quickly evolving thanks to the work of the open-source community with AMD still sadly holding the datasheets concerning the power/temperature registers close to their vest. A new version of k10temp was sent out on Wednesday.
As reported earlier this week, these k10temp improvements could land for the upcoming Linux 5.6 but additional testing is needed. While Zen 2 CPUs have been shipping for months, these k10temp improvements are only coming now thanks to HWMON maintainer Guenter Roeck who continues working on this driver in cooperation with the community as AMD currently isn't releasing documentation/datasheets concerning the power/thermal registers or any reference code for that matter... Many Linux desktop users dream of seeing something someday like AMD Ryzen Master coming to Linux.
With the v4 driver improvements it normalizes current calculations but still depends upon some potential per-board scaling to be set in the configuration file of LM-Sensors, makes a clarification, and adds DebugFS support for better debugging/analysis on different CPUs.
Due to the lack of official AMD information to assist in figuring out the correct power/thermal reporting information, the DebugFS data is designed to assist in reading the raw information. Dumped via DebugFS is now the raw thermal and SVI register dumps. The focus is on helping to discover additional sensors on various CPUs and/or for helping to fix problems for buggy information on CPUs. Ideally this shouldn't be needed but due to k10temp still being community-driven without much support from AMD, this DebugFS support should be a big help until AMD steps up to the plate.
If all goes well, these k10temp improvements will make it into Linux 5.6.
| 20
| 1,760,719,510.596403
|
https://www.phoronix.com/news/K10temp-Could-Improve-5.6
|
AMD Zen Thermal/Power Reporting Improvements Could Hit Linux 5.6 But More Testing Needed
|
Michael Larabel
|
Last week I eagerly reported on Ryzen CPUs on Linux finally seeing CCD temperatures and current/voltage reporting thanks to new patches to the k10temp driver by Google's Guenter Roeck who oversees the kernel's hardware monitoring "HWMON" subsystem. The patches seem to be working well and are tentatively queued in hwmon-next, but more testing is still needed.
Those k10temp patches were quickly revised and currently at version two. They seem to be working well in the extra thermal and power reporting where capable for Zen/Zen+/Zen2 CPUs.
Excitingly, on Sunday I noticed the AMD thermal/power driver improvements did land in hwmon-next normally meaning they are queued and ready for the upcoming kernel merge window, Linux 5.6
However, in communicating with Guenter in my testing results of the patches, he indicated he still would like to see more tests/confirmation that these driver improvements are working fine. He wants to get as much feedback as possible and then make up his mind whether to submit for Linux 5.6.
I have tested the k10temp patches on a number of AMD Zen/Zen+/Zen2 systems so far and have not run into any problems. Those wanting to test these patches and share their feedback can find the current v2 patches here as well as the contact information for the mailing lists to volley your data (or if you can't be bothered to do that, post your sensors output in the forums as a comment and I can relay the data as well).
To hopefully motivate more to test these patches, here is my Ubuntu/Debian kernel build of Linux 5.5 patched with the v2 k10temp patches, if you were holding off on testing to avoid building your own kernel. Of course use at your own risk but so far these driver patches have been working great.
Here's to hoping that these long-awaited AMD thermal/power CPU reporting improvements can make it into Linux 5.6!
| 23
| 1,760,719,511.551669
|
https://www.phoronix.com/news/AMD-Zen-3-Patches-Start-5.6
|
AMD Zen 3 "Family 19h" Enablement Beginning With The Linux 5.6 Kernel
|
Michael Larabel
|
With the upcoming Linux 5.6 kernel cycle will be the first of many patches to come surrounding AMD Zen 3 "Family 19h" support.
So far there haven't been any AMD Family 19h patches to the Linux kernel besides k10temp driver support. But queued up ahead of the weekend were a couple changes relating to Zen3/19h beginning to collect in ras/core for the Linux 5.6 merge window kicking off in the next week or two.
That includes adding the Family 19h PCI IDs (0x1654 and 0x1653) for the amd_nb driver.
And within the AMD64 EDAC driver for error detection and correction it's extended the Family 17h code
There are also a few other patches for future-proofing the AMD EDAC/MCE Linux kernel code.
That's it for what we have seen so far. For those reporting this weekend and saying AMD released Zen 3 microcode in the Linux kernel, that is not the case.
As the AMD Zen 3 / Family 19h bring-up continues for Linux, we'll keep you up-to-date on the happenings.
| 20
| 1,760,719,512.033041
|
https://www.phoronix.com/news/AMD-k10temp-Better-Zen-V2
|
The AMD Ryzen Thermal / Power Linux Reporting Improvements Working Well - V2 Up For Testing
|
Michael Larabel
|
A few days ago I reported on AMD's "k10temp" Linux kernel driver finally seeing the ability to report CCD temperatures and CPU current/voltage readings as a big improvement to this hardware monitoring driver. The work hasn't yet been queued for inclusion into the mainline kernel, but initial testing is working well and a second revision to the patches has been sent out.
Linux HWMON maintainer Guenter Roeck who spearheaded this work independent of AMD sent out the "v2" k10temp driver improvements on Saturday. This allows core complex tie temperature reporting for Zen 2 CPUs and allows current and voltage reporting for Ryzen CPUs. While this information has long been available to Windows users, sadly it's not been the case for Linux at least as far as mainline drivers go -- the out-of-tree Zenpower driver and other third-party attempts have been available but nothing mainline.
With the revised patches there are various fixes and code improvements. Those patches are up for testing here.
Roeck is still looking for more testing by AMD Ryzen Linux owners, so if you try out the work be sure to chime in with the results. Ideally these k10temp improvements could be queued for inclusion in the upcoming Linux 5.6 kernel if there is enough testing completed otherwise would have to wait until a kernel release later in the year before merging.
| 25
| 1,760,719,513.461524
|
https://www.phoronix.com/news/ASUS-TUF-Ryzen-Thermal-Fix
|
ASUS TUF Laptops With Ryzen Are Now Patched To Stop Overheating On Linux
|
Michael Larabel
|
The AMD Ryzen Linux laptop experience continues improving albeit quite tardy on some elements of the support. In addition to the AMD Sensor Fusion Hub driver finally being released and current/voltage reporting for Zen CPUs on Linux, another step forward in Ryzen mobile support is a fix for ASUS TUF laptops with these processors.
The ASUS TUF laptops with Ryzen CPUs like the FX705DY, FX505DY, and GA502DU turned out to have quite a show-stopping problem on Linux. At boot its ACPI implementation would put the laptop cooling in a "silent mode" causing the CPU and GPU cooling to operate at a minimum. But any game or CPU intensive workload would end up heating up the CPU to the point of throttling around 399MHz and being stuck that way until next reboot.
An independent Linux developer has now had his patches accepted for addressing this issue by exposing the ACPI thermal throttle policy for this ASUS TUF laptops. By default now rather than being limited to the "silent" mode, the default thermal throttling policy is enabled at boot where the cooling fans will operate as intended and similar to the Windows defaults. This in turn should avoid the ASUS TUF laptops with Ryzen CPUs from thermal throttling so easily.
This new support to the asus_wmi Linux driver also now exposes the "overboost" capability of the cooling fans where they will ramp up quicker to help avoid throttling in heavy workloads. With the patches that will land for Linux 5.6, the ASUS thermal policy can be configured now via /sys/devices/platform/asus-nb-wmi/throttle_thermal_policy.
Those wanting to patch the driver themselves to avoid having to wait for the next kernel cycle can find the asus_wmi patches here and here. These patches are already residing with linux-platform-drivers-x86.git until the forthcoming Linux 5.6 merge window.
This problem for ASUS laptops with Ryzen CPUs has been known since at least last summer with bug reports where numerous users reported this behavior and some resorting to disabling the CPU's boost capability outright in an attempt to avoid throttling.
| 46
| 1,760,719,513.572271
|
https://www.phoronix.com/news/AMD-Ryzen-k10temp-CCD-V-Current
|
Ryzen CPUs On Linux Finally See CCD Temperatures, Current + Voltage Reporting
|
Michael Larabel
|
One of the few frustrations with the AMD Ryzen CPU support on Linux to date has been besides the often delayed support for CPU temperature reporting has been the mainline kernel not supporting voltage readings and other extra sensors. But that is finally changing with the "k10temp" driver being extended to include current and voltage reporting plus CCD temperature reporting on Zen 2 processors.
There has been the out-of-tree Zenpower driver and other efforts to provide this information on Linux but hasn't been officially backed by AMD and not mainlined in the kernel, thus greatly reducing the exposure to potential users. But now the k10temp driver is finally being extended to include these extra information outputs.
Linux HWMON maintainer Guenter Roeck has been working on these driver improvements to k10temp. Besides some code improvements, the new patches support reporting Core Complex Die (CCD) temperatures on Zen 2 processors. Additionally, for Ryzen CPUs (Zen 1 included) are core/SoC current and voltage information.
With this for current Ryzen 3000 series processors the patched k10temp Linux driver should expose Vcore, VSoc, Tdie, Tctl, Tccd1, Tccd2, Icore, and Isoc outputs.
Guenter posted these patches to the kernel mailing list. He's looking for more testing on these k10temp improvements before he will queue them up for mainline kernel inclusion... Thus if you are hoping to see this work for the upcoming Linux 5.6 cycle, you better start doing some testing on AMD CPUs ASAP and pass along your findings (and, yes, I'll be joining in on this testing party). Regardless of whether it happens for Linux 5.6 or takes another cycle, at least these driver improvements are now happening and hopefully moving forward AMD engineers will be able to make more proactive contributions.
| 42
| 1,760,719,515.017743
|
https://www.phoronix.com/news/LLVM-10-AMD-Znver2-Sched-Model
|
LLVM/Clang 10.0 Adds AMD Zen 2 Scheduler Model For Optimized Code Generation
|
Michael Larabel
|
It's too bad that it has taken so many months after AMD Zen 2 based Ryzen and EPYC processors began shipping to see this compiler support in place, but the good news now is that for the upcoming release of LLVM 10.0 is now the Zen 2 scheduler model being added to the "znver2" target.
Going back to before the Zen 2 processors began shipping last summer, in February AMD Znver2 support was added for LLVM Clang 9.0. But like the GCC compiler support at the time, it added new instructions supported by these CPUs but didn't update the scheduler model / cost tables. In July AMD-partner SUSE added Znver2 tuning to GCC including a new scheduler model that was wired up for GCC 10 and back-ported to GCC 9.2.
Now a half-year later and just days ahead of the LLVM 10.0 feature freeze, mainline LLVM 10.0 is seeing the Zen 2 scheduler model added. AMD's Ganesh Gopalasubramanian provided the Znver2/Rome scheduler and it was committed last week. "The patch gives out the details of the znver2 scheduler model. There are few improvements with respect to execution units, latencies and throughput when compared with znver1."
It's a pity that it has taken so long after Zen 2 CPUs began shipping to see this compiler support through, but at least it's here now. LLVM/Clang 10.0 should be released as stable next month while GCC 10 will be releasing in the March~April timeframe once it clears out its high priority bugs. LLVM 10.0 will hopefully make it in time for inclusion in the upcoming Ubuntu 20.04 LTS.
Moving forward hopefully AMD will be more punctual with their Zen 3 "znver3" compiler tuning, which we have yet to see for GCC or LLVM. The compiler targets and software support in general for Linux/open-source has been one of the areas where Intel has excelled very well over the years thanks to their Open-Source Technology Center and sizable open-source development team while AMD has generally been belated in elements of their open-source support. With some Intel generations we've seen their GCC + LLVM support ironed out a year or more in advance, even before their manufacturing rut has led to delays in new microarchitecture launches. With AMD continuing to gain marketshare both on desktops and servers, hopefully we'll see improvements from AMD in this area in 2020 and beyond.
I'll have up some fresh LLVM Clang -march=znver2 comparison benchmarks with EPYC up shortly.
| 5
| 1,760,719,515.299821
|
https://www.phoronix.com/news/AMD-Linux-5.6-DP-MST-DSC
|
AMD Has DP MST DSC Support Ready For The Linux 5.6 Kernel
|
Michael Larabel
|
In addition to their AMDGPU/AMDKFD feature updates sent out on Thursday, AMD also sent in a special pull request to DRM-Next on its own of a new feature: DP MST DSC.
DP MST DSC is DisplayPort Multi-Stream Transport Display Stream Compression. AMD and the other common Linux display drivers have long supported DP MST for driving multiple displays off a single DisplayPort cable thanks to this specification. DP MST has been around since DisplayPort 1.2 and allows multiple independent displays to be served off a single DP port thanks to multiplexing the streams. MST also supports daisy-chaining and other convenient features. What's new now is supporting DSC (Display Stream Compression) for MST.
DSC itself is the VESA-backed low-latency compression algorithm for serving higher resolution content on reduced bandwidth. DSC is important for 8K and above outputs in particular or 4K@120Hz. DSC with MST becomes all the more important due to sharing a single cable for multiple displays, thus opening up setups to higher resolutions / higher refresh rates.
MST hubs and other devices do need to support DSC for this to work out, besides this new kernel support.
This pull request plumbs DSC into the core Direct Rendering Manager subsystem and ultimately exposes it for operation with the AMDGPU DRM driver.
| 18
| 1,760,719,516.614688
|
https://www.phoronix.com/news/AMD-Sensor-Fusion-Hub-Linux
|
AMD Finally Publishes Sensor Fusion Hub Driver For Linux
|
Michael Larabel
|
One of the missing features for those with AMD Ryzen laptops has been the lack of a Sensor Fusion Hub driver that is needed for supporting the accelerometer and gyroscopic sensors for the display and related laptop sensor functionality. This week AMD finally posted patches for a Sensor Fusion Hub Linux driver.
Going back to 2018 have been requests for Sensor Fusion Hub support on Linux and AMD confirmed last year they were working on a driver they hoped to have ready during summer 2019. This week we were finally greeted by this driver.
AMD's Sandeep Singh posted the new amd-sfh-hid driver for supporting the Sensor Fusion Hub HID-based driver. This driver communicates with the MP2 processor, which was wired up back in Linux 5.2. The AMD Sensor Fusion Hub is akin to the Intel Sensor Hub that has been supported since it first premiered within Intel mobile devices.
The AMD Sensor Fusion Hub Linux driver patches for now are on linux-iio. If the patches are reviewed quickly enough, there's still time for hopefully seeing the AMD SFH driver premiere in Linux 5.6. This driver amounts to just under 2,500 lines of code.
| 10
| 1,760,719,516.646586
|
https://www.phoronix.com/news/AMD-Pollock-Linux-Driver
|
AMDGPU Linux Driver Adding Support For The AMD Pollock
|
Michael Larabel
|
AMD "Pollock" is a new chip similar to Dali and looking like it may be used for some Ryzen embedded purposes. AMD Pollock was plumbed into the Linux driver yesterday and was the first time we've heard this codename.
AMD sent out a patch for the AMDGPU DRM driver adding support for Pollock. The Pollock enablement for this Linux kernel graphics driver primarily comes down to adding the new IDs and updating some conditional statements to basically take the same driver code-paths as Dali.
The Pollock IDs added are 0x94, 0x95, 0xE9, 0xEA, and 0xEB. So at least from the graphics side it's quite close to Dali and other "Raven 2" parts.
This AMD Pollock enablement will presumably make it into the upcoming Linux 5.6 cycle. Given the rest of the AMD landscape at this point and being close to Dali, it's looking like Pollock may be for some embedded SoC purposes, but we'll see what else turns up in Linux patch form over the weeks ahead.
| 10
| 1,760,719,518.060529
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.