1,600 AMD open-source and Linux related news articles on Phoronix since 2006.
With the newest AMD Ryzen 7040 series laptops there is "Ryzen AI" as a dedicated AI engine based on Xilinx IP to help accelerate machine learning with the likes of PyTorch and TensorFlow. Sadly though this Ryzen AI with their new Zen 4 laptops is only supported under Microsoft Windows at this point. But it could change with sufficient customer interest.
It was just over one week ago that AMD announced plans to acquire Nod.ai to enhance their open-source AI software capabilities. Today already they announced that acquisition has completed.
AMD today published a new CPU microcode revision for Family 19h processors, which include Zen 3 / Zen 3+ / Zen 4 processors. As is sadly standard, no change-log is provided.
AMD today posted a set of interesting patches for enabling better integration of their AMDKFD (Kernel Fusion Driver, what is their compute kernel driver) memory management with Linux's DRM GEM ioctl API. In turn the code allows managing virtual address (VA) mappings in compute VMs with the GEM_VA ioctl interface for greater control of buffers imported via DMA-BUF.
Over the past two weeks there has been a pleasant uptick in new commits to Coreboot as part of enabling EPYC 9004 "Genoa" series platform support.
AMD has published the Advanced Media Acceleration (AMA) v1.0 SDK as a software stack aimed for hardware video acceleration and capable of high density real-time transcoding. The AMD AMA SDK integrates with the likes of the open-source FFmpeg and GStreamer projects for fitting nicely in the Linux software ecosystem as well as sporting its own C-based API.
Merged today to Linux 6.6 Git is a new patch for fixing an AMD erratum CPU bug affecting Zen 4 based processors like the Ryzen 7000 series and EPYC 8004/9004 series.
It was just a few weeks ago that AMD acquired AI software company Mipsology to help their AI software efforts on FPGAs. Today AMD announced another notable AI software acquisition: open-source AI software vendor Nod.ai.
While the AMD P-State driver is working quite well for Ryzen systems already with the default on Linux 6.5, one of the additions we are still waiting to land is the AMD "Preferred Core" functionality. An eighth version of those patches were posted on Monday for inching this feature closer to the mainline kernel.
The upcoming Linux 6.7 merge window is set to include a new AMD driver for supporting error detection and correct (EDAC) for their Versal SoCs.
Earlier this year at the OCP Regional Summit in Prague AMD first presented openSIL as their new open-source CPU silicon initialization effort that can integrate with Coreboot and open-source boot firmware solutions. AMD openSIL is currently being prototyped on Genoa platforms but in a few years will eventually replace AGESA on both client and server processors. Later this month at the OCP Global Summit, there will be a new presentation on AMD openSIL.
A fix has made its way into the Linux PCI subsystem's power management branch to address various AMD Ryzen Rembrandt and Phoenix generation laptops failing to resume from suspend when external USB devices are attached for initiating the system resume.
As written about last month, AMD Linux engineers have been working on PMF Linux driver support for a "Smart PC Solutions Builder". The AMD Smart PC Solutions Builder feature is intended to provide OEMs with more control over system power/performance policies. It looks like systems making use of this feature are already to the marketplace or imminent with AMD having already landed the PMF firmware.
Checking for the presence of Intel virtualization (VMX) support and it being enabled can be easily achieved by looking at the flags in /proc/cpuinfo. But to this point AMD virtualization (SVM) has always been shown to user-space via /proc/cpuinfo even when the BIOS/platform has disabled SVM functionality. Finally for Linux 6.7 this oversight is being corrected.
The long ongoing saga of upstreaming AMD's Pensando Elba SoC support is now partially over with the initial enablement patches around the DeviceTree being queued as part of the SoC changes destined for the Linux 6.7 kernel cycle.
For a year and a half now Pensando has been working on enabling their Elba SoC support for the mainline Linux kernel - a process that coincidentally began just days after AMD announced it was acquiring Pensando. Over the past 18 months the AMD-Pensando Elba SoC enablement work has now been through 16 rounds of code review but still isn't over the finish line yet but some of the initial enablement code might finally land with Linux 6.7.
The AMD Platform Management Framework (PMF) Linux driver is preparing to support a new but potentially controversial feature that's seen little public information so far: the Smart PC Solutions Builder.
AMD's GPUOpen team today released a number of updated components for graphics application/engine developers.
AMD on Thursday released ZenDNN 4.1 as the newest release of their Zen Deep Neural Network Library for accelerating inference workloads on AMD Ryzen and EPYC processors. The ZenDNN library remains API compatible with Intel's oneDNN library and helps for building out optimized AI workloads for use on AMD Zen processors.
AMD announced today that their Versal HBM adaptive SoCs have entered production for these newest wares with high bandwidth memory.
Since early this year AMD has been working on Linux enablement patches for Dynamic Boost Control (DBC). This is a new feature of some AMD SoCs that allow an "authenticated entity" to have greater control over certain SoC characteristics to improve the power/performance. AMD DBC was merged for Linux 6.6 just days ago while already new patches have been posted that extend the supported platforms for this Dynamic Boost Control functionality.
AMD today published version 1.2 of their Unified Inference Frontend (UIF) that is for supporting deep learning inference across AMD CPUs, GPUs, as well as Versal adaptive SoCs and FPGAs. AMD UIF aims to accelerate AI inference across all AMD compute platforms with machine learning frameworks like PyTorch and TensorFlow. With UIF 1.2 adds initial support for Radeon GPUs.
On the AMD CPU side of the house, one of the patch series we are looking forward to seeing upstreamed in the Linux kernel is the AMD Preferred Core functionality that was initially sent out this summer. This AMD Preferred Core handling is built onto the AMD P-State driver and has been undergoing a few rounds of iteration with the latest "v5" patches having been posted this week.
AMD by way of their Xilinx acquisition has been running the Open Hardware Competition since 2015 for helping to drive new innovations around FPGAs and the ACAP compute platform. The winners of the AMD Open Hardware Competition 2023 were recently announced.
One of the patches to be picked up by the Linux 6.6 kernel this week brings back REP MOSQ for user-access on CPUs without Enhanced REP MOVSB (ERMS) support. In turn this can equate to some performance benefits on AMD CPUs lacking ERMS.
While I have been eagerly following the AMD openSIL project for open-source CPU initialization that will eventually replace AGESA, today AMD announced a new open-source firmware drop: the SEV firmware has been made open-source.
While we are eagerly awaiting ROCm support for more RDNA3 GPUs said to be coming later this calendar year, shipping Tuesday night was ROCm 5.6.1 as the newest point release for this open-source GPU compute stack.
Over the summer AMD Linux engineers began sending kernel patches for Family 1Ah (Family 26) processors that almost surely are for next-gen Zen 5 processors. Among the initial bits sent out for this next-generation AMD CPU Family are HWMON temperature monitoring and EDAC reporting, which are now on their way for the Linux 6.6 kernel.
Back in April AMD Linux engineers posted enabling a new CPU feature called Dynamic Boost Control to be found with some unspecified Ryzen SoCs for tuning the processor cores for optimal performance. The Dynamic Boost Control functionality depends upon the AMD Cryptographic Co-Processor (CCP) / Platform Security Processor (PSP) and this functionality was submitted today as part of the crypto updates for Linux 6.6.
As part of the Reliability, Availability and Serviceability (RAS) updates submitted today for the Linux 6.6 kernel is adding a quirk/workaround for dealing with current AMD Zen systems where a processor bug could lead to erroneously increased error severity and unneeded kernel panics.
It looks like the further-tuned AMD Inception / SRSO (Speculative Return Stack Overflow) mitigation code will be submitted for the upcoming Linux 6.6 merge window.
Following AMD recently posting Linux graphics driver patches for enabling a GFX 11.5 graphics engine and new DCN 3.5 display block that are presumably for an RDNA3 refresh such as for the Ryzen 8000 series APUs, today AMD posted additional open-source driver patch series for enabling additional new IP.
While AMD has acquired a number of hardware companies in the past several years, software company acquisitions by AMD has been much more rare. This morning AMD announced the acquisition of Mipsology as an AI software company.
Earlier this month when the AMD Inception CPU vulnerability was disclosed the initial mitigation was merged to Linux kernel right away for what there is referred to as the Speculative Return Stack Overflow (SRSO). Within a day of that code being published there were already efforts to clean it up and merged last week for Linux 6.5-rc7 was that AMD Inception code cleaning. This week a new set of 22 patches were published for further improving the AMD Inception/SRSO mitigation code.
For those that have been eyeing an AMD Ryzen 7 7040 "Phoenix" series laptop for Linux use, over the coming weeks ahead there will be benchmarks and a review on the Lenovo ThinkPad P14s Gen4 with AMD Ryzen 7 PRO 7840U laptop. With this 8-core / 16-thread Zen 4 mobile processor clocking up to 5.1GHz, 64GB of LPDDR5x-6400 memory, 1TB NVMe SSD, and 2.8K OLED display it should be a real treat if the Linux support is all in good shape.
Earlier this month the AMD Inception vulnerability was disclosed and quickly mitigated within the mainline Linux kernel and back-ported to the stable kernels. In the rush to get the code merged and the mitigation being under embargo until the disclosure date, some bugs and clean-ups with the mitigation code were discovered. That revised code was now submitted today for merging ahead of the Linux 6.5-rc7 kernel release this weekend.
A few minutes ago AMD sent out the very first open-source Linux kernel graphics driver patches for enabling the "GFX 11.5" graphics engine.
AMD's Preferred Core feature continues working its way toward the Linux kernel for this functionality that's been around since Zen 2.
Following last week's AMD Inception vulnerability another AMD Zen CPU bug came to light and that was performing a divide by zero on Zen 1 could end up leaking data with this DIV0 speculation bug. The original workaround was performing a dummy division 0/1 within the #DE exception handler but that's now turned out to be inadequate.
AMD Linux engineers continue work on enabling the Family 1Ah CPU models for the Linux kernel as what would appear to almost surely be the next-gen Zen 5 processors.
After a rather busy Patch Tuesday with the AMD Inception vulnerability and Intel Downfall going public, the Linux kernel saw a new bug fix merged today for a different issue... It turns out original AMD Zen 1 processors could end up leaking data in certain conditions after a divide by zero occurs.
Following yesterday's disclosure of the AMD "Inception" security vulnerability and the Linux kernel patches merged for reporting the mitigation status as well as the kernel-based handling for earlier generation Zen CPUs, the Family 19h microcode mitigations have now been picked up by the linux-firmware.git repository.
A new set of patches have been posted for the Linux kernel that implement AMD P-State Preferred Core handling for the amd-pstate driver.
AMD has kicked off a busy Patch Tuesday by disclosing INCEPTION, a new speculative side channel attack affecting Zen 3 and Zen 4 processors that require new microcode while prior Zen CPUs require a kernel-based solution.
As a follow-up to the first-on-Phoronix article last month that highlighted Linus Torvalds' frustrated views on the AMD fTPM random number generator continuing to cause problems for users even with updated firmware/BIOS, as of today the Linux kernel has gone ahead and blanket disabled RNG use for all current AMD fTPMs.
An updated patch-set was sent out on Wednesday to address fixing wake-up problems for some AMD client platforms when going through a suspend/resume cycle.
Linux's cpupower utility lives within the Linux kernel source tree for reading and tuning various CPU power settings rather than poking at sysfs files directly or other means of adjusting your processor power-related tunables. With the upcoming Linux 6.6 kernel cycle the cpupower utility is adding support for adjusting new AMD P-State driver features.
Last weekend I wrote about Zen 4's Automatic IBRS security feature needing STIBP enabled for protecting user-space processes. Single-Threaded Indirect Branch Predictors though haven't been enabled up to now with the Auto IBRS functionality on Linux. But the x86/urgent pull request sent out today ahead of the Linux 6.5-rc4 tagging makes that change.
Going back to May I wrote about AMD's "k10temp" Linux temperature driver being updated to handle negative temperature readings and now finally merged on Friday as a fix ahead of Linux 6.5-rc4 is a change to that open-source driver for properly displaying negative temperatures.
AMD lifted the embargo this evening on the Ryzen 9 7945HX3D, their first mobile processor sporting 3D V-Cache technology for boosting gaming performance and other cache-happy workloads.
1600 AMD news articles published on Phoronix.
