AMD News Archives

The ACPI, power management, and thermal subsystem pull requests have been sent in for the in-development Linux 6.1 kernel.

AMD on Friday upstreamed new Family 19h CPU microcode to linux-firmware.git.

While some Linux enthusiasts eagerly recommend users boot their systems with the "mitigations=off" kernel parameter for run-time disabling of various relevant CPU security mitigations for Spectre, Meltdown, L1TF, TAA, Retbleed, and friends, with the new AMD Ryzen 7000 "Zen 4" processors while still needing some software mitigations, it's surprisingly faster for the most part leaving the relevant mitigations enabled.

One of the new drivers set to make its debut with Linux 6.1 is the AMD Platform Management Framework "PMF" with an intent on "making AMD PCs smarter, quieter, power efficient by adapting to user behavior and environment" with next-generation hardware. Another part of AMD PMF, the Cool and Quiet Framework (CnQF) has also been queued up for introduction in Linux 6.1.

This morning I called attention to some pending work around a 20 year old chipset workaround in the Linux kernel had been hurting modern AMD systems by erroneously still applying the change to modern hardware. Fortunately, that patch has now been picked up by Linus Torvalds in time for the Linux 6.0 kernel expected for its stable debut next weekend.

AMD engineer K Prateek Nayak recently uncovered that a ~20 year old chipset workaround in the Linux kernel still being applied to modern AMD systems is responsible in some cases for hurting performance on modern Zen hardware. Fortunately, a fix is on the way for limiting that workaround to old systems and in turn helping with performance for modern systems.

While Blender 3.2 introduced AMD HIP on Linux support for GPU acceleration and the recent Blender 3.3 extended the AMD GPU Cycles acceleration back to GFX9/Vega GPUs, for those wanting AMD ray-tracing support within Blender it's not expected to come until Blender 3.5.

AMD is expected to announce their Radeon RX 7000 "RDNA 3" graphics cards on 3 November.

Earlier this month I wrote about AMD working on s2idle fixes for some AMD Ryzen 6000 series "Rembrandt" laptops. At the time it was just for select ASUS laptops known to have a bug in the firmware resulting in suspend-to-idle issues while now additional models not only from ASUS but also Lenovo have been uncovered.

While we are used to seeing Intel engineers dominating the speaker lists at various Linux events around the world, it's been a number of years since AMD engineers held multiple presentations like they did this week for the Linux Plumbers Conference and now the Open-Source Summit EU taking place in Dublin, Ireland.

For those recently picking up an ASUS laptop powered by AMD Ryzen Mobile 6000 series "Rembrandt" SoCs or considering such a device, AMD has prepared a set of fixes for the suspend-to-idle support.

AMD engineers are working to better optimize the Linux kernel's scheduler for split-LLC (last level cache) processor designs, namely to benefit their EPYC server processors.

Over the past year AMD engineers have been developing the AMD P-State CPU frequency scaling driver as an alternative to the long-used ACPI CPUFreq driver to provide better performance/power efficiency with Zen 2 and newer Ryzen/EPYC processors. Today they have now introduced AMD P-State EPP as they aim to deliver better performance-per-Watt.

Due to bringing a number of new SoC designs to market next year and trying to make their model numbers easier to decipher, AMD announced this morning a new naming system for Ryzen mobile processors.

Back in early June AMD engineers began posting support for enabling Virtual NMI on Linux for AMD CPUs with KVM and permitting hardware support. VNMI is expected to finally happen on the AMD side with Zen 4 processors and today they posted their latest revision of this work.

Earlier this month AMD posted Linux kernel patches preparing LbrExtV2 as updated Last Branch Record functionality being introduced with upcoming AMD Zen 4 processors. That LbrExtV2 support for the Linux kernel's "perf" subsystem has now been queued up in its respective branch ahead of the Linux 6.1 feature merge window beginning in early October.

One of the questions that has come up following my AMD Ryzen Threadripper PRO 5965WX Linux testing has been how well air-cooling is working out for the 280 Watt workstation CPU. Water cooling is, of course, most ideal but there are air coolers that can work out sufficiently too. Here are some quick reference results.

In addition to AMD-Xilinx working on new network driver code, a new DRM display driver, and other kernel features recently covered on Phoronix, they are also preparing upstream Linux kernel support for the "CDX" bus with their FPGA devices.

As I've written about the past several weeks, AMD engineers have been preparing a Platform Management Framework (PMF) driver for Linux. The AMD Platform Management Framework for future hardware appears similar to Intel's Dynamic Platform and Thermal Framework (DPTF) and designed to enhance the thermal/power performance of future platforms.

While Linux 6.0 won't be out for another month and a half, on deck already for Linux 6.1 is the AMD Platform Management Framework "PMF" driver being queued into platform-drivers-x86.git.

Up to this point loading updated CPU microcode on AMD processors under Linux has checked just to ensure every physical CPU core was loaded with the new microcode but not sibling threads for SMT processors. While logically that makes sense, it turns out some AMD microcode updates do carry out per-thread modifications that means the microcode updating needs to be carried out on every thread. A Linux fix is on its way to the kernel to adjust that behavior.

AMD today sent out revised patches for improving the AMD P-State CPU frequency scaling Linux driver that aims to provide better power efficiency than the generic ACPI CPUFreq driver that has long been relied upon for AMD processors.

Sent out this morning is a Linux kernel "fix" that now enabled STIBP when using the IBPB mode for Retbleed mitigations on AMD processors. In other words, more protections needed for this enhanced mode of Retbleed mitigation.

The colorful fishy codenames are not over for AMD's Linux driver crew! While on the GPU side they have moved to IP block-by-block enablement strategy for their future GPUs, over on the audio co-processor side AMD posted a series of patches today under the "Pink Sardine" codename.

While all of the key Zen 4 CPU functionality appears in place for the mainline Linux kernel, AMD engineers continue working to enable other new Zen 4 features for use under Linux. The newest patches out of AMD this morning are for LbrExtV2.

With AMD EPYC showing some nice gains on Linux 6.0, I've been eager to begin testing Linux 6.0 on more systems especially now that the v6.0 merge window is winding down... With now having the shiny new AMD Ryzen Threadripper 5965WX, I decided to take this high-end 24-core chip for a run with Linux 6.0 Git to see how it performs over Linux 5.19 stable.

In addition to Intel's busy Patch Tuesday, AMD today made public CVE-2021-46778 that university researchers have dubbed the "SQUIP" attack as a side channel vulnerability affecting the execution unit scheduler across Zen 1/2/3 processors.

AMD engineers have released an updated version of AOMP, their LLVM/Clang downstream that carries the company's latest patches around OpenMP offloading to Radeon GPUs.

AMD Automatic Mode Transition (AMT) is a new feature wired up for Ryzen-powered ThinkPad laptops that is being introduced with the Linux 6.0 kernel.

You may recall the Phoronix news earlier this year around an AMD "Sabrina" SoC appearing in Coreboot for open-source system firmware support. Over the past few months we've cited a number of AMD Sabrina hits in open-source code but outside of that haven't heard much else about "Sabrina" or seen it on AMD's roadmaps.

Earlier this year AMD began posting Linux kernel patches for >a href="https://www.phoronix.com/news/AMD-Zen-4-IBS-Linux">Instruction Based Sampling (IBS) extensions coming with Zen 4 processors. With Linux 5.19 the Zen 4 IBS extensions landed while now with Linux 6.0 the perf tools have been updated for dealing with Zen 4 IBS.

With Linux 6.0 having some big scheduler changes and tuning that should specifically benefit AMD Zen systems, I've been eager to see how some high core count EPYC servers will benefit from this next version of the Linux kernel. While just a few days into the Linux 6.0 merge window, here are some early benchmarks showing some of the areas where Linux 6.0 is allowing higher performance out of existing AMD EPYC 7003 series hardware.

AMD recently started posting Linux patches for a Platform Management Framework "PMF" driver that is designed to "enhance end user experience by making AMD PCs smarter, quieter, power efficient by adapting to user behavior and environment."

Some of the newest Linux patches out of AMD for new processors are implementing support for some recently-documented Quality of Service extensions: L3SBE and BMEC.

Among the early pull requests for the now-open Linux 6.0 merge window (nee Linux 5.20) are a few AMD additions worth mentioning.

The newest AMD Linux optimization patch for the kernel aims to introduce a cool down period for the AMD Platform Security Processor (PSP) after each I2C transaction between the x86 CPU and the PSP.

With the upcoming Linux 5.20 cycle is support for AMD's Sensor Fusion Hub v1.1 revision being found in newer Ryzen laptops.

AMD today posted new open-source Linux driver code for enabling the audio co-processor (ACP) with the upcoming Raphael platform with the Ryzen 7000 series processors.

Last week AMD quietly released AOCL 3.2 as the newest version of their optimized CPU software libraries for use across Ryzen, Ryzen Threadripper, and EPYC platforms.

Following some weekend benchmarks here are more complementary numbers on the Retbleed mitigation performance benchmark costs. These additional numbers are on a Zen 2 based AMD Ryzen 7 4800U APU that has been common both to laptops as well as embedded/low-profile devices for thin client computing, IoT / edge use-cases, and more.

Back in Linux 5.17 the AMD P-State "amd_pstate" driver was introduced for Ryzen and EPYC systems as an alternative to the ACPU CPUFreq frequency scaling driver with an emphasis on delivering better power efficiency for modern AMD Zen 2 and newer systems. Since the mainlining there hasn't been too much change to this driver but now a new patch series has been sent out with some updates.

It's great seeing AMD continuing to hire for more Linux/open-source driver developers. Beyond their many roles they are still working to fill on the CPU side of the house, they have a new job posting in hiring for their open-source GPU driver stack with a focus on multimedia efforts.

Today's Coreboot code now has AMD Rembrandt SoC support by splitting it out from the Sabrina SoC support that has been in the works the past several months for this open-source firmware project.

Somewhat surprisingly, AMD engineers have been working on some new Linux kernel patches for their aging Jadeite platform.

After AMD announced FidelityFX Super Resolution 2.0 back in March, as of today they have made good on their word to open-source it.

AMD is using the Embedded World conference in Nürnberg to launch the Ryzen Embedded R2000 series for industrial use-cases along with IoT, thin clients, and edge computing.

After announcing the Threadripper PRO 5000 WX series back in March and with Lenovo being their launch partner for these Zen 3 Ryzen Threadripper CPUs, AMD today shared an update on availability.

Earlier this year Pensando engineers began posting Linux patches for enabling their Elba DPU SoC. This data processing unit is powered by 16 x Arm Cortex-A72 cores and designed for supporting up to dual 200GE networking with this SoC intended for high-end networking equipment. It didn't take long for the AMD integration less than one month after AMD completed its Pensando acquisition with the new Linux patches now reflected as the AMD Pensando Elba.

One of my personal gripes with AMD's Zen CPU support on Linux has been the lack of timely support for CPU temperature monitoring with their "k10temp" driver. Even though usually just new IDs are often needed and sometimes needing to adjust offsets or other minor changes, it has traditionally been done post-launch and sometimes left up to patches from the open-source community. Thankfully that has been changing and with Zen 4 it looks like that support will be ready for launch-day with the mainline Linux kernel.

While not record-shattering like the 1.1 Exaflops Frontier supercomputer at ORNL that took the Top500 spot this year from Fugaku, LUMI was inaugurated today with the claim of Europe's most powerful supercomputer.