The Rust-based Redox OS open-source operating system provided a status update on all of their interesting development activities during the month of March. In addition to a lot of code improvements, Redox OS also enhanced its documentation as well as added an AI policy to reject any contributions relying on large language models...

At the end of March AMD announced the Ryzen 9 9950X3D2 processor with both dies having 3D V-Cache. This 16 core processor with 206MB of total cache is quit exciting and will be on sale later this month but AMD refrained from commenting on the price until today...

Last month Intel announced the Arc Pro B70 with 32GB of GDDR6 video memory for this long-awaited Battlemage G31 graphics card. This new top-end Battlemage graphics card with 32 Xe cores and 32GB of GDDR6 video memory offers a lot of potential for LLM/AI and other use cases, especially when running multiple Arc Pro B70s. Last week Intel sent over four Arc Pro B70 graphics cards for Linux testing at Phoronix. Given the current re-testing for the imminent Ubuntu 26.04 release, I am still going through all of the benchmarks especially for the multi-GPU scenarios. In this article are some initial Arc Pro B70 single card benchmarks on Linux compared to other Intel Arc Graphics hardware across AI / LLM with OpenVINO and Llama.cpp, OpenCL compute benchmarks, and also some OpenGL and Vulkan benchmarks. More benchmarks and the competitive compares will come as that fresh testing wraps up, but so far the Arc Pro B70 is working out rather well atop the fully open-source Linux graphics driver stack.

Set to be merged for the upcoming Linux 7.1 merge window is the "Yogafan" hardware monitoring driver to provide fan speed monitoring not only for Lenovo Yoga laptops but also various Legion and IdeaPad laptops too...

Intel's OpenVINO toolkit for optimizing and deploying AI inferencing across their range of hardware platforms is out with its newest quarterly feature update. There is official support for Intel's latest hardware as well as enabling more large language models and other new AI innovations for this excellent open-source Intel software project...

Google on Tuesday announced the Chrome 147 stable release to all Windows, macOS, and Linux users. There are a number of refinements in this latest routine Chrome stable update paired with various fixes and new developer capabilities...

Announced today from the PyTorch Conference EU in Paris is word that Hugging Face has contributed their Safetensors project to the PyTorch Foundation, which is an umbrella organization under the Linux Foundation for hosting AI initiatives. Safetensors aims to help mitigate arbitrary code execution risks and more...

In addition to the release today of Flatpak 1.16.4 for shipping new security fixes including a sandbox escape and the ability to delete host files, XDG-Desktop-Portal 1.20.4 is also now available with another security fix of its own to prevent sandboxed apps from being able to trash arbitrary host files...

That Flatbox app sandboxing and distribution system is out today with important security updates...

Jay is a new open-source shader compiler being developed for Intel's open-source OpenGL and Vulkan Linux drivers. Ultimately this Jay shader compiler should help in delivering better Linux graphics performance with modern Intel hardware...

For those that like to make remarks about AMD "fine wine" especially when it comes to open-source Linux drivers and/or nostalgic about feature work on really old hardware, to much amusement there are new patches today for the AMD InterWave ISA sound card from the 1990s...

Last week I provided benchmarks to quantify how the AMD Strix Halo graphics performance has evolved since launch one year ago, in today's article is a look at how the Zen 5 CPU performance with the flagship Ryzen AI Max+ 395 has evolved under Linux in the year since these exciting APUs began making their way to high-end laptops and desktops. Complementing the nice Radeon 8060S performance gains are also some nice CPU performance benefits quantified when using Ubuntu 26.04.

The Intel QuickAssist "QAT" driver for the mainline Linux 7.1 kernel is adding support for Zstd offloading across QuickAssist Gen 4 / Gen 5 / Gen 6 accelerators for Zstandard compression as well as Zstandard decompression (limited there to the latest Gen 6 hardware)...

Greg Kroah-Hartman, the main Linux stable kernel maintainer and typically viewed as the second-in-command to the Linux kernel development, has turned to new "gregkh_clanker_t1000" fuzzing tooling to help uncover new kernel bugs...

Framework Computer on Monday issued their latest update concerning the ongoing price increases for memory and solid state drives affecting the industry. There has been some more price increases, signs of some temporary reprieve, and then a bit of good news on pricing for select Framework hardware...

It looks like with the Linux 7.2 kernel later in the year the AMD ISP4 driver will finally be merged to mainline. This driver is needed for the web camera on the HP ZBook Ultra G1a Strix Halo laptop and other future AMD Ryzen laptops...

Introduced in Linux 5.13 back in 2021 was eXecute In Place "XIP" support for RISC-V that allows for the kernel image to be executed from ROM. The intent is on allowing the kernel to run from non-volatile storage like NOR flash that is directly addressable by the CPU and to reduce RAM usage. But after RISC-V XIP support is broken for months at a time, the feature is now set to be retired from the mainline kernel...

It doesn't change too much in practice with how Mesa updates typically have been handled under Fedora Linux, but now it's officially documented: Mesa graphics drivers have a permanent updates exception so new Mesa versions can be shipped as updates in Fedora stable releases...

Following last month's Lemonade SDK 10.0 release that finally makes AMD Ryzen AI NPUs under Linux useful for running large language models (LLMs) where as before the Linux build could only target GPUs, released on Monday was Lemonade 10.1 with more enhancements to this local LLM solution...

Building on prior Mesa contributor guidelines and discussions among upstream Mesa developers, there are two generative AI "GenAI" policies that have now been decided upon for Mesa development moving forward...