AMD Wants Your Logs To Help Optimize PyTorch & ComfyUI For Strix Halo, Radeon GPUs

If you are not satisfied with the current performance for PyTorch or ComfyUI / Stable Diffusion on your Strix Halo APU system or with other consumer RDNA3/RDNA4 Radeon consumer GPUs, AMD engineers are interested in your logs to help better optimize the performance going forward.

For any PyTorch program or ComfyUI / Stable Diffusion scenario especially where the performance is coming up short on the likes of the Strix Halo Radeon 8060S, Radeon RX 9000 series, or other RDNA3/RDNA4 GPUs,
Continue reading...

Open Source LLM

Opus 1.6 Audio Codec Adds New Machine Learning Functionality

Version 1.6 of libopus as the library for the open-source Opus audio codec is now available. Opus 1.6 brings new machine learning "ML" based features in building atop the machine learning features initially added to Opus 1.5 .

Opus 1.6 changes include:
- A new wideband-to-fullband bandwidth extension (BWE) module
- Support for 96 kHz audio with Opus HD
- Significant improvement to Deep Redundancy (DRED)
- A new 24-bit encoder/decoder API
Continue reading...

Linux Exposing Support For Lenovo ThinkPads Being Able To Detect Hardware Damage

Newer Lenovo ThinkPads are adding the ability to detect and report varying degrees of hardware damage. The Lenovo ThinkPad ACPI driver for Linux is being adapted for being able to communicate said hardware damage to user-space Linux software.

It turns out newer ThinkPads will begin communicating detected hardware damage that can then be parsed by the OS. A new patch to Lenovo's ThinkPad ACPI open-source driver explains:
"Thinkpads are adding the ability to detect and report hardware damage status. Add new sysfs interface to identify whether hardware damage
Continue reading...

Intel's Cache Aware Scheduling Presentation At LPC 2025

One of the exciting Intel innovations to the Linux kernel this year has been around the Cache Aware Scheduling for helping to deliver better performance on modern CPUs with multiple last level caches. The kernel patches have yet to be upstreamed but testing has shown to be quite promising for grouping tasks sharing data to the same LLC domain to help reduce cache misses and cache bouncing. Those wishing to learn more about Cache Aware Scheduling, there was a presentation on it last week by Intel engineers Tim Chen and Chen Yu at the Linux Plumbers Conference
Continue reading...

The Significant Performance Gains For Radeon RADV Ray-Tracing Performance In 2025

As part of my various year-end comparison benchmarking, I recently ran some tests looking at how the Radeon RX 9000 series RDNA 4 performance has evolved since its debut near the beginning of the year. The Vulkan ray-tracing performance in particular was standing out this year as having evolved quite nicely while for conventional OpenGL and Vulkan performance the performance has been largely stable this year with its great at-launch support.

The biggest area of performance gains this year for the open-source Radeon RADV driver within

Continue reading...

Mozilla Names New CEO, Firefox To Evolve Into A "Modern AI Browser"

Mozilla Corporation has named its new CEO in replacing interim CEO Laura Chambers.

Anthony Enzor-DeMeo has been named the new CEO of Mozilla Corporation. Anthony Enzor-DeMeo was SVP of Firefox from 2004 to July 2005 and then from July until now was the GM of Firefox at Mozilla. He's written a public message today in his first day serving as the new chief executive for Mozilla.
"People want software that is fast, modern, but also honest about what it does.

Red Hat Acquires Another AI Company

Last year Red Hat acquired Neural Magic as part of their AI acquisitions and to bolster the open-source AI ecosystem. Today they announced another AI acquisition.

Red Hat announced today they have acquired Chatterbox Labs, a private company that began in 2011 and is focused on AI model testing and generative AI guardrails. Chatterbox Labs' AIMI platform provides quantitative AI risk metrics and other features.

Continue reading...

ZLUDA For CUDA On Non-NVIDIA GPUs Enables AMD ROCm 7 Support

The ZLUDA open-source project that has been through several incarnations but ultimately about getting CUDA software up and running on non-NVIDIA GPUs now supports the AMD ROCm 7 series.

ZLUDA is working on bringing CUDA to non-NVIDIA GPUs . While there were prior versions focused originally on Intel GPUs and then for a while AMD-financed work on Radeon/ROCm support, the current take is on being a multi-vendor CUDA implementation and with a special focus on getting CUDA AI workloads up and running.

The latest achievement
Continue reading...

Servo 0.0.3 Browser Engine Brings Better Performance, Embedding Improvements

Servo 0.0.3 is out today as the newest update to this Rust-based, open-source web layout engine. Servo 0.0.3 incorporates many enhancements made over the past month from better performance to enhancing its embedding API and other improvements like context menus for more web content.

Servo 0.0.3 released today and the Servo project also put out a blog post to highlight its improvements as part of developments over the month of November. The highlights in the new release include:

-
Continue reading...

MidnightBSD 4.0 Brings Many Changes To This FreeBSD 13 Derived OS

While FreeBSD 15 stable was officially released earlier this month, MidnightBSD continues plotting its own course atop its FreeBSD 13 base. Out today is MidnightBSD 4.0 as the latest iteration of this desktop-minded BSD operating system.

MidnightBSD 4.0 incorporates the latest updates from the FreeBSD 13 series, updates many third-party libraries, and features the latest iteration of its mport package manager. The mport verify functionality is now more robust, improved color handling for the TUI, and other changes.

MidnightBSD 4
Continue reading...