As part of my various end-of-year benchmarks, recently I looked at the Linux LTS kernel performance on AMD EPYC 9005 over the past year, the AMD EPYC Milan-X performance over the past four years, and various other performance comparisons over time to look the evolution of the Linux software performance. Another run I had carried out was looking at the AMD EPYC 8004 "Siena" series since its launch just over two years ago. Here is a look at how an up-to-date Linux software stack can deliver some additional performance gains for these energy efficiency and cost-optimized server processors.
A patch queued up into tip/tip.git's x86/cpu Git branch ahead of the upcoming Linux 6.20~7.0 kernel cycle enables the Intel Transactional Synchronization Extensions (TSX) functionality by default on the mainline kernel for capable CPUs and those not affected by side-channel attacks due to TSX Async Abort (TAA) and similar vulnerabilities. For newer Intel CPUs with safe TSX support, this change can mean better performance with the kernel defaults...
If all goes well the upcoming Ubuntu 26.04 LTS release will further enhance the NVIDIA graphics performance under its default GNOME Wayland session. The improvements might be upstreamed to GNOME 50 in time but otherwise it's looking like Ubuntu 26.04 will carry its own patch(es) for improving the NVIDIA Wayland performance...
The Linux kernel patches talked about at the start of the year for more easily changing the boot logo of Tux are now queued into a "for-next" branch and thus expected to be submitted for the upcoming Linux 6.20~7.0 kernel cycle. Those wanting to replace the Tux icon with an alternative logo during the Linux kernel boot process could already patch the file manually but this new code allows for an easy replacement via Kconfig options...
For those looking for a speedy Basic Linear Algebra Subprograms "BLAS" library, OpenBLAS 0.3.31 is now available for this optimized open-source implementation...
One of the initiatives launched by Intel in 2025 was LLM-Scaler as part of Project Battlematrix. The open-source LLM Scaler is a Docker-based solution for helping to deploy Generative AI "GenAI" workloads on Intel Battlemage graphics cards with frameworks like vLLM, ComfyUI, SGLang, and more. There continues to be routine new feature releases of LLM Scaler for broadening the large language models supported and other improvements...
Pages