Desir AI Ops Briefing: Week of 2026-08-12
Linux Finally Seeing Patches For Better Hybrid Graphics On 2018~2019 Era MacBook Pros
Patches posted today to the Linux kernel mailing list provide support for Apple GMUX hybrid graphics support on 2018~2019 Apple MacBook Pros. This allows for better power savings on these nearly decade old MacBook Pro laptops by allowing the discrete GPU to power down and be properly restored as needed…
Source: https://www.phoronix.com/news/Linux-2026-Patches-For-2018-MBP
Intel ANV Driver Enables Vulkan Video AV1 Encoding For DG2/Alchemist GPUs
Hyunjun Ko of Igalia continues working on Vulkan Video support for Intel graphics hardware on Linux with the ANV driver. Merged now for Mesa 26.3-devel is enabling AV1 video encoding via Vulkan Video on DG2/Alchemist hardware…
Source: https://www.phoronix.com/news/Intel-DG2-Vulkan-Video-AV1
Linux 7.2 To Revert Back To The FIFO DRM Scheduler Policy Due To “Fair” Regressions
One of the Direct Rendering Manager changes made for Linux 7.2 was switching to “fair” as the DRM scheduler default in place of the existing FIFO policy. The intent was on making the DRM scheduler policy more fair with this being used by AMDGPU and various other DRM kernel graphics drivers. But due to a last minute regression report, the fair policy is being reverted for Linux 7.2…
Source: https://www.phoronix.com/news/Linux-7.2-Reverting-DRM-Fair
Intel AMX Performance With The Xeon 678X Workstation CPU
One of the most compelling features for Intel Xeon 6 Granite Rapids alongside MRDIMMs is Advanced Matrix Extensions (AMX) for helping accelerate AI workloads and more. AMX is also found with the new Granite Rapids WS processors and in currently having my hands on an Intel Xeon 678X in the HP Z4 G6i workstation, I ran some benchmarks looking at the AMX performance and power impact for this high-end workstation.
Source: https://www.phoronix.com/review/intel-amx-xeon-678x
Intel LLM-Scaler Ready With Muse Glimmer Support, Other LLMs & Features
Intel’s LLM-Scaler project that was born out of their Project Battlematrix initiative aims to make it easier to run generative AI on Arc (Pro) B-Series graphics cards with the likes of vLLM, ComfyUI, SGLang, and other popular AI software in this Docker-based pre-configured AI stack. This week new LLM-Scaler releases brought same-day support for new models and other enhancements…
Source: https://www.phoronix.com/news/Intel-LLM-Scaler-Muse-Glimmer
FAQ
What changed this week?
ai accounted for 6 high-signal items this week.
How should operators respond?
Prioritize reliability runbooks before adding new tooling layers.