The NPU in AMD’s upcoming Ryzen AI Max 400 series, codenamed “Gorgon Halo”, will be supported in Linux 7.4. The enablement patch has been applied to drm-misc-next, the staging tree for the next merge window.

A two-line patch

The patch adds the AIE2 revision 9 NPU, identified as Gorgon Halo (“GRGH”), to the device table of the AMDXDNA accelerator driver. That is the whole change: the NPU is effectively the same design already supported in Strix Point, Strix Halo, and Gorgon Point, so the existing driver works once it recognises the hardware.

It is a small patch with an outsized effect: without that table entry, the NPU on a brand-new laptop simply does not appear to Linux. It had apparently been overlooked until shortly before the hardware reached the market.

What the NPU is for

An NPU (neural processing unit) is a low-power accelerator for machine learning inference. AMD’s NPUs have become genuinely usable under Linux for running small language models and features like background blur and speech recognition, at far lower power than the GPU. For local AI work more broadly, our Ollama guide covers running models on the CPU and GPU, which remains the most common path.

Timing

Linux 7.3 is due in mid-October (it reached rc5 on September 27). The 7.4 merge window opens immediately after, and 7.4 should be released around December. Distributions will pick it up from there: rolling releases within weeks, fixed releases in their next version.

Other AMD work queued for 7.4 includes PerfOpt, a new performance optimisation feature for AMD integrated GPUs. Our Linux GPU drivers guide covers how AMD’s kernel drivers fit together.