[juujian]

Except AMD still has not figured out how to deliver on LLM acceleration on Linux.

[d3Xt3r]

Not sure what you mean by that, but ROCm works just fine. I'm using a patched ollama (for APU support) [1] and I can run models like llama3.1:8b and deepseek-r1:14b just fine on my Radeon 780M iGPU, on Arch (btw).

[1] https://github.com/rjmalagon/ollama-linux-amd-apu