Comment by aruggirello

8 days ago

Minisforum, Gmktec also have Ryzen AI HX 370 mini PCs with 128Gb (2x64Gb) max LPDDR5. It's dirt cheap, you can get one barebone with ~€750 on Amazon (the 395 similarly retails for ~€1k)... It should be fully supported in Ubuntu 25.04 or 25.10 with ROCm for iGPU inference (NPU isn't available ATM AFAIK), which is what I'd use it for. But I just don't know how the HX 370 compares to eg. the 395, iGPU-wise. I was thinking of getting one to run Lemonade, Qwen3-coder-next FP8, BTW... but I don't know how much RAM should I equip it with - shouldn't 96Gb be enough? Suggestions welcome!

4 comments

aruggirello

Reply
cpburns2009  8 days ago

I benchmarked unsloth/Qwen3-Coder-Next-GGUF using the MXFP4_MOE (43.7 GB) quantization on my Ryzen AI Max+ 395 and I got ~30 tps. According to [1] and [2], the AI Max+ 395 is 2.4x faster than the AI 9 HX 370 (laptop edition). Taking all that into account, the AI 9 HX 370 should get ~13 tps on this model. Make of that what you will.

[1]: https://community.frame.work/t/ai-9-hx-370-vs-ai-max-395/736...

[2]: https://community.frame.work/t/tracking-will-the-ai-max-395-...