Hacker News Reader

I'm one month into owning a GMKTec Evo-X2 with the new mad AMD gpu in it and so far I've managed to get Ollama running. ROCm doesn't officially support the GPU and everything seems to be hacky workaround on top of hacky workaround. Starting to wish I'd just waited for NVIDIA to actually release the project digits thing.

View on HN

So, how did you get it running? I‘m about to get a Bosgame M5 (same CPU).

I would recommend not wasting time with ollama, https://github.com/containers/ramalama just works on that chip.

Installing the ROCm nightlies from TheRock and upgrading the linux kernel to 6.16. However the kernel upgrade breaks a ton of other things, notably dkms.

I've been trying for weeks to get ComfyUI to not explode without success.

Reply on news.ycombinator.com