Hacker News Reader

CUDA runs on essentially all NVIDIA consumer gpus. ROCm is supported on a single 5 year old model and two 2 year old models.

https://rocm.docs.amd.com/projects/install-on-linux/en/docs-...

View on HN

The people who write their documentation aren't very good. "Support" seems to mean something like tested + enterprise grade support. It works on a fair number of officially unsupported cards. Eventually it'll work well on everything even though the support matrix is unlikely to ever get bigger.

Although, as mentioned, works under a cycloptic vision where the card is only doing compute tasks. I'd be interested to know if even supported cards can multitask GPU and pure compute tasks without crashing because it looks like it might be a design issue. Hard to tell with driver corruption. Maybe the testing catches that on supported cards; who knows.

Reply on news.ycombinator.com