ROCm Status in mid 2026 [D]
Reddit - Machine Learning
1 min read
About this article
Hey folks I'm starting to hear that ROCm works fine for inference now. But, I've not seen any reports on how viable it is for training. I have a couple of RTX 3090s I use for prototyping models, but I'm considering switching to a pair of RX7900XTX instead. On paper at least, the RX7900XTX can output about 4 times the throughput at FP16 with a similar power draw, VRAM, and cost. Based on PyTorch docs, it seems like ROCm is now fully supported, but I'm struggling to find user reports on how wel...
You've been blocked by network security.To continue, log in to your Reddit account or use your developer tokenIf you think you've been blocked by mistake, file a ticket below and we'll look into it.Log in File a ticket
Originally published on May 07, 2026.
Curated by AI News.