News for hackers. Stuff that matters.
Re:AMD's CDNA 3 Compute Architecture by lostmsu on Tuesday December 19, 2023 @09:53AM (#38696238)
Microsoft has DirectML and a PyTorch backend for it. Although it is stale and significantly slower on 3090 than CUDA (5x+ as of 2 years ago).

There are no comments on this comment yet.