TBH the A100 is the chip powering the boom. Thats what most of the breakthrough models/papers seem to be trained on. And not just because its fast, but becaus

TBH the A100 is the chip powering the boom. Thats what most of the breakthrough models/papers seem to be trained on. And not just because its fast, but because it just works. Most researchers aren't going to waste time figuring out how to run an MI250 or Gaudi 2 or whatever in PyTorch.