FP32/64 performance is going to be capped, as with prior generations.
Do you mean "enterprise ML"? Like, if I ran an ML company I wouldn't be looking at these cards?
Reply on news.ycombinator.com