2 ms·
Good point - I wonder what the market split is there. Even so, wouldn’t it be far more economical to stack a bunch of 6800xts for ML, or is the ML support that
by taurath 6y ago
Good point - I wonder what the market split is there. Even so, wouldn’t it be far more economical to stack a bunch of 6800xts for ML, or is the ML support that far behind on AMD?
- lumost 6y agoML on AMD is currently impractical, the mainstream tools have hard deps on CUDA and the AMD projects are so far half-baked.
- alfalfasprout 6y agoI used to think so but recent testing with ROCM has shown the tables are starting to turn.