2 ms·
Well, one confounding factor is that CPU Flops are more generic, for any algorithm. GPU Flops as mentioned work better on tensor cases. However, when we do hav
by etaioinshrdlu 5y ago
Well, one confounding factor is that CPU Flops are more generic, for any algorithm. GPU Flops as mentioned work better on tensor cases.
However, when we do have tensors, the GPU and CPU would both work to their full potential, and thus the flops comparison ought to be valid.