One Petaflop on a single chip.10x Better than Current DL Processors. Data movement, not compute, is the bottleneck. Current deep learning processors target the compute problem, ignoring 90% of the problem. Vathys’ hardware and software takes a different approach.
