Reaction
Nvidia GB300 FP4 reportedly trails TPUv7 FP8 on performance per dollar for external customers across most of the latency curve
SemiAnalysis says GB300 FP4’s performance-per-dollar is 1.64 times worse than TPUv7’s for external customers across most of the curve.
TLDR
SemiAnalysis says Nvidia’s first GB300 FP4 counter-submission is 1.64 times worse than TPUv7 FP8 on performance per dollar for external customers across most of the end-to-end latency curve. It expects further Nvidia submissions as the company tries to narrow the gap.
Combined views
—
1 Source, first seen ago
