Manual · GPU

H100: Hopper

The FP8 machine that made frontier inference affordable. Transformer Engine, NVLink 4, and the architecture every other GPU is compared to.

H100 is the reference point. When people say "a GPU", they usually mean an H100. It's the Ampere A100's successor, and it's the GPU that made FP8 inference real.

What changed from A100

The numbers

Why it matters for inference

FP8 is the reason H100 became the default serving GPU. A 70B model in FP8 is ~70GB, which fits on one H100 with room for KV cache. Before FP8, you needed two A100s. That single fact reshaped the economics of serving: one GPU per frontier-ish model, instead of two.

Hopper also introduced CUDA graphs as a first-class optimization (capture the kernel graph, replay it), which became the standard way to cut launch overhead. And its DPX instructions (for dynamic programming) matter less for LLMs but show the specialization trend.

Sources

Back to the manual