Glossary3 sections1 min read
What is Hopper?
Hopper is the NVIDIA data center GPU architecture behind the H100 and H200. It introduced FP8 Tensor Cores and the Transformer Engine, which manages precision automatically for transformer models.
Why it matters when renting a GPU
Hopper GPUs remain the reference for LLM training and inference, with mature software support.
H100 and H200 share the same compute; the H200 adds more and faster memory.
Worked example
EverXCloud describes 6 Hopper-based models (including Grace Hopper). The H100 SXM and the H200 are both rated at 989 dense BF16 TFLOPS; memory goes from 80 GB to 141 GB.
Frequently asked questions
- What is the Transformer Engine?
- A Hopper feature that switches between FP8 and 16-bit precision layer by layer to speed up transformer models while preserving accuracy.