NVIDIAAI GPU · PCIe (NVL)
H200 NVL
NVIDIA H200 NVL is a Hopper data-center accelerator with 141 GB HBM3e and 4.8 TB/s of memory bandwidth.
Specifications
Dense throughput, no sparsity| Architecture | Hopper |
|---|---|
| Form factor | PCIe (NVL) |
| Process | TSMC 4N |
| FP16 / BF16 tensor (dense) | 835.5 TFLOPS |
| FP8 tensor (dense) | 1,670.5 TFLOPS |
| FP4 tensor (dense) | — |
| Memory | 141 GB HBM3e |
| Memory bandwidth | 4.8 TB/s |
| Power | Up to 600 W |
| Launch | Nov 2024 |
AI index
Tensor × bandwidth100 × √((835.5 ÷ 989.4) × (4.8 ÷ 3.35)) = 110.0
The AI index is the geometric mean of dense 16-bit tensor throughput and memory bandwidth, each relative to the H100 SXM. It is a spec-sheet comparison, not a measured benchmark.