NVIDIA H100 NVL
Data center GPU · Hopper
Launched
Q1 2023
VRAM
94 GB
Mem. bandwidth
3,900 GB/s
On-demand from
Tech specs
VRAM 94 GB HBM3
Memory bandwidth 3,900 GB/s
Interface PCIe
CUDA cores 14,592
Tensor cores 456 (Gen 4)
TDP 400 W
Supported data types
FP64FP32FP16BF16FP8INT8
Cloud rental prices
5 providers · available in 18 countries · lowest price per GPU, per hour
On-demand
$2.92 – $6.98 /GPU/h
Reserved
$3.58 – $5.24 /GPU/h
Spot
$0.81 – $1.40 /GPU/h
| Provider | On-demand | Reserved | Spot |
|---|---|---|---|
 Atlantic.Net | $3.94 | $3.58 | — |
| | $6.98 | $3.84 | $1.40 |
 Massed Compute | $2.92 | — | — |
 Runpod | $3.19 | — | — |
 Vast.ai | — | — | $0.81 |
Models that fit in VRAM
Open-weights models that fit in VRAM. Estimated using 🤗 accelerate, plus approximation for up to 8K context.
Media
Peak theoretical performance
FP8 Tensor Core 1,670 TFLOPS
INT8 Tensor Core 1,670 TOPS
BF16 Tensor Core 835 TFLOPS
FP16 Tensor Core 835 TFLOPS
TF32 Tensor Core 417 TFLOPS
FP32 60 TFLOPS
FP64 30 TFLOPS
FP64 Tensor Core 60 TFLOPS
Performance figures assume no sparsity; in cases where only sparse performance figures are published by the manufacturer, these are halved to give approximate dense performance.
Resources
Detailed documentation from the manufacturer.
Similar GPUs
Other accelerators you might compare.