Active Models2,847indexed
Model Rankings165updated
Global Latency11msoptimal
Active Models2,847indexed
Model Rankings165updated
Global Latency11msoptimal

NVIDIA H100 SXM5 80GB

NVIDIA Data Center80GB HBM3 • 3.35 TB/s • 67 TFLOPS FP32
$25,000 - $40,000-2.56%In Stock
TDP
700W
VRAM
80GB HBM3
Bandwidth
3.35 TB/s
Compute
67 FP32 / 3,958 FP8
Released
Sep 2022
Availability
In Stock

Description

The H100 is NVIDIA's flagship data center GPU built on the Hopper architecture. Features fourth-gen Tensor Cores with FP8 precision, Transformer Engine, and NVLink for massive GPU clusters.

Benchmarks

MLPerf Training
2.8x vs A100
HPC Linpack
51.2 TFLOPS
ResNet-50
28,848 img/s
BERT-Large
12.1 min