Live
AI hardware

NVIDIA H100 PCIe

FP16 throughput
7.6×10¹⁴ FLOP/s
Memory
74.5 GB
TDP
350 W

NVIDIA H100 PCIe is an AI accelerator designed by NVIDIA (GPU), released in September 2022. It is built on a 4 nm process at TSMC, packing 80B transistors.

It delivers 7.6×10¹⁴ FLOP/s of tensor FP16/BF16 throughput. It carries 74.5 GB of memory at 1.9 TB/s of bandwidth. Its thermal design power is 350 W.

Full record
Manufacturer
NVIDIA
Type
GPU
Release date
Sep 20, 2022
Tensor FP16/BF16
7.6×10¹⁴ FLOP/s
FP32
5.1×10¹³ FLOP/s
FP64
2.6×10¹³ FLOP/s
Memory
74.5 GB
Memory bandwidth
1.9 TB
TDP
350 W
Process node
4 nm
Transistors (millions)
80,000
Die size (mm2)
814
Foundry
TSMC
More from NVIDIA
SourceEpoch AI, 'Machine Learning Hardware'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/data/machine-learning-hardware. Licensed under CC BY 4.0.
← All ai hardware