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

NVIDIA A100 80GB PCIe

NVIDIA Data Center80GB HBM2e • 2 TB/s • 19.5 TFLOPS FP32
$10,000 - $15,0007.37%In Stock
TDP
400W
VRAM
80GB HBM2e
Bandwidth
2.0 TB/s
Compute
19.5 FP32 / 624 TF16
Released
May 2020
Availability
In Stock

Description

The A100 is NVIDIA's previous-generation data center GPU based on the Ampere architecture. Still widely used for training and inference in cloud environments.

Benchmarks

MLPerf Training
Baseline
HPC Linpack
19.5 TFLOPS
ResNet-50
10,320 img/s
BERT-Large
34.5 min