NVIDIA V100

Data center GPU · Volta · SXM2 and PCIe variants

Summary

NVIDIA's flagship Volta-generation data centre GPU, and the first to feature Tensor Cores. Three variants were produced, differing in memory capacity and form factor: V100 32GB SXM, V100 16GB SXM and V100 16GB PCIe. The V100 has since been superseded by the A100 and newer GPUs.

Launched
2017
VRAM
16 - 32 GB
Mem. bandwidth
900 GB/s
On-demand from
Compare prices from 7 providers →

Tech specs

Spec V100 32GB SXMV100 16GB SXMV100 16GB PCIe
VRAM 32 GB HBM216 GB HBM216 GB HBM2
Memory bandwidth 900 GB/s900 GB/s900 GB/s
Interface PCIe SXM PCIe
CUDA cores 5,120 5,120 5,120
Tensor cores 640 (Gen 1)640 (Gen 1)640 (Gen 1)
TDP 300 W300 W250 W
Supported data types
FP64FP32FP16INT8

Cloud rental prices

7 providers · available in 19 countries · lowest price per GPU, per hour

Prices updated

On-demand
$1.53 – $3.90 /GPU/h
Reserved
$1.30 – $2.40 /GPU/h
Spot
$0.28 /GPU/h
Provider On-demand Reserved Spot
AWS ? $3.90 $1.30 —
Azure ? $1.53 $1.35 $0.28
Sesterce ✓ $2.57 — —
Compare all V100 32GB SXM cloud providers & configurations →
On-demand
$0.78 – $3.06 /GPU/h
Provider On-demand Reserved Spot
AWS ? $3.06 — —
Oracle ? $2.95 — —
OVHcloud ? $0.78 — —
Compare all V100 16GB SXM cloud providers & configurations →
On-demand
$0.23 – $3.06 /GPU/h
Reserved
$0.21 – $0.23 /GPU/h
Spot
$0.12 – $0.57 /GPU/h
Provider On-demand Reserved Spot
Azure ? $3.06 — $0.57
Lambda ? $0.79 — —
Sesterce ✓ $2.57 — —
Verda ? $0.23 $0.21 $0.12
Compare all V100 16GB PCIe cloud providers & configurations →

Models that fit in VRAM

Open-weights models that fit in VRAM. Estimated using 🤗 accelerate, plus approximation for up to 8K context.

Resources

Detailed documentation from the manufacturer.

Similar GPUs

Other accelerators you might compare.