Skip to content
NVIDIA

Tesla V100 SXM2 32GB

125 FP16 · 32GB HBM2 · 300W

9Score
Budget ML TrainingClassic Deep LearningLegacy Pipelines

Specifications

ArchitectureVolta (GV100)
Memory32GB HBM2
Memory Bandwidth900 GB/s
FP16 TFLOPS125
FP8 TFLOPS0
BF16 TFLOPS0
INT8 TOPS62
TDP300W
InterconnectNVLink 2.0 (300 GB/s) (300 GB/s)
EcosystemCUDA
GenerationPrevious
Est. Price$3,000

Memory

Generation

HBM2

Capacity

32 GB

Bandwidth

0.9 TB/s

900 GB/s

Bus Width

4,096-bit

HBM Vendor(s)SK Hynix

Recommended Configuration

8× V100 SXM2 in DGX-1 or DGX-2

Training Intelligence

◆CUDA
◆PyTorch
◆TensorFlow
◆JAX
◆DeepSpeed
Training Time Estimates
LLaMA 7B(7B)
~2 days8 GPUs
GPT-3 175B(175B)
~38 days1024 GPUs
SDXL(3.5B)
~5 days8 GPUs

Cloud cost: $0.80/hr

Rent this GPU in the Cloud

Cloud rental rates not tracked for this GPU.

Methodology

Specs come from the manufacturer's published datasheets. Cloud rates are regularly tracked from each provider's own pricing page and shown per GPU-hour, on-demand. GPUAdvisor does not rate or rank providers. Full methodology →