QuData

Rent H100 NVL by the Hour by QuData Ai

Explore the specifications and benchmarks of the H100 NVL. Compare hardware configurations and performance metrics for scalable AI and HPC workloads with the Hopper architecture in NVLink configuration.

GPU Description

Explore the specifications and benchmarks of the H100 NVL. Compare hardware configurations and performance metrics for scalable AI and HPC workloads with the Hopper architecture in NVLink configuration.

GPU Characteristics

GPU typeH100 NVL
GPU connectionNVLink
GPU model nameNVIDIA H100 NVL
Driver version535.183.01
GPU memory94GB HBM3
Power Limit (W)800.0
GPU temperature (°C)56
GPU Clock Speed (MHz)1980
Memory Clock Speed (MHz)2619
Memory bandwidth3350 GB/s

GPU Configuration

CUDA cores16896
Tensor core528 (4th generation)
Support for FP16/INT8Yes
ArchitectureHopper
VirtualizationNVIDIA MIG

Benchmarks

ffmpeg~5ms
Coremark (iterations per second)~155000-185000
Llama2 Inference (tokens per second)~480-600
Tensorflow Mixed Training~15.8-19.8
Stable Diffusion (images/min)~120-150
Mixed Precision Training~22.5-28.2x

Providers with H100 NVL