NVIDIA DGX H200

rack
8× H200
NVSwitch + NVLink
2024
FP64
0.3
PFLOPS
FP32
0.5
PFLOPS
Power
8.5
kW Total
Memory
1128
GB Total

Latest DGX system with H200 GPUs and 141GB HBM3e memory per GPU

FP64
0.27
PFLOPS
64.0 TFLOPS/kW
FP32
0.54
PFLOPS
126.1 TFLOPS/kW
TF32
3.96
PFLOPS
930.8 TFLOPS/kW
FP16
7.92
PFLOPS
1862.6 TFLOPS/kW

System Details

GPU Configuration

GPU Count: 8 GPUs
Architecture: Hopper
Interconnect: NVSwitch + NVLink

System Specifications

Form Factor: rack
Total Power: 8.5 kW
Total Memory: 1128 GB
Memory Bandwidth: 38400 GB/s

Precision Performance Breakdown

PrecisionSystem PerformancePer GPUEfficiency
FP64 0.272 PFLOPS 34.0 TFLOPS 64.0 TFLOPS/kW
FP32 0.536 PFLOPS 67.0 TFLOPS 126.1 TFLOPS/kW
TF32 3.956 PFLOPS 494.5 TFLOPS 930.8 TFLOPS/kW
FP16 7.916 PFLOPS 989.5 TFLOPS 1862.6 TFLOPS/kW
BF16 7.916 PFLOPS 989.5 TFLOPS 1862.6 TFLOPS/kW
FP8 15.832 PFLOPS 1979.0 TFLOPS 3725.2 TFLOPS/kW
INT8 15.832 PFLOPS 1979.0 TFLOPS 3725.2 TFLOPS/kW

Powered by NVIDIA H200

This system utilizes 8 × NVIDIA H200 SXM 141GB GPUs, each delivering exceptional performance for AI and HPC workloads.

Per GPU TDP

700W

Per GPU Memory

141 GB

Process Node

4nm

Architecture

Hopper

Documentation & Resources

workstation-datasheet-dgx-spark-gtc25-spring-nvidia-us-3716899-web.pdf

NVIDIA • 2025-10-16

View Document ↗

NVIDIA DGX Documentation

System guides, deployment resources

Browse ↗

Typical Use Cases

Large Language Model Training
Distributed Deep Learning
Multi-GPU Inference
HPC Simulations
Scientific Computing

The NVIDIA DGX H200 with 8× H200 GPUs is designed for enterprise-scale workloads requiring 0.3 PFLOPS of compute power with NVSwitch + NVLink interconnect for efficient multi-GPU communication.