NVIDIA GB200 NVL32

rack
32× B200
NVLink
2025
FP64
1.2
PFLOPS
FP32
2.4
PFLOPS
Power
53.3
kW Total
Memory
5760
GB Total

16 GB200 superchips (32 B200 GPUs + 16 Grace CPUs) in a rack-scale system

FP64
1.18
PFLOPS
27.0 TFLOPS/kW
FP32
2.40
PFLOPS
54.0 TFLOPS/kW
TF32
35.20
PFLOPS
1501.0 TFLOPS/kW
FP16
72.00
PFLOPS
2701.0 TFLOPS/kW

System Details

GPU Configuration

GPU Count: 32 GPUs
Architecture: Blackwell
Interconnect: NVLink

System Specifications

Form Factor: rack
Total Power: 53.3 kW
Total Memory: 5760 GB
Memory Bandwidth: 246400 GB/s

Precision Performance Breakdown

PrecisionSystem PerformancePer GPUEfficiency
FP64 1.184 PFLOPS 37.0 TFLOPS 27.0 TFLOPS/kW
FP32 2.400 PFLOPS 75.0 TFLOPS 54.0 TFLOPS/kW
TF32 35.200 PFLOPS 1100.0 TFLOPS 1501.0 TFLOPS/kW
FP16 72.000 PFLOPS 2250.0 TFLOPS 2701.0 TFLOPS/kW
BF16 72.000 PFLOPS 2250.0 TFLOPS 2701.0 TFLOPS/kW
FP8 144.000 PFLOPS 4500.0 TFLOPS 5403.0 TFLOPS/kW
FP4 288.000 PFLOPS 9000.0 TFLOPS 10806.0 TFLOPS/kW
INT8 144.000 PFLOPS 4500.0 TFLOPS 5403.0 TFLOPS/kW

Powered by NVIDIA B200

This system utilizes 32 × NVIDIA B200 SXM 180GB GPUs, each delivering exceptional performance for AI and HPC workloads.

Per GPU TDP

1000W

Per GPU Memory

180 GB

Process Node

4nm

Architecture

Blackwell

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

AI/ML Training
High-Performance Computing
Data Analytics

The NVIDIA GB200 NVL32 with 32× B200 GPUs is designed for enterprise-scale workloads requiring 1.2 PFLOPS of compute power with NVLink interconnect for efficient multi-GPU communication.