NVIDIA DGX SuperPOD GB200

pod
2304× B200
NVLink, NVSwitch, InfiniBand
2025
FP64
85.2
PFLOPS
FP32
172.8
PFLOPS
Power
3840.0
kW Total
Memory
414720
GB Total

32-rack SuperPOD configuration with GB200 NVL72 systems

FP64
85.25
PFLOPS
27.0 TFLOPS/kW
FP32
172.80
PFLOPS
54.0 TFLOPS/kW
TF32
2534.40
PFLOPS
1500.0 TFLOPS/kW
FP16
5184.00
PFLOPS
2700.0 TFLOPS/kW

System Details

GPU Configuration

GPU Count: 2304 GPUs
Architecture: Blackwell
Interconnect: NVLink, NVSwitch, InfiniBand

System Specifications

Form Factor: pod
Total Power: 3840.0 kW
Total Memory: 414720 GB
Memory Bandwidth: 17740800 GB/s

Precision Performance Breakdown

PrecisionSystem PerformancePer GPUEfficiency
FP64 85.248 PFLOPS 37.0 TFLOPS 27.0 TFLOPS/kW
FP32 172.800 PFLOPS 75.0 TFLOPS 54.0 TFLOPS/kW
TF32 2534.400 PFLOPS 1100.0 TFLOPS 1500.0 TFLOPS/kW
FP16 5184.000 PFLOPS 2250.0 TFLOPS 2700.0 TFLOPS/kW
BF16 5184.000 PFLOPS 2250.0 TFLOPS 2700.0 TFLOPS/kW
FP8 10368.000 PFLOPS 4500.0 TFLOPS 5400.0 TFLOPS/kW
FP4 20736.000 PFLOPS 9000.0 TFLOPS 10800.0 TFLOPS/kW
INT8 10368.000 PFLOPS 4500.0 TFLOPS 5400.0 TFLOPS/kW

Powered by NVIDIA B200

This system utilizes 2304 × 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 DGX SuperPOD GB200 with 2304× B200 GPUs is designed for enterprise-scale workloads requiring 85.2 PFLOPS of compute power with NVLink, NVSwitch, InfiniBand interconnect for efficient multi-GPU communication.