NVIDIA HGX A100 4-GPU
baseboard
4× A100
NVLink
2020
FP64
0.0
PFLOPS
FP32
0.1
PFLOPS
Power
3.2
kW Total
Memory
320
GB Total
HGX A100 4-GPU configuration
FP64
0.04
PFLOPS
12.1 TFLOPS/kW
FP32
0.08
PFLOPS
24.4 TFLOPS/kW
TF32
0.62
PFLOPS
195.0 TFLOPS/kW
FP16
1.25
PFLOPS
390.0 TFLOPS/kW
System Details
GPU Configuration
System Specifications
Form Factor: baseboard
Total Power: 3.2 kW
Total Memory: 320 GB
Memory Bandwidth: 8156 GB/s
Precision Performance Breakdown
| Precision | System Performance | Per GPU | Efficiency |
|---|---|---|---|
| FP64 | 0.039 PFLOPS | 9.7 TFLOPS | 12.1 TFLOPS/kW |
| FP32 | 0.078 PFLOPS | 19.5 TFLOPS | 24.4 TFLOPS/kW |
| TF32 | 0.624 PFLOPS | 156.0 TFLOPS | 195.0 TFLOPS/kW |
| FP16 | 1.248 PFLOPS | 312.0 TFLOPS | 390.0 TFLOPS/kW |
| BF16 | 1.248 PFLOPS | 312.0 TFLOPS | 390.0 TFLOPS/kW |
| INT8 | 2.496 PFLOPS | 624.0 TFLOPS | 780.0 TFLOPS/kW |
Powered by NVIDIA A100
This system utilizes 4 × NVIDIA A100 SXM4 80GB GPUs, each delivering exceptional performance for AI and HPC workloads.
Per GPU TDP
400W
Per GPU Memory
80 GB
Process Node
7nm
Architecture
Ampere
Documentation & Resources
workstation-datasheet-dgx-spark-gtc25-spring-nvidia-us-3716899-web.pdf
NVIDIA • 2025-10-16
NVIDIA DGX Documentation
System guides, deployment resources
Typical Use Cases
Large Language Model Training
Distributed Deep Learning
Multi-GPU Inference
HPC Simulations
Scientific Computing
The NVIDIA HGX A100 4-GPU with 4× A100 GPUs is designed for enterprise-scale workloads requiring 0.0 PFLOPS of compute power with NVLink interconnect for efficient multi-GPU communication.