Flopper.io System Specification Sheet
August 10, 2026
flopper.io/system/nvidia-dgx-h100NVIDIA DGX H100
8x H100 Hopper rack 2022
Overview
15.8
FP8 PFLOPS Dense
0.6 TB
Aggregate Memory
26.8 TB/s
Aggregate Bandwidth
8.5 kW
Aggregate Power
Performance Metrics
| Precision | Peak (PFLOPS) | Per GPU (TFLOPS) | TFLOPS/W |
|---|---|---|---|
| FP64 | 0.3 | 34.0 | 32.00 |
| FP32 | 0.5 | 67.0 | 63.06 |
| TF32 | 4.0 | 494.5 | 930.82 |
| FP16 | 7.9 | 989.5 | 1862.59 |
| BF16 | 7.9 | 989.5 | 1862.59 |
| FP8 | 15.8 | 1979.0 | 3725.18 |
| INT8 | 15.8 | 1979.0 | 3725.18 |
All figures are dense. Vendors commonly headline the sparse number, which is twice the dense one.
Configuration
- Accelerator
- H100 SXM5 80GB
- Count
- 8
- Memory per GPU
- 80 GB
- Bandwidth per GPU
- 3.4 TB/s
Power & Efficiency
- Aggregate Power
- 8.5 kW
- TDP per GPU
- 700 W
- FP8 Efficiency
- 3725.18 TFLOPS/W
- Aggregate Bandwidth
- 26.8 TB/s
System Details
Vendor: NVIDIA
Family: DGX
Architecture: Hopper
Form Factor: rack
Release Year: 2022
Process Node: 4nm
Interconnect
NVSwitch + NVLink
Notes
Enterprise AI system with 8x H100 SXM5 GPUs