NVIDIA H200 SXM5
NVIDIA · Hopper (refresh)
Shipping
traininginferenceflagshiphbm3e
Memory
141GB HBM3e
FP8 TFLOPS
3958
Power
700W
Released
Nov 1, 2024
Technical Specifications
| Manufacturer | NVIDIA |
|---|---|
| Architecture | Hopper (refresh) |
| Memory | 141GB HBM3e |
| Memory Bandwidth | 4.8 TB/s |
| FP16 TFLOPS | 1979 |
| FP8 TFLOPS | 3958 |
| Interconnect | NVLink4 (900GB/s) |
| Power (TDP) | 700W |
| Released | Nov 1, 2024 |
| Status | Shipping |
Availability
Available via neoclouds; hyperscaler rollouts ongoing
Pricing
Retail (estimated)
35000-45000
Cloud Pricing ($/hr)
| coreweave | $2.49/hr |
|---|---|
| lambda_labs | $2.49/hr |
Key Specs
Memory: 141GB HBM3e
Bandwidth: 4.8 TB/s
FP8: 3958 TFLOPS
Power: 700W
Datacenters Using This GPU
- xAI Memphis Supercluster
- Microsoft Quincy Data Center Campus
- Microsoft San Antonio Data Center
- Microsoft West Des Moines AI Campus
- Microsoft Cheyenne Data Center
- AWS Ashburn Data Center Complex
- AWS Umatilla Data Center
- Meta Temple Data Center
- Oracle Mesa Data Center Campus
- CoreWeave New Jersey AI Campus
- CoreWeave Illinois AI Campus
- Lambda Cloud GPU Infrastructure
- Together AI Training Cluster
- Sakura Internet Ishikari AI Campus
- Yotta D1 AI Data Center
- Jean Zay Supercomputer Expansion
- Canada AI Compute Infrastructure (Toronto)
- Singapore AI Compute Cluster
- NVIDIA Israel-1 Supercomputer