← Back to catalog
NVIDIADatacenter
NVIDIA H100 SXM5 80GB
The Hopper workhorse. Still the backbone of most training clusters.
$38,000
80GB HBM3
Memory
1,979 TFLOPS
FP16 Tensor (sparse)
NVLink 4.0, 900 GB/s
Interconnect
The GPU that built the current AI boom. SXM5 form factor with 4.0 for multi-GPU scaling inside a compatible server baseboard (e.g. DGX H100).
Full Specifications
Compute
- Architecture
- Hopper
- FP16 Tensor (sparse)
- 1,979
Memory
- Memory
- 80GB
- Memory Bandwidth
- 3.35 TB/s
Interconnect & Networking
- Interconnect
- 4.0, 900 GB/s
Power & Cooling
- 700W
General
- Form Factor
- SXM5 module
Compatible & Required
Pairs well with
Add software & support
NVIDIA AI Enterprise
$4,500/yr