NVIDIA H200 NVL 141GB HBM3e
NVIDIA DATA CENTER GPU / AI TRAINING / LLM
A data center accelerator with 141GB of HBM3e and 4.8 TB/s of bandwidth in a dual-slot, air-cooled PCIe form factor. Two- or four-way NVLink bridging makes it the mainstay for large language model training and inference.
Specifications
| Architecture | NVIDIA Hopper |
|---|---|
| Single-Precision Performance | 60 TFLOPS |
| AI Performance | FP8 Tensor Core 3,341 TFLOPS |
| GPU Memory | 141 GB HBM3e |
| Memory Bandwidth | 4.8 TB/s |
| Max Power Consumption | Up to 600W (Configurable) |
| Multi-Instance GPU | Up to 7 MIGs @ 16.5GB |
| Confidential Computing | Supported |
| Graphics Bus | PCIe Gen5 x16 (128 GB/s) |
| Interconnect | 2- or 4-way NVLink bridge, 900 GB/s per GPU |
| Form Factor | PCIe, Dual-slot air-cooled |
| Thermal Solution | Passive |
| NVENC | NVDEC | JPEG | — | 7x NVDEC | 7x JPEG |