Data Centre GPUs
NVIDIA H200 NVL
141 GB HBM3e Hopper GPU for large-model inference and HPC
Starting at
₹69,85,999
H200-NVL-141G · excluding GST · final price confirmed on quotation
- GPU Memory
- 141 GB HBM3e
- AI Performance
- 3,341 TFLOPS FP8 (sparse)
- Board Power
- Up to 600 W, dual slot
Not sure this is the right model? Ask an engineer
Overview
The H200 NVL is the PCIe version of NVIDIA's H200, built for air-cooled enterprise racks. Its 141 GB of HBM3e memory at 4.8 TB/s holds large language models in memory for fast inference, and it accelerates fine-tuning and HPC workloads.
Up to four cards can be joined with NVLink bridges at 900 GB/s per GPU, and each card splits into as many as seven MIG instances. It includes a five-year NVIDIA AI Enterprise subscription. We supply it in NVIDIA-Certified Systems and size power and cooling for it.
Specifications
| H200 NVL | |
|---|---|
| Performance | |
| Architecture | NVIDIA Hopper |
| FP64 | 30 TFLOPS |
| FP64 Tensor Core | 60 TFLOPS |
| FP32 | 60 TFLOPS |
| TF32 Tensor Core (sparse) | 835 TFLOPS |
| BF16 / FP16 Tensor Core (sparse) | 1,671 TFLOPS |
| FP8 / INT8 Tensor Core (sparse) | 3,341 TFLOPS / TOPS |
| Memory | |
| GPU Memory | 141 GB HBM3e |
| Memory Bandwidth | 4.8 TB/s |
| Interconnect and Features | |
| Interconnect | 2- or 4-way NVLink bridge, 900 GB/s per GPU; PCIe Gen 5, 128 GB/s |
| Decoders | 7x NVDEC, 7x JPEG |
| Multi-Instance GPU (MIG) | Up to 7 instances of 16.5 GB |
| Confidential Computing | Supported |
| NVIDIA AI Enterprise | Five-year subscription included |
| Power and Form Factor | |
| Max Power | Up to 600 W (configurable) |
| Form Factor | PCIe, dual slot, air cooled |
| Server Options | NVIDIA MGX H200 NVL and NVIDIA-Certified Systems with up to 8 GPUs |
Specifications are published by NVIDIA and may change between hardware revisions. We confirm the exact figures on your quotation.
Pricing & options
Indicative prices in Indian rupees, excluding GST. Choose an option and send a buy request; we confirm availability and the final price on a formal quotation.
1 options
More Data Centre GPUs
All NVIDIA products-
RTX PRO 6000 Blackwell Server Edition
96 GB Blackwell GPU for AI, rendering and virtual workstations in servers
- GPU Memory
- 96 GB GDDR7 with ECC
- AI Performance
- 4 PFLOPS FP4
- Board Power
- Up to 600 W, passive
Price on request
Buy -
