NVIDIA 900-21010-0040-000 H200 NVL
Brand: NVIDIA | Category: GPUs
SKU: NVI-900-21010-0040-000 | Part #: 900-21010-0040-000 | MPN: 900-21010-0040-000
Contact for Pricing — Request a Quote
About the NVIDIA 900-21010-0040-000 H200 NVL
The NVIDIA 900-21010-0040-000 H200 NVL from NVIDIA is enterprise-grade GPUs hardware built for data centers. It features 141GB capacity, PCIe 5.0 x16 interface. The NVIDIA H200 NVL (part number 900-21010-0040-000) is a high-capacity GPU module built on the NVIDIA Hopper architecture, engineered to address the most demanding large-scale AI inference and training workloads in enterprise data centers. The H200 NVL features HBM3e memory technology, delivering substantially higher memory bandwidth and capacity compared to prior-generation HBM2e solutions, enabling practitioners to run larger AI models entirely in GPU memory without resort Available in brand new condition. NVIDIA gpus products are trusted by enterprises and data centers worldwide. Available from Omnixon Global. Contact our sales team to request a quote.
Technical Specifications
| Brand | NVIDIA |
| Category | GPUs |
| SKU | NVI-900-21010-0040-000 |
| Part Number | 900-21010-0040-000 |
| Condition | New |
| Capacity | 141GB |
| Interface | PCIe 5.0 x16 |
| Manufacturer Part Number | 900-21010-0040-000 |
| Product Name | NVIDIA H200 NVL |
| GPU Architecture | NVIDIA Hopper (GH200) |
| GPU Configuration | Dual-GPU NVLink module (2x H200 GPUs on one board) |
| Memory Type | HBM3e |
| Total Memory Capacity | 141 GB per GPU / 282 GB aggregate (NVLink pair) |
| Memory Bandwidth | Up to 4.8 TB/s per GPU (aggregate ~9.6 TB/s across NVLink pair) |
| NVLink Interconnect Bandwidth | 900 GB/s bidirectional (NVLink 4.0) |
| FP8 Tensor Core Performance | Up to 3,958 TFLOPS per GPU (sparsity) |
| FP16 / BF16 Tensor Core Performance | Up to 1,979 TFLOPS per GPU (sparsity) |
| FP32 (CUDA Core) Performance | Up to 66.9 TFLOPS per GPU |
| Form Factor | SXM5 |
| Thermal Design Power (TDP) | 700 W per GPU |
| Interconnect | NVLink 4.0, PCIe Gen 5 |
| CUDA Compute Capability | 9.0 |
| ECC Support | Yes (HBM3e on-die ECC) |
| Multi-Instance GPU (MIG) | Supported (up to 7 instances per GPU) |
| Supported Precision Formats | FP64, TF32, FP32, BF16, FP16, FP8, INT8 |
Frequently Asked Questions about NVIDIA 900-21010-0040-000 H200 NVL
What does the NVIDIA 900-21010-0040-000 H200 NVL do?
The NVIDIA 900-21010-0040-000 H200 NVL accelerates AI/ML training, inference, scientific HPC, and virtualization (vGPU) workloads. Typical deployments include LLM training clusters, computer-vision pipelines, financial risk modeling, and rendering farms.
What are the headline specs of the NVIDIA 900-21010-0040-000 H200 NVL?
Key specifications for the NVIDIA 900-21010-0040-000 H200 NVL: capacity 141GB; interface PCIe 5.0 x16; new condition; model H200 NVL; headline specification 141GB HBM3e; platform detail PCIe 5.0 x16 accelerator; use case AI, graphics, simulation, analytics, edge, or network acceleration. Manufacturer part number 900-21010-0040-000. For the full datasheet with electrical, environmental, and compliance details, contact our pre-sales engineering team.
Is the NVIDIA 900-21010-0040-000 H200 NVL compatible with my infrastructure?
The NVIDIA 900-21010-0040-000 H200 NVL requires a PCIe Gen4 or Gen5 x16 slot, server power adequate for the card's TDP, and CUDA/ROCm driver support in your hypervisor or bare-metal OS. Sales engineering will confirm chassis fit (1U/2U/4U), PCIe lane count, and PSU headroom before quoting.