Brand: NVIDIA | Category: GPUs
SKU: NV-DCH200-NVL-HOP | Part #: 900-2G670-2500-000 | MPN: 900-2G670-2500-000
Contact for Pricing — Request a Quote
The NVIDIA H200 NVL 141GB HBM3e NVL (Dual) GPU delivers unparalleled performance for AI training and inference in dual GPU configurations. Equipped with 141 GB of HBM3e memory, this GPU leverages NVIDIA's NVLink technology to facilitate high-speed communication between GPUs, enabling efficient multi-GPU processing for large-scale AI applications. Its dual architecture is optimized for demanding workloads such as data science, deep learning, and enterprise AI inference, providing a flexible and scalable solution for modern data centers. Designed for maximum throughput, the H200 NVL ensures rapid training times and real-time inference capabilities.
| Brand | NVIDIA |
| Category | GPUs |
| SKU | NV-DCH200-NVL-HOP |
| Part Number | 900-2G670-2500-000 |
| Condition | New |
| Capacity | 141GB |
| Form Factor | NVL (Dual) |
| Interface | NVLink |
| GPU Architecture | Hopper |
| RAID Support | H200 |
| GPU Support | NVIDIA H200 |
| AI Optimized | Yes |
| GPU Form Factor | NVL (Dual) |
| GPU Interface | NVLink |
| GPU Memory | 141GB |
| GPU Memory Type | HBM3E |
| NVLink Support | Yes |
| GPU Platform | Data Center |
The NVIDIA H200 NVL 141GB HBM3e NVL (Dual) NVLink 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.
Key specifications for the NVIDIA H200 NVL 141GB HBM3e NVL (Dual) NVLink: new condition; gpu architecture Hopper; gpu form factor NVL (Dual); gpu interface version NVLink; gpu memory size 141GB; gpu memory type HBM3e; gpu nvlink support Yes. Manufacturer part number 900-2G670-2500-000. For the full datasheet with electrical, environmental, and compliance details, contact our pre-sales engineering team.
The NVIDIA H200 NVL 141GB HBM3e NVL (Dual) NVLink 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.