More memory for large AI models
NVIDIA H200 is designed for applications that benefit from greater memory capacity and bandwidth. We will configure the server, accelerator count and network architecture for your performance target.
Specifications
Review the main product parameters for a quick comparison. Before delivery, we confirm the exact configuration, compatibility, warranty and required quantity.
| Architecture | Hopper |
|---|---|
| Memory | 141 GB HBM3e |
| Memory bandwidth | 4.8 TB/s |
| Form factor | SXM / PCIe NVL by SKU |
| MIG | up to 7 MIG instances |
| Workloads | memory-bound LLM / inference / training |
Best-fit applications
A high-performance option for large-language-model training, inference, retrieval, analytics and long-context applications.
Configuration and delivery
Important for a reliable deployment
H200 requires compatible server nodes, cooling and power. We include these requirements in the project configuration and delivery plan.
Review the main product parameters for a quick comparison. Before delivery, we confirm the exact configuration, compatibility, warranty and required quantity.
Other solutions
Compare similar solutions or describe your application and we will recommend the best-fit model.
Tailored commercial offer
Send the model and quantity or attach an equipment list. If the exact configuration is not yet selected, describe your application and we will prepare a suitable offer.