NVIDIA DGX SuperPOD

NVIDIA DGX Superpod is the reference architecture and shows best practice for large scale AI infrastructure deployments. All aspects of a SuperPOD are optimized to provide everything needed for rapid deployment and successful operations while ensuring great return-on-investment and total cost of ownership.The exact specification of a NVIDIA DGX SuperPOD will be adapted to the requirement of individual customers, but generally includes the following:

• At least 31 latest generation DGX Servers
• Non-blocking high speed InfiniBand network, allowing for multiple DGX servers to work simultaneously on large tasks. Typically, one dedicated InfiniBand port is used per GPU.
• At least 500TB of shared parallel storage, enabling the DGX servers to work directly on large datasets, without the need to move the data to local storage in each server.
• Installation, configuration and handover testing on-site by NVIDIA certified technicians.
• NVIDIA management software, providing an easy-to-use interface to maintain the solution and prioritize workloads.
• A dedicated NVIDIA professional services contact person.
• At least 3 years of warranty and support. Support is provided directly from NVIDIA.