Summary

DGX SuperPOD with NVIDIA DGX H100 systems is the next generation of data center scale architecture to meet the demanding and growing needs of AI training. This RA document for DGX SuperPOD represents the architecture used by NVIDIA for our own AI model and HPC research and development. DGX SuperPOD continues to build upon its high-performance roots to enable training of the largest NLP models, support the expansive needs of training models for automotive applications, and scaling-up recommender models for greater accuracy and faster turn-around-time.

DGX SuperPOD represents a complete system of not just hardware but all the necessary software to accelerate time-to-deployment, streamline system management, proactively identify system issues, and support the same accelerated software that your leverage across DGX SuperPOD, on laptops, or other NVIDIA GPU-based systems. The combination of all these components keeps systems running reliably, with maximum performance, and enables users to push the bounds of state-of-the-art. The platform is designed to both support the workloads of today and grow to support tomorrow’s applications.