Resource Library

This document represents the reference architecture for the NVIDIA Enterprise AI Factory for Supermicro SYS-212GB-FNR systems. It is a comprehensive explanation of the technology used, each component’s purpose and configuration.

Supermicro and AMD Instinct MI355X Benchmarking Report: Results demonstrate a double-digit performance uplift over CPU-only systems where the workloads can take advantage of GPU’s compute acceleration, enabling workloads that previously required months of simulation time to be completed in days with GPU acceleration. Read the report to learn more.

Supermicro Building Block solutions for 5G networks, the intelligent edge, and embedded applications.

Supermicro’s portfolio includes several high-performance storage platforms specifically engineered for AI workloads on object storage, delivering the high throughput and low latency required for both inference and training.

Seamless Liquid-to-Air Cooling Solution for Air-Cooled Data Centers

Liquid-Cooled Infrastructure-Scale Solution

Infrastructure-Scale Facility Water Cooling Designed for Low-PUE/WUE Data Centers

High-Efficiency Liquid-to-Liquid Cooling Solution for Cluster-Level Deployment

Compact, Plug-and-Play Cooling Solution for High-Density AI and HPC Racks

Autonomous Multi-Agent Monitoring, Debugging, and Remediation on Supermicro A+ Platforms with AMD Instinct™ GPUs.

In this solution brief, we will describe a high-performance storage cluster, purpose-built for the most demanding AI training and inference workloads running over an Ethernet network. The key components of the storage architecture include Supermicro’s Petascale server equipped with Micron E3.S NVMe, connected via NVIDIA Spectrum-X Ethernet.

All-in-One Edge Infrastructure Solution for Stores