GPU Servers Supporting H100 PCIe
Quad GPU 2U Server with 1x 4th Gen EPYC 9004
Quad GPU Dual Intel Xeon Scalable Workstation
8x GPU 4U Server with Dual 4th Gen EPYC 9004
NVIDIA HGX H100 SXM Solutions
NVIDIA HGX 4x H100 Dual Intel Xeon Scalable 4U
NVIDIA HGX 8x H100 Dual AMD EPYC 9004 8U
NVIDIA HGX 8x H100 Dual Intel Xeon Scalable 8U
9 Times the Performance per Chip
NVIDIA H100 is the world's most advanced chip built with the largest generational leap. The H100 incorporates the ability to processing on FP8 for AI training, calculates using mixed-precision processing, introduces a transformer engine, and executes via DPX Instructions to accelerate dynamic programming algorithms. The H100 provides up to 9x faster in training and 30x faster in inferencing.
NVIDIA H100 is the building block of performance prowess. Scale AI factories with numerous H100 SXM5 (found in DGX and HGX) deliver the highest threshold of computing or super charge mainstream data centers with H100 PCIe for easy to implementation.
Scalable AI Factory with NVIDIA DGX
NVIDIA DGX H100 is a scalable server designed to accelerate machine learning for HPC and AI workloads. 8 NVIDIA H100 SXM5 GPUs are NVLinked together to operate as one giant GPU capable of 32 PFLOPS of AI Performance reducing your AI model's total time to market from weeks to days.
NVIDIA DGX H100 is just the building block; The DGX SuperPOD™ is 32 DGX H100s (256 H100s) to power the most complex AI systems. Contact our team for more information on NVIDIA DGX H100.
Build your ideal system