After announcing the launch of HGX H200 on Microsoft, Google Cloud, Oracle, and AWS, NVIDIA is also tailoring supercomputing solutions for Hewlett Packard Enterprise (HPE). HPE has unveiled a supercomputing solution tailored for generative AI applications, targeting large enterprises, research institutions, and government organisations, powered by NVIDIA GH200. 

The comprehensive solution comprises a software suite facilitating the training and tuning of AI models using private datasets, coupled with liquid-cooled supercomputers, accelerated compute capabilities, networking, storage, and services.

The generative AI supercomputing solution is set to be generally available in December, accessible through HPE in more than 30 countries.

This HPE initiative, developed in collaboration with NVIDIA, emphasises purpose-built solutions to accelerate AI model training and outcomes. The software suite integrates with HPE Cray supercomputing technology, powered by NVIDIA Grace Hopper GH200 Superchips, providing unprecedented scale and performance for substantial AI workloads. Notably, the system’s advanced capabilities enhance system performance by 2-3X.

Key components of the supercomputing solution include AI/ML acceleration software, featuring the HPE Machine Learning Development Environment, NVIDIA AI Enterprise, and the HPE Cray Programming Environment suite. 

The solution, based on the HPE Cray EX2500 and the NVIDIA GH200 Grace Hopper Superchips, can scale up to thousands of GPUs, dedicating the full capacity of nodes to a single AI workload for faster time-to-value. The network infrastructure, HPE Slingshot Interconnect, supports real-time AI with high-speed networking.

HPE emphasises turnkey simplicity with its Complete Care Services, providing global specialists for setup, installation, and full lifecycle support to simplify AI adoption.

Looking ahead, HPE is committed to sustainability, aiming to address the anticipated surge in AI workloads’ power requirements by delivering solutions with liquid-cooling capabilities. These solutions can potentially drive up to a 20% performance improvement per kilowatt over air-cooled alternatives while consuming 15% less power. The supercomputing solution for generative AI, incorporating direct liquid cooling (DLC), aligns with HPE’s focus on energy efficiency.