HPE to Build Exascale Supercomputer “Discovery” and AI Cluster “Lux”

Introducing HPE Cray Supercomputing GX5000 that includes the HPE Cray Supercomputing Storage Systems K3000 to allow supercomputing breakthroughs within the converged AI and HPC period

In this text

• HPE-built Discovery will bolster productiveness up to 10X and unlock new scientific horizons in precision drugs, most cancers analysis, nuclear vitality and aerospace
• The Lux system, to be constructed by HPE, will provide a versatile, multi-tenant AI cloud platform to help coaching and inference
• Discovery will characteristic the brand new HPE Cray Supercomputing GX5000, a subsequent technology supercomputer constructed for the converged AI and HPC period, to drive better efficiency and productiveness for the lab’s physics-based modeling, simulation and data-driven AI fashions, in addition to offering testbed capabilities for quantum computing
• The new HPE Cray Supercomputing Storage Systems K3000, an possibility for the HPE Cray Supercomputing GX5000, is the business’s first factory-built storage system with embedded Distributed Asynchronous Object Storage (DAOS) open supply software program

HPE (NYSE:HPE) right now introduced it was chosen to construct two techniques for the U.S. Department of Energy’s (DOE) Oak Ridge National Laboratory (ORNL) as a part of the DOE’s mission to advance American management in synthetic intelligence (AI) and supercomputing that help science, vitality and nationwide safety. The new techniques embody a second technology exascale supercomputer, “Discovery,” which is the successor to ORNL’s Frontier – an HPE-built system that broke the exascale velocity barrier – and a brand new AI cluster “Lux” that can help the DOE’s initiatives in advancing AI and machine studying with a multi-tenant cloud-like platform.

Discovery can be primarily based on the brand new HPE Cray Supercomputing GX5000, HPE’s next-generation supercomputing platform for management class techniques that leverages a unified AI and excessive efficiency computing (HPC) structure to streamline operations site-wide and throughout distributed clusters. It can be augmented by a brand new DAOS-based HPE Cray Supercomputing Storage Systems K3000, a storage possibility for HPE Cray Supercomputing GX5000. Discovery will ship new capabilities for AI, HPC and quantum computing and is anticipated to enhance choose utility productiveness tenfold¹, enabling scientists to speed up breakthroughs in areas similar to precision drugs, most cancers analysis, nuclear vitality and aerospace.

“When we constructed Frontier for Oak Ridge National Laboratory and ushered in exascale, we achieved the head in supercomputing historical past and a triumph for the U.S.,” mentioned Antonio Neri, president and CEO at HPE. “We are proud to construct on that management innovation and sturdy public-private partnership with the U.S. Department of Energy, ORNL and AMD, to construct Discovery and Lux, accelerating the subsequent period of scientific discovery and AI innovation.”

Lux can be a devoted AI system primarily based on the direct liquid-cooled HPE ProLiant Compute XD685 and characteristic AMD Instinct MI355X GPUs, AMD EPYC CPUs and AMD Pensando networking. Designed to bolster entry to AI sources, Lux will present researchers throughout the U.S. with cloud-like entry to a sovereign AI manufacturing unit particularly resourced for coaching and inference.

Discovery will elevate the exascale computing capabilities first developed for the HPE-built Frontier supercomputer at ORNL. As a end result, Discovery will unlock new scientific horizons in varied scientific fields whereas advancing the lab’s mission of innovation and safety.

“We are excited for Discovery and Lux to increase the science that researchers are ready to do at Oak Ridge,” mentioned Bronson Messer, Director of Science for the Oak Ridge Leadership Computing Facility. “Discovery will set the stage for a brand new degree of converged HPC, AI and quantum computing capabilities, offering extra perception in reference to different techniques, whereas Lux tremendously expands researcher entry to devoted AI sources. As a end result, we anticipate each techniques will contribute to a paradigm shift in our productiveness, reaching unparalleled good points in varied, important areas of scientific analysis and management.”

“For greater than a decade, AMD and HPE have partnered to push the bounds of high-performance computing, delivering options that allow discoveries and change the world,” mentioned Dr. Lisa Su, chair and CEO, AMD. “Together with Oak Ridge National Laboratory, we’re advancing the subsequent technology of AI techniques with Discovery and Lux—empowering researchers to speed up innovation and strengthen America’s management in science and expertise.”

Inside Discovery: the next-generation exascale supercomputer

Discovery’s scientific developments will stem from using the HPE Cray Supercomputing GX5000 unveiled right now. Building upon 50 years of supercomputing innovation relationship again to the Cray-1 introduced in 1975, HPE has designed its next-generation infrastructure for supercomputing within the converged AI and HPC period.

The HPE Cray Supercomputing GX5000 is purpose-built for exascale and options state-of-the-art end-to-end capabilities throughout CPUs, GPUs, accelerators, networking, software program, storage and liquid cooling. By leveraging the brand new structure, Discovery will ship:

• Greater efficiency with optimized house – The new platform is purpose-built to scale to exascale efficiency with better density in contrast to the earlier model², utilizing 25 p.c much less information heart house per rack.
• High efficiency interconnect with HPE Slingshot – The subsequent technology HPE Slingshot gives Discovery a contemporary, high-performance interconnect to ship high-bandwidth and low-latency for HPC, machine studying and analytics purposes.
• Industry-first HPC DAOS storage efficiency³ – Augmented by the brand new HPE Cray Supercomputing Storage Systems K3000, Discovery can have 300 p.c⁴ extra enter/output operations per second (IOPS) per storage rack in contrast to Frontier to allow AI purposes to run with larger productiveness. As the business’s first factory-built storage system with embedded Distributed Asynchronous Object Storage (DAOS) open supply software program, the HPE Cray Supercomputing Storage Systems K3000 is an economical, all-flash storage system that enhances the Lustre file system-based HPE Cray Supercomputing Storage Systems E2000, which will even be featured in Discovery.
• Next-generation, liquid-cooled and accelerated compute – Discovery will characteristic next-generation AMD EPYC processors, codenamed “Venice,” with AMD Instinct MI430X GPUs, which supply superior efficiency and accuracy for modeling, simulation and AI initiatives. Leveraging HPE’s 50 years⁵ of liquid cooling innovation, Discovery’s compute infrastructure can be totally liquid-cooled to optimize vitality effectivity and cost-effectiveness in supercomputing environments.

As a world chief in supercomputing⁶, HPE delivers end-to-end options and companies to prospects with best-in-class AI and HPC experience. As an integral accomplice, HPE supercomputing companies assist improve outcomes by way of a completely unified administration method of a company’s infrastructure and purposes with a key give attention to core enterprise wants and steady innovation.

The publish HPE to Build Exascale Supercomputer “Discovery” and AI Cluster “Lux” first appeared on AI-Tech Park.