HPE Selected to Build Next-Gen Supercomputer and AI Cluster for Oak Ridge National Laboratory
In a significant advancement for the fields of artificial intelligence (AI) and supercomputing, Hewlett Packard Enterprise Co. (HPE) has announced that it has been selected to construct two groundbreaking systems for the U.S. Department of Energy’s (DOE) Oak Ridge National Laboratory (ORNL). The systems include a next-generation exascale supercomputer named “Discovery” and a dedicated AI cluster called “Lux.” This initiative is part of the DOE’s mission to reinforce American leadership in science, energy, and national security.
1. Discovery: The Next-Generation Exascale Supercomputer
Building on a Legacy
“Discovery” will succeed ORNL’s Frontier, an HPE-built system that made history by breaking the exascale speed barrier. This new supercomputer will be based on the HPE Cray Supercomputing GX5000, which integrates a unified AI and high-performance computing (HPC) architecture, enabling streamlined operations across distributed clusters. Complementing this, Discovery will utilize the new DAOS-based HPE Cray Supercomputing Storage Systems K3000, further enhancing its capabilities.
Enhanced Performance and Applications
Discovery is projected to increase the productivity of select applications tenfold, revolutionizing research in critical domains such as precision medicine, cancer research, nuclear energy, and aerospace. HPE CEO Antonio Neri expressed pride in continuing a strong public-private partnership with the U.S. Department of Energy, ORNL, and AMD, underscoring the project’s potential to accelerate scientific discovery and AI innovation.
2. Lux: A Dedicated AI Cluster
Tailored for AI Advancement
The AI cluster “Lux” will be a dedicated system designed to bolster access to AI resources for researchers across the U.S. It will employ direct liquid-cooled HPE ProLiant Compute XD685, powered by AMD Instinct MI355X GPUs and AMD EPYC™ CPUs, along with AMD Pensando™ networking. This configuration aims to provide a cloud-like environment for training and inference, effectively creating a sovereign AI factory.
Expanding Research Access
Bronson Messer, Director of Science for the Oak Ridge Leadership Computing Facility, highlighted the transformative potential of both systems. He noted that Discovery will elevate the integration of HPC, AI, and quantum computing capabilities, while Lux will significantly enhance researcher access to essential AI resources.
3. Technical Innovations in Discovery
Unveiling the HPE Cray Supercomputing GX5000
The HPE Cray Supercomputing GX5000, unveiled as part of this initiative, represents the culmination of over 50 years of supercomputing innovation. This next-generation infrastructure is purpose-built for exascale performance and boasts several key features:
- Optimized Space Utilization: The GX5000 is designed to occupy 25% less data center space per rack compared to its predecessor, while still achieving greater density.
- High-Performance Interconnect: The system will utilize the advanced HPE Slingshot interconnect, providing high bandwidth and low latency essential for HPC, machine learning, and analytics applications.
- Industry-Leading DAOS Storage Performance: The new storage system promises to deliver 300% more input/output operations per second (IOPS) compared to Frontier, enabling AI applications to run with higher efficiency.
Liquid Cooling and Advanced Compute
Discovery will feature next-generation AMD EPYC processors, codenamed “Venice,” alongside AMD Instinct MI430X GPUs. Utilizing HPE’s extensive experience in liquid cooling, Discovery’s entire compute infrastructure will be fully liquid-cooled, optimizing energy efficiency and reducing operational costs.
4. A Collaborative Vision for the Future
Dr. Lisa Su, CEO of AMD, emphasized the long-standing partnership between AMD and HPE, noting their collective efforts to push the boundaries of high-performance computing. With the introduction of Discovery and Lux, they aim to empower researchers and bolster America’s position in science and technology.
5. Conclusion
The formation of these advanced systems not only underscores HPE’s leadership in supercomputing but also represents a critical step forward in harnessing technology for scientific advancement. As these projects come to fruition, they are poised to impact various fields and enhance the capabilities of researchers at Oak Ridge National Laboratory and beyond. HPE continues to pioneer the integration of AI and supercomputing, further solidifying its role as a leader in enterprise technology solutions.
