Aurora (supercomputer)

Aurora (supercomputer). Aurora is an exascale computing system located at the Argonne National Laboratory, a United States Department of Energy facility in Illinois. As one of the world's most powerful supercomputers, it is designed to enable groundbreaking scientific research across fields such as climate science, materials discovery, and fundamental physics. The system is a cornerstone of the DOE Exascale Computing Project, aimed at maintaining United States leadership in high-performance computing.

Overview

Aurora represents a monumental leap in computational capability, built to surpass the exaflop barrier, which denotes the ability to perform at least one quintillion calculations per second. It is a joint project involving the primary contractor Hewlett Packard Enterprise and key technology partners Intel and Cray Inc.. The supercomputer is installed at the Argonne Leadership Computing Facility, which provides resources to the research community through programs like the Innovative and Novel Computational Impact on Theory and Experiment program. Its deployment marks a significant achievement for the broader Exascale Computing Initiative, a national effort coordinated by the DOE Office of Science.

Development and construction

The development of Aurora was a multi-year endeavor marked by significant technical challenges and evolving specifications. Initially announced in 2015 with a different architecture and timeline, the project was substantially redesigned in 2019 to incorporate then-novel technologies from Intel, including its Xeon Scalable processor family and GPU architectures. The construction was led by Hewlett Packard Enterprise utilizing its Cray EX supercomputer platform, a design also used for other leading systems like Frontier (supercomputer) at Oak Ridge National Laboratory. Major installation phases occurred throughout 2023, with teams from Argonne National Laboratory, Intel, and Hewlett Packard Enterprise collaborating to integrate over 10,000 compute blades and the massive cooling system required for operation.

Technical specifications

Aurora's architecture is based on the Hewlett Packard Enterprise Cray EX platform, which utilizes a Slingshot (interconnect) networking fabric for high-speed data transfer. Each compute node combines one Intel Xeon CPU Max Series processor with six Intel Data Center GPU Max Series units, leveraging a heterogeneous computing model. The full system comprises more than 60,000 GPUs and exceeds 2 exaFLOPS of peak theoretical performance, with a sustained performance on the High Performance Linpack benchmark also expected to be exascale-class. It is supported by a vast hierarchical storage system featuring Solid-state drive technology and is cooled by an advanced, energy-efficient liquid cooling system to manage the immense thermal load.

Scientific applications and research

Aurora is engineered to tackle some of the most complex and data-intensive challenges in modern science. Key research domains include the development of high-fidelity climate models for predicting extreme weather events, the use of artificial intelligence to accelerate the discovery of new pharmaceuticals and battery materials, and large-scale simulations for particle physics experiments related to the Standard Model. Projects will also advance cosmology by simulating the formation of galaxies and support the design of future nuclear fusion reactors. The computational power enables researchers to run unprecedented simulations that integrate vast datasets from facilities like the Large Hadron Collider and the Advanced Photon Source.

Timeline and milestones

The project timeline began with the initial award to Intel and Cray Inc. by the United States Department of Energy in 2015. A major restructuring was announced in 2019, shifting the delivery target and technical approach. The "A21" system, as it was known internally, commenced the installation of its first hardware racks in late 2022. In 2023, the system passed early acceptance tests and began running initial benchmark (computing) codes. Aurora was officially unveiled and entered full production for scientific research in 2024, achieving its exascale performance goal and becoming operational for users of the Argonne Leadership Computing Facility. This milestone positioned it alongside other leading exascale systems like Frontier (supercomputer) and Fugaku (supercomputer) in the global TOP500 rankings.

Category:Supercomputers Category:Argonne National Laboratory Category:Exascale computers