Hewlett Packard Enterprise (HPE) has powered up the latest in its AI-capable supercomputers, Aurora.
Aurora is the second exascale system delivered by the company. Having reached 1.012 exaflops on 87% of the system, is the second fastest in the world after Frontier – also created by HPE – according to Top500.
Built in collaboration with Intel for the United States Department of Energy’s Argonne National Laboratory, the system uses HPE Cray EX infrastructure and features 75,000 compute node endpoints, 2,400 storage and service network endpoints, and 5,600 switches. The 10,624 compute blades that make up Aurora also feature 21,248 Intel Xeon CPU Max Series processors and 63,744 Intel Data Center GPU Max units.
According to HPE, researchers using Aurora will be able to run generative AI models to “accelerate scientific discovery”. Early experiments run on the system include brain mapping, developing methods for detecting neutrino interactions, and fusion reactor simulations.
Rick Stevens, associate laboratory director and distinguished fellow at Argonne National Laboratory, described Aurora as a "first of its kind supercomputer" that will be a gamechanger for researchers.
"Reaching this milestone with a second exascale system in the US is an incredibly significant achievement that will advance open science initiatives globally." he said.
For HPE, AI and HPC are inseparable
High-performance computing (HPC) has been an important part of HPE’s portfolio since its acquisition of SGI almost 10 years ago and Cray in 2019. It was only last year, however, that the company really started to put it front and center in its public announcements.
This shift has been spurred at least in part by the rise of generative AI and the company’s move to embrace it. At HPE Discover 2023, held in June last year, CEO Antonio Neri revealed HPE’s new focus, telling attendees that it had moved from being the "edge to cloud" company to "edge to cloud to AI".
This direction of travel was underscored at the time by the release of HPE GreenLake for Large Language Models, an HPC-powered infrastructure offering that allows customers to train and deploy large-scale AI projects without having to invest in their own specialist hardware at extremely high cost.
Subsequently, it has built several AI-focused HPC offerings, including Isambard-AI, the UK’s most powerful supercomputer, Helios, Poland’s fastest system, Venado in New Mexico, and now Aurora.
