Energy plans delivery of first exascale supercomputer
- By Chase Gunter
- Mar 19, 2019
The Argonne National Laboratory will soon be home to the nation's first exascale computer capable of 1 exaflop performance -- 1 quintillion calculations per second. Aurora is expected to be delivered in 2021 under a contract to Intel Corp. and Cray worth more than $500 million.
The Department of Energy expects to use the system to develop massive-scale space simulations, model drug response predictions and improve extreme weather forecasting, among other scientific problems.
First announced in 2015, Aurora was envisioned as a 180 petaflop supercomputer. Those plans were revised in 2017 when DOE announced it was devoting $258 million to research the hardware, software and applications necessary to achieve exascale computing.
Aurora will feature new technologies Intel has designed specifically for exascale-based artificial intelligence applications, including an Intel Xeon Scalable processor, Optane DC Persistent Memory, Xe compute architecture and One API software. It will use Cray’s next-generation Shasta family of supercomputers, which feature the company's scalable, high-performance network, codenamed Slingshot.
Argonne is already a supercomputing powerhouse. It is home to Mira, a 10-petaflop IBM Blue Gene/Q system capable of 10 quadrillion calculations per second, and Theta, an 11.69-petaflops Intel/Cray system.
“This platform is designed to tackle the largest AI training and inference problems that we know about," said Argonne's Associate Lab Director Rick Stevens.
"What excites me the most about exascale is the fact that we have in one environment the ability to mix simulation and artificial intelligence," Stevens said. "When we have this integration of big data, of data-driven science, streaming data and simulation -- that's where science is going to happen. That is where breakthroughs will happen."
Energy Secretary Rick Perry said the field of AI and supercomputing is one "we cannot fall behind in."
“That's why we have committed to building three new exascale machines which will be 10 to 20 times faster than the fastest machines operating today."
The announcement comes amid a global arms race for supercomputing power. China especially has invested heavily in emerging technology and supercomputing and, by comparison, has plans to build an exascale-level computer by 2020.
“This is a big deal,” said Stephen Ezell, vice-president of global innovation policy at the Information Technology and Innovation Foundation, noting that the development of an exaflop computer has enormous implications for national security, cryptography and in the commercial sector.
"It signals that the United States is keeping pace internationally being at the leading edge of development of the world’s fastest computing systems," Ezell said.
The commitment to exascale represents another step in the government's focus on emerging and future computing powers.
In April 2018, Energy put out a request for proposals for two other exascale computer systems to be deployed by 2023, one at Oak Ridge and another at Lawrence Livermore National Laboratory in Livermore, Calif. The two could be worth as much as $1.8 billion.
The White House’s budget for fiscal year 2020 requested $500 million for exascale computing.
This article was first posted to FCW, a sibling site to GCN.
Chase Gunter is a staff writer covering civilian agencies, workforce issues, health IT, open data and innovation.
Prior to joining FCW, Gunter reported for the C-Ville Weekly in Charlottesville, Va., and served as a college sports beat writer for the South Boston (Va.) News and Record. He started at FCW as an editorial fellow before joining the team full-time as a reporter.
Gunter is a graduate of the University of Virginia, where his emphases were English, history and media studies.
Click here for previous articles by Gunter, or connect with him on Twitter: @WChaseGunter