BREAKING: Argonne, Intel to Build First U.S. Exascale Supercomputer
The Department of Energy has picked Argonne National Laboratory and its industry partner Intel Corp. to build the United States’ first exascale supercomputer by 2021, officials announced March 18.
The system — known as Aurora — is slated for completion in 2021, said Paul Dabbar, undersecretary for science at the Energy Department.
Aurora will jump “to the top performance computer of the world,” he told reporters during a media call before the official announcement. “It will hit an exaflop of performance — so that is 10 to the 18th [power], or a billion-billion calculations per second.”
The system will push forward the bounds of human knowledge and discovery science, he said. The contract is valued at more than $500 million and work will be performed by Intel with its subcontractor Cray Computing, he added.
Aurora is the next step in a long history of the Energy Department and U.S. national labs' pursuit of high-performance computing, he noted. Just last year, the department unveiled the Summit supercomputer that — with a peak performance of 200,000 petaflops — surpassed China’s Sunway TaihuLight in capability. Summit is based at Oak Ridge National Laboratory.
The Aurora supercomputer will be used for a variety of purposes, said Energy Secretary Rick Perry.
“Achieving exascale is imperative not only to better the scientific community, but also to better the lives of everyday Americans,” he said in a statement. “Aurora and the next-generation of exascale supercomputers will apply HPC and AI technologies to areas such as cancer research, climate modeling and veterans’ health treatments. The innovative advancements that will be made with exascale will have an incredibly significant impact on our society.”
Rick Stevens, Argonne’s associate laboratory director for computing, environment and life science, said the system will be optimized for big data analytics and deep learning. It will be particularly beneficial for “the kinds of streaming data problems that we have in DoE where we have data coming off of accelerators and detectors, telescopes and so forth,” he added.
Dabbar noted that the system will have national security and defense applications. For example, the United States no longer actively tests nuclear warheads, but instead relies on simulations to do so.
“High-performance computing [is used] to fill the gap for those analyses,” he said. “This capability allows us to continue [that] and to accelerate our understanding of those needs.”
Stevens noted that Aurora will also be available to U.S. universities and industry.