Exascale computing by the numbers

Exascale computing by the numbers

Mind-bending amounts of data, converging workloads, and the digital transformation race. These are some of the things driving the need for exascale computing.

Exascale computing will benefit the world through things like personalized medicine, manufacturing innovation, and space, nuclear, and climate research.

October 18 is the first annual Exascale Day. To celebrate, we’ve assembled a short timeline of how we got to where we are today—where a quintillion computations per second is possible. Plus, we’ve highlighted some related reading to get you up to speed.

Join the technologists who are on point for the first three exascale systems

Six experts discuss all things exascale in this one-hour webcast and panel discussion. Register, and submit your questions today.

When: 11 a.m. to 12 p.m. ET, Friday, October 18

Panel: Doug Kothe, ECP; Steve Scott, Cray; Rick Stevens, ANL; Jeff Nichols, ORNL; Michel McCoy, LLNL; Earl Joseph (moderator), Hyperion Research

Exascale computing by the numbers

1 exaFLOPS—A quintillion (a billion billion) floating-point operations per second (10¹⁸FLOPS).

6—Number of labs involved in the Exascale Computing Project.

6—Number of U.S. technology companies that received funding from the U.S. Department of Energy’s Exascale Computing Project as part of its new PathForward program, accelerating the research necessary to deploy the nation’s first exascale supercomputers.

2008—PetaFLOPS barrier was broken (10¹⁵floating-point operations per second).

March 18, 2019—The U.S. DOE announced Intel and Cray will build Aurora, the first exascale supercomputer in the U.S. It will be delivered to Argonne National Laboratory in 2021.

May 7, 2019—The U.S. DOE announced AMD and Cray will build the Frontier supercomputer for the Oak Ridge National Laboratory. Frontier is expected to go online in 2021 with 1.5 exaFLOPS of processing power.

Aug. 14, 2019—The U.S. DOE, the National Nuclear Security Administration, and Lawrence Livermore National Laboratory announced the signing of contracts with Cray to build the NNSA’s first exascale supercomputer, El Capitan. It will have a peak performance of more than 1.5 exaFLOPS, with anticipated delivery in late 2022. The total contract award is valued at $600 million.

18—Number of companies on the Exascale Computing Project’s Industry Council.

1972—Cray Research was founded.

1976—First commercial supercomputer, Cray®-1, was installed at Los Alamos National Laboratory. It offered 160 megaFLOPS and cost $8.8 million.

~5—Number of years the Cray-1, first shipped in 1976, was the fastest computer in the world.

~1—Number of years it took to assemble a Cray-1, each of which were hand-wired.

4 feet—Length of the longest wire in the Cray-1, with a total of 60-plus miles of wire snaking through it.

256,000,000—Number of 8-byte words of memory (2 GB of memory) supported by the Cray C90, introduced in 1991. Editor’s note: In 1991, that was a very significant amount of memory.

$10 billion—What the major contenders in the exascale race—the U.S., Europe, Japan, and China—together may spend between now and 2022 on developing and purchasing early exascale supercomputers. This is a projection from Hyperion Research based largely on public information.