U.S. Takes Supercomputer Crown with Sequoia

The United States has apparently reclaimed the top spot on the influential Top500 list of the world’s fastest supercomputers.

The top supercomputer, known as “Sequoia,” is an IBM BlueGene/Q system installed in the Department of Energy’s Livermore National Laboratory in Livermore, Calif. Thanks to its 1,572,864 processing cores, it can deliver 16.32 sustained petaflops (that’s a quadrillion floating point operations per second) on the Linpack benchmark, a performance yardstick based on the ability to solve a dense system of linear equations.

That’s a ton of computing power, bent toward a very specific task: simulating nuclear weapons tests. Specifically, Sequoia can offer researchers an extremely granular view of how materials will react at extreme pressures and temperatures. It can also effectively perform what’s known as uncertainty quantification calculations, enabling nuclear stewards to more effectively anticipate potential issues as the nation’s weapons stockpile continues to age.

“Computing platforms like Sequoia help the United States keep its nuclear stockpile safe, secure, and effective without the need for underground testing,” National Nuclear Security Administration (NNSA) administrator Thomas D’Agostino wrote in a statement. “While Sequoia may be the fastest, the underlying computing capabilities it provides give us increased confidence in the nation’s nuclear deterrent as the weapons stockpile changes under treaty agreements.”

Coming in at second on the latest Top500 list is Fujitsu’s “K Computer,” installed at the RIKEN Advanced Institute for Computational Science (AICS) in Kobe, Japan. Ranked the world’s top supercomputer on the two previous Top500 lists, the K Computer can leverage its 705,024 cores to deliver 10.51 petaflops on the Linpack benchmark.

Third on the list is another IBM BlueGene/Q system, this one based at the Argonne National Laboratory in Illinois. That “Mira” supercomputer, relying on 786,432 cores, is capable of 8.15 sustained petaflops on the Linpack benchmark.

The “SuperMUC” supercomputer, installed at Leibniz Rechenzentrum in Germany, came in fourth on the list, followed by the “Tianhe-1A” at the National Supercomputing Center in Tianjin. The full list can be found on the Top500 Website.

With regard to manufacturers, IBM kept the overall lead with 213 systems on the list (42.6 percent), followed by Hewlett-Packard with 138 systems (27.6 percent), Cray with 5.4 percent, Appro with 3.6 percent, and SGI and Bull tied with 3.2 percent each. Combined, the companies’ systems delivered some 123.4 petaflops’ worth of performance—up considerably from the November 2011 edition of the list, when total performance hit 74.2 petaflops.

Processing Power

In terms of hardware, some 74.4 percent of the systems on the list rely on Intel processors, a dip from 76.8 percent on the previous edition of the list. AMD Opteron processors came in second with 12.6 percent of the list, basically unchanged, while the number of supercomputer systems relying on IBM Power processors increased slightly to 11.6 percent.

Some 58 systems used accelerators or co-processors, up from 39 systems on the previous list. Of those, 53 relied on Nvidia, two on ATI Radeon, two on Cell processors, and one on Intel MIC (Many Integrated Core) architecture.

The Top500 list, issued twice a year, is currently on its 39th edition. It is assembled by Hans Meuer of the University of Mannheim, Germany, along with Erich Strohmaier and Horst Simon of Lawrence Berkeley National Laboratory, and Jack Dongarra of the Universirt of Tennessee, Knoxville.


Image: Dario Lo Presti/Shutterstock.com