The 10 fastest supercomputers on Earth

China retakes the top spot for first time since 2010

Maximum speed

The latest twice-yearly Top 500 listing of the fastest supercomputers in the world is out, and this time the winner is the monolithic Tianhe-2, built by the Chinese government. Here’s a walkthrough of the 10 mightiest computing machines out there.

Tianhe-1A
Credit: Nvidia
Tianhe-1A

Fittingly, China’s last top dog comes in at number 10 on the latest list. Its 2.5 petaflop performance is powered by a Linux-based proprietary interconnect running 186,386 processor cores. It’s the least energy-efficient machine in the top 10, however, producing just 635 megaflops per watt.

SuperMUC
Credit: LRZ.de
SuperMUC

The German system drops three places in this list, having peaked on its debut of fourth place one year ago. This IBM-built machine was measured at 2.89 petaflops produced by 147,456 cores.

Vulcan
Credit: Lawrence Livermore National Laboratory
Vulcan

It was recently announced that the Vulcan supercomputer, operated by Lawrence Livermore National Laboratory, is available for use on a contract basis. So if you have a need for 4.29 petaflops of processing power, provided by 393,216 cores, you’re in luck. (Vulcan is also the most energy-efficient machine in the top 10, at 2177 megaflops per watt.)

JUQUEEN
Credit: Forschungzentrum Jülich
JUQUEEN

JUQUEEN, based at the Forschungzentrum Jülich research center in Germany, has been built out from 131,072 cores as of a year ago to 458,752 in the latest list, and its top speed has increased commensurately. It’s currently sitting just a hair north of 5 petaflops.

Stampede
Credit: University of Texas/Texas Advanced Computing Center
Stampede

The University of Texas’ Stampede clocked in at 5.16 petaflops, just edging out JUQUEEN. The most powerful machine run by an academic institution in the top 10, Stampede supplements the 462,462 main processor cores with 366,366 of Intel’s Xeon Phi SE accelerator cores for extra speed.

Mira
Credit: Argonne National Laboratories
Mira

One of four entries in the top 10 owned by the U.S. Department of Energy (along with numbers 8, 3, and 2), the Argonne National Labs’ Mira clocks in at 8.58 petaflops of measured performance thanks to 786,432 total processing cores.

K Computer
Credit: RIKEN
K Computer

The lone SPARC64-powered entry in the top 10, the Fujitsu-built K Computer at the RIKEN Advanced Institute for Computational Sciences in Japan is powered by 705,024 cores, achieving a speed of 10.5 petaflops. The K Computer is a long-standing presence on the top 10 list, having taken the top spot in both 2011 lists and remained in the top five ever since.

Sequoia
Credit: Lawrence Livermore National Laboratory
Sequoia

Another entry from the Livermore Labs, Sequoia packs in a total of 1,572,864 cores, which provide 17.1 petaflops of performance. It was the fastest supercomputer in the world as of two lists ago, but has slipped a place in both subsequent rankings.

 

Titan
Credit: Oak Ridge National Laboratory
Titan

The last list’s top dog, the Oak Ridge National Lab’s Titan is unique among the top 10 in its use of 261,632 Nvidia K20X-model GPU cores for acceleration. Even when that’s combined with 560,640 main cores, Titan still uses far fewer total processor cores than rival Sequoia to produce 17.6 petaflops of power.

Tianhe-2
Credit: Jack Dongarra, via Ars Technica
Tianhe-2

China’s latest supercomputer is a monster: Powered by a whopping 3,120,000 processor cores and 2,736,000 accelerators, Tianhe-2 nearly doubled Titan’s best performance, clocking in at fully 33.8 petaflops.

Email Jon Gold at jgold@nww.com and follow him on Twitter at @NWWJonGold.