The Chinese supercomputer Tihane-2 still tops the list of the most powerful supercomputers in the world.
The latest Top500 list was released today, and the Tianhe-2, developed at China’s National University of Defense Technology, was the fastest system on the list by far, reaching speeds of 33.86 petaflop per second -- quadrillions of calculations per second -- on the Linpack benchmark. It initially topped the second fastest machine -- the U.S.-based Titan -- in the last set of rankings, released in June.
This time around, the Titan -- a Cray XK7 system installed at the Department of Energy’s (DOE) Oak Ridge National Laboratory -- held on to second place with 17.59 Pflop/s worth of computing power. Sequoia, an IBM BlueGene/Q system installed at DOE’s Lawrence Livermore National Laboratory, came in third with 17.17 Pflop/s.
Although the Chinese are still sitting in the top spot, the U.S. is the leading consumer of supercomputers, with 265 of the top 500 systems. That's up from 253 last time. China, meanwhile, has edged out Japan as the second largest consumer of supercomputers.
The only new entry to the top 10 was the Piz Daint, a Cray XC30 system installed at the Swiss National Supercomputing Centre (CSCS) in Lugano, Switzerland. At 6.27 Pflop/s, it's Europe's most powerful system. It also happens to be the most energy efficient supercomputer in the top 10.
Intel is still the top provider the processors, powering 82.4 percent of the TOP500 systems. The number of systems using accelerator or co-processing technology, such as graphics processors, remained steady at 53 systems. Of these, 28 use NVIDIA chips, two use ATI Radeon, and 13 systems use Intel MIC technology.
The TOP500 list began in 1993. It's compiled by Hans Meuer of the University of Mannheim, Germany, Erich Strohmaier, and Horst Simon of Lawrence Berkeley National Laboratory, and Jack Dongarra of the University of Tennessee, Knoxville.
One possible change in the supercomputing industry is the growth of open source, scale-out solutions. IBM's Watson, the computer that won at Jeopardy, is powered by the open source platform Hadoop. While supercomputers have traditionally stuffed increasing numbers of increasingly fast processors into one system, Hadoop distributes problems across a large number of cheaper, less powerful servers. In other words, it scales out instead of up, and it's a popular choice for a growing number of computing needs. Other high performance computing companies, such as Cray and SGI, are starting to use Hadoop as well.
But supercomputing research continues apace. Earlier this year, Stanford University researchers broke a new record in supercomputing with a system that performed a calculation across one million cores. The largest known Hadoop cluster, by contrast, had just 8,800 cores.
