China's Tianhe-2 Still the World's Fastest Supercomputer
There isn't much change at the top of the newest list of the world's fastest supercomputers.
According to the Top500 list released Nov. 18 at the SC '13 supercomputer show in Denver, the five fastest systems in the world are the same as when the last list was released in June, with China's massive Tianhe-2 "Milky Way" still holding the top spot. The system, housed at China's National University of Defense Technology, had a performance of 33.86 petaflops per second (quadrillions of calculations per second) and packed more than 3.1 billion Intel Xeon E5-2692 cores. It also leverages Intel's many-core Xeon Phi coprocessors to help speed up performance.
It was the second consecutive list topped by Tianhe-2, which in June overtook Titan, a supercomputer based on Cray's XKY system and housed at the U.S. Department of Energy's Oak Ridge National Laboratory in Tennessee.
Overall, the top 10 saw only one change, with the Piz Daint system at the Swiss National Supercomputing Centre coming into the sixth-place slot. The supercomputer, which is a Cray XC30 system, hit 6.27 petaflops per second (Pflops/s) in performance, making it’s the most powerful system in Europe.
The system also is the most energy-efficient system in the top 10, consuming 2.33 megawatts of power and delivering 2.7 gigaflops per watt of performance.
Piz Daint's entrance into the top 10 helped push China's Tianhe-1, a system at the country's National Supercomputer Center and at one time the world's fastest system, out of the top 10. In the just-released list, the Tianhe-1 is now the 12th-fastest system.
IBM powered half the systems in the top 10, including four using is BlueGene technology. Cray had two systems in the top 10. In addition, there were four systems in the top 10 running IBM's Power processors, and another four powered by Intel's Xeon chips. There was one system running Advanced Micro Device's Opteron processor, and another running SPARC64 chips from Fujitsu.
According to the organizers of the Top500 list, Intel chips powered 82.4 percent of all the systems on the list, and 94 percent of the supercomputers use processors with six or more cores, and 75 percent leverage chips with eight or more.
The use of accelerators and coprocessors continues to draw interest from system makers. In all, 53 systems on the list use either GPU accelerators or coprocessors to help increase the performance of the systems without driving up the power consumption. Thirty-eight of the systems use Nvidia's GPU accelerators, while another use Intel's x86-based Xeon Phi coprocessors, which offer up to 60 cores. Two other system use Advanced Micro Devices' ATI Radeon GPUs.
In the top 10, Tianhe-2 and the number-seven supercomputer—Stampede, at the Texas Advanced Computing Center, which comprises Intel-based Dell systems—use Xeon Phis, while Titan and Piz Daint use Nvidia GPUs.
As an indication in the growth of compute power, the last supercomputer on the list—a cluster developed by Hewlett-Packard for a bank with a performance of 117.8 teraflops per second—was number 363 on the list in June.
The United States still houses the most high-performance computing (HPC) systems, with 265 of the top 500, up from 253 in the June list. The Asia-Pacific region was the second highest—led by China's 63—followed by Europe.