When the 36th edition of the TOP500 Supercomputer Sites was released last November, two industry trends were confirmed: China’s new leadership in the industry, and the impressive performance gains registered by the adoption of GPUs (Graphics Processing Units). The Chinese Tianhe-1A, seen below, is housed at the National Supercomputer Center in Tianjin. Achieving a performance level of 2.57 petaFLOPS (quadrillions of FLoating point OPerations per Second) it took the #1 spot.
Tianhe-1A Supercomputer at the National Supercomputing Center in Tianjin, China
Besides grabbing two of the top five positions on the list, China now claims 42 systems in the TOP500, moving past past Japan, France, Germany and the UK to become the number two country behind the United States, which remains the overall leader with 282 machines in the top 500.
The former number one system, seen below, is the visually stunning Cray XT5 “Jaguar” located at the U.S. Department of Energy’s (DOE) Oak Ridge Leadership Computing Facility. Achieving a level of 1.75 petaFLOPS, it was still bested from a performance standpoint by 47%.
Cray XT5 Jaguar Supercomputer located at Oak Ridge Leadership Computing Facility (OLCF)
The Rise of GPGPU Supercomputing
Three of the top five supercomputers on the list, the Tianhe-1A, Nebulae and Tsubame 2.0, are using NVIDIA GPUs. In all, 17 systems on the TOP500 use non-CPU hardware accelerators, with ten of them using NVIDIA GPUs, six using the Cell processor, and one using ATI Radeon GPUs. This system trend represents a dramatic shift from supercomputing’s CPU-centric, massive parallel processing roots.
China has clearly tied its supercomputing strategy to GPU Computing and has stated that it wants to be the first nation to build an exascale machine. To that end they’ve been developing proprietary network interconnect and I/O processor technology to maximize system performance. But the United States has not been standing still, as IBM announced plans to build a 750,000 core supercomputer for the Department of Energy’s Argonne National Laboratory and expects performance of 10 petaFLOPS.
The Top Five Supercomputers
| Tianhe-1A | 2.57 petaFLOPS | China |
| Cray XT5 “Jaguar” | 1.75 petaFLOPS | United States |
| Nebulae | 1.27 petaFLOPS | China |
| Tsubame 2.0 | 1.19 petaFLOPS | Japan |
| Cray XE6 “Hopper” | 1.05 petaFLOPS | United States |
Out of the top 10 systems, five are in the United States, two in China and one each in Japan, France and Germany. The most powerful supercomputer within Europe is a Bull system at the French CEA (Commissariat à l’énergie atomique et aux énergies alternatives or Atomic and Alternative Energies Commission) that is now ranked in sixth place.
While Trenton is not building supercomputers yet, our membership in the NVIDIA Tesla Preferred Partner program lays the foundation allows us to create customer-driven, long-life GPU Computing solutions that will satisfy computationally intensive applications within Government & Defense, Video Processing and Medical Research markets.
{ 2 trackbacks }