Since 1993, the TOP500 project has been publishing a semi-annual list of the 500 most powerful computer systems in the world, as a barometer of trends and accomplishments in high-performance computing. The systems are ranked based on their speed in floating-point operations per second (FLOP/s), measured on the LINPACK benchmark, which involves the solution of a dense system of linear equations.
This fall’s edition of the list has recently been announced, in conjunction with the SC11 supercomputer conference being held this week in Seattle. The fastest system is still the Japanese K Computer.
Japan’s “K Computer” maintained its position atop the newest edition of the TOP500 List of the world’s most powerful supercomputers, thanks to a full build-out that makes it four times as powerful as its nearest competitor. Installed at the RIKEN Advanced Institute for Computational Science (AICS) in Kobe, Japan, the K Computer it achieved an impressive 10.51 Petaflop/s on the Linpack benchmark using 705,024 SPARC64 processing cores.
The K Computer is the first system to surpass 10 petaflops (1.0 × 1016 floating point operations per second). Unlike many large systems, it does not use graphics processors or other special chips as accelerators; it is also one of the more energy-efficient systems on the list.
The second place system is still the Chinese Tianhe-1A system with 2.57 Petaflops. In fact, the top ten systems have remained in place from June’s ranking. The entry level for the Top 500 has gone up, though.
In the latest list, the level to the list moved up to the 50.9 Teraflop/s mark on the Linpack benchmark, compared to 39.1 Teraflop/s six months ago.
The total performance of all systems on the Top 500 list is now 74.2 petaflops, compared to 58.7 petaflops in the previous survey.
Intel is the leading supplier of processors for these systems; its processors are used in 384 systems. AMD Opteron processors are used in 63 systems, and IBM POWER processors in 49 systems. Graphics processors, principally from NVIDIA, are used to accelerate computations in 39 systems.
As one might expect, the K Computer, now fully built out, uses a good deal of electricity, 12.66 megawatts; however, it is one of the most efficient systems, delivering 830 megaflops per watt. The average system’s efficiency is 282 megaflops / watt, up from 248 megaflops / watt in the last survey. The most efficient systems are the IBM BlueGene/Q systems, at 2,029 megaflops / watt.
As has been true for some time, the distribution of operating systems used is rather different from that in the desktop computing market:
| OS Family | Number | % of Capacity |
|---|---|---|
| Linux | 457 | 91.4 |
| Unix | 30 | 6.0 |
| BSD-based | 1 | 0.2 |
| Windows | 1 | 0.2 |
| Mixed | 11 | 2.2 |
In the early days of computing, supercomputers often used highly specialized components and software, and were built in very small numbers. Today, although the complete systems are still customized for their intended use, the use of commodity processors and open source software has become the norm. It is interesting, for example, that the number 42 system on the list is Amazon’s EC2 cluster. Who would have guessed, even twenty years ago, that a bookseller would have more computing horsepower than Sandia National Laboratory (at number 50)?
Random Walks 