Roadrunner runs away with supercomputing prize

Roadrunner has topped the Top500 supercomputers released yesterday at the International Supercomputing Conference in Dresden, Germany.

Roadrunner, the world's fastest supercomputer, can perform one thousand trillion calculations per second.

Twice yearly, the list measures the 500 most powerful computer systems available commercially. This year - the 31st time the list has been put together - the honour of top supercomputer goes to IBM's Roadrunner, which is housed at the US Department of Energy's Los Alamos National Laboratory. It's the first system to reach 1.026 petaflops - one petaflop is equal to a quadrillion, or one thousand trillion, calculations per second.

Last year's most powerful computer, Lawrence Livermore National Laboratory's BlueGene/L - also made by IBM - reached 208.6 teraflops. This year that computer ranked number two, reaching a maximum processing speed of 478.2 teraflops.

The fastest supercomputer in the world - used to monitor the US nuclear weapons stockpile - is based on the IBM QS22 blades, which are built using advanced versions of the Cell processor in Sony's PlayStation 3.

IBM, which continues its dominance of supercomputing, makes 210 of the 500 systems, including five of the top 10. HP is close behind, however. HP makes 183 of the fastest computers, including the number eight fastest system known as EKA, located in Computational Research Laboratories' data centre in Pune, India.

Rounding out the top 10 is Sun Microsystem's Ranger at number four, Cray's Jaguar at number five, SGI's Encanto at number seven, and SGI's Altix at number 10.

Supercomputing, which pits the highest-end machines against challenges such as forecasting the global climate in coming decades or finding oil reservoirs underground, is a fast-changing field. The Top500 list once again had the most turnover compared with the preceding list, according to the researchers who compile it.

The main measurement used in compiling the list is the Linpack measurement, which puts each system through its paces by having to solve a dense system of linear equations.

The Top500 acknowledges Linpack isn't a complete test of system performance but it's a way to test for performance on a similar problem across each system. The need for a more complete benchmarking system has been under discussion for several years.