IBM Roadrunner tops supercomputer list

IBM's supercomputers claimed the top three spots

The IBM Roadrunner cluster at Los Alamos Labs has officially become the fastest supercomputer in the world.

The latest edition of the Top 500 supercomputer ranking places the US Department of Energy's machine ahead of IBM's BlueGene L and P systems at the Lawrence Livermore and Argonne National Laboratories.

Roadrunner was formally launched earlier this month. The cluster uses a combination of AMD Opteron and IBM Cell processors to achieve speeds of more than a petaflop, or one thousand trillion operations per second.

Rounding out the top five was Sun Microsystems' Ranger cluster at the University of Texas, followed by Cray's Jaguar system in Oak Ridge National Lab in Tennessee.

Outside the US, the IBM BlueGene/P system at the Forschungszentrum Juelich in Germany ranked sixth overall.

The top supercomputer in the UK was an IBM PowerPC cluster located at the European Centre for Medium-Range Weather Forecasts in Reading. The system ranked 18th overall.

IBM dominated the list with the top three spots and 207 other spots in the rankings. The company estimates the total computing power of all its supercomputers on the list at around 5.6 petaflops.

Also claiming victory was Intel, which had its chips running in 375 of the 500 top systems. The company said that in just 18 months of availability, its quad-core chips had been installed in more than half of the supercomputers on the list.

Microsoft also made an appearance on the list. For the first time, the high-performance version of Windows Server was able to crack the top 25.

The software was running on a Dell PowerEdge system at the National Center for Supercomputing Applications in Illinois that ranked number 23 overall.

Windows Server HPC was also found on the 39-ranked Akka cluster at Sweden's Umea University.

"The systems at Umea University and NSCA demonstrate that Windows can scale to the rarefied atmosphere of the top 25 supercomputing systems in the world, which up to now have relied on dedicated, specialised hardware and software," said Kyril Faenov, general manager of HPC at Microsoft.