IBM again tops supercomputing list

The biannual list of the world's fastest supercomputers has been announced, with IBM's BlueGene/L System in Livermore, California again taking the top spot - but new entrants, including India's first ever top ten listing, have shaken up the top ten.

The leading project, a joint effort between IBM and the US National Nuclear Security Administration, has lead the TOP500 list since November 2004, has seen some updates and has boosted its speed from 280.6 teraFLOPS a second to 478.2 teraFLOPS/sec - three times the speed of the next fastest.

Despite Livermore's hold on the number one spot, the top ten has seen some changes. The second fastest is a new installation of IBM's BlueGene/P System in Germany, with a speed of 167.3 teraFLOPS/sec. The third fastest is from SGI, and is the first system at the recently opened New Mexico Computing Applications Centre, with a speed of 126.9 teraFLOPS/sec.

The first Indian system to crack the top 10 came in at fourth. The system, a Hewlett-Packard Cluster Platform, is installed at the Computational Research Laboratries in Pune, and features their own routing technology - and 117.9 teraFLOPS/sec performance.

In this latest ranking, IBM has regained the lead from HP. Six months ago, HP had more systems, but IBM is again dominating with 46 per cent of the total, compared to HP's 33 per cent. Multi-core processors were the leading chip architecture, while the biggest growth was in Intel's Clovertown quad core chips - which jumped from 19 to 102 supercomputers in the past six months. Over 70 per cent of the top 500 use Intel processors.

With the top three in the list secured, the US dominates the entire list, with 284 of the 500. The UK is the second rated country by number, with 48, followed by Germany's 31. The UK's top system came in at 17, and is a Cray based at the University of Edinburgh.

The lowest ranking system on the list is now at 5.9 teraFLOPS/sec, compared to 4.0 teraFLOPS/sec in the last list. To make the top 100 list, supercomputers must hit 12.97 teraFLOPS/sec. The total performance of all 500 systems is up to 6.97 petaFLOPS a second from 3.54 petaFLOPS/sec a year ago.

Earlier this year, IBM said it was working on topping a petaFLOP, processing 1,000 trillion calculations each second.

It will be delivering such a computer, nicknamed "Roadrunner", to the US Department of Energy's Los Alamos National Laboratory next summer. Roadrunner is a hybrid which blends PC-style proccessors from AMD with the Cell Broadband Engine, the graphics processor used in the Sony Playstation 3.

Email to a friend

Print this page

Social Bookmark this article: What is this?

Digg

Delicious

Reddit

StumbleUpon

Slashdot

Google

Facebook