Network Computing is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

US Reclaims Supercomputing Crown

As predicted last week on NDCF, the U.S. has regained pole position in the latest Top 500 list of supercomputers, with the Department of Energys Blue Gene/L (BG/L) and NASA ’s Columbia system claiming first and second place (see US Stakes Supercomputing Claim and DOE Gains Supercomputing Crown).

After topping the list for the last five editions, Japan’s Earth Simulator, built by NEC Corp. (Nasdaq: NIPNY; Tokyo: 6701) in Yokohama, was pushed into third place by the raw power of BG/L and Columbia. The last U.S. supercomputer to top the charts was the Lawrence Livermore National Lab's ASCI White back in November 2001.

The list, unveiled today at the SC2004 conference in Pittsburgh, reveals that clusters are fast becoming the most popular supercomputing architecture. In the current list, clusters account for 296 systems, up from 291 six months ago and just 208 a year ago.

This will be music to the ears of the InfiniBand vendors. The major challenge that most organizations face when deploying clusters remains the issue of bandwidth. Not surprisingly, suppliers such as Voltaire Inc., Topspin Communications Inc., and InfiniCon Systems Inc. have been busy banging the InfiniBand drum at SC2004 (see Voltaire Expands InfiniBand Roster, Topspin, Sun Add Grid Solutions, and InfiniCon Delivers InfiniBand for Apple).

With bandwidth as the major choke point, the popularity of clusters also spells good news for the 10-Gbit/s Ethernet vendors, according to Max Flisi, research analyst at IDC. “Supercomputing clusters are an environment where we see a lot of 10-Gig happening,” he says.

  • 1