Dell to build fastest Linux supercomputer

The National Center for Supercomputing Applications (NCSA) plans to use more than 1,450 Dell servers in a powerful Linux supercomputer that will be used for everything from predicting the demise of the universe to discovering new drugs to keep you alive until that happens.

The National Center for Supercomputing Applications (NCSA) plans to use more than 1,450 Dell servers in a powerful Linux supercomputer that will be used for everything from predicting the demise of the universe to discovering new drugs to keep you alive until that happens.

Linux is growing in popularity among supercomputer designers as an operating system for high-performance computing. IBM and Fujitsu recently unveiled plans to build Linux supercomputers that promised to be the most powerful Linux systems yet devised, until Dell’s announcement Tuesday at Linuxworld in San Francisco.

The NCSA plans to use 1,280 Dell PowerEdge servers, each with two 3.06-GHz Xeon DP processors from Intel, in a cluster running Red Hat Linux, Dell said Tuesday in a release. The cluster will also use 106 PowerEdge servers for I/O services and storage.

The cluster will produce theoretical peak performance of 17.7 trillion floating point operations per second (T FLOPS), which would make it the third most powerful supercomputer in the world based on the latest Top 500 list of the world’s supercomputers. The NCSA cluster is scheduled to come online later this year at the NCSA facilities at the University of Illinois at Urbana-Champaign.

IBM’s Linux supercomputer will use Advanced Micro Devices’ Opteron processor and Intel’s Itanium processor to deliver peak theoretical performance of 11.2T FLOPS. Fujitsu did not release the chip vendor that will power its supercomputer, but said it plans to deliver 12.4T FLOPS of peak theoretical performance. Both systems are scheduled to be released in March.

Cray will follow with a Linux supercomputer using Opteron processors that is expected to deliver 40T FLOPS of peak theoretical performance when it is released late in 2004 for Sandia National Laboratories.

Lawrence Livermore National Laboratory currently operates the fastest Linux supercomputer, which was built with clustering technology from Linux Networx. It delivers 11.1 TFLOPS of peak theoretical performance.