PC clusters ranked in supercomputer top 10

Two PC-based cluster computers have entered the top 10 list of the most powerful supercomputers in the world, according to the Top 500 supercomputer list released this week.

The 20th edition of the list was released in time for the annual SC2002 supercomputer conference in Baltimore, Md., due to run from Nov. 16-22. The list is compiled twice a year by researchers at the University of Mannheim, Germany, the University of Tennessee and the U.S. National Energy Research Scientific Computing Center (NERSC) in Berkeley, Calif.

The Earth Simulator supercomputer, installed earlier this year by NEC of Tokyo at the Earth Simulation Center in Yokohama, Japan, retained its No. 1 position with a benchmark performance of 35.86T flops (trillions of floating-point operations per second), far ahead of any competition.

Second and third positions have been taken by two new, identical Hewlett-Packard ASCI Q systems at the Los Alamos National Laboratories. The HP computers are based on the AlphaServer SC computer system and offer a 7.73T flops performance each. They take the places of former IBM’s 7.23 T flops ASCI White Computer at the Lawrence Livermore National Laboratory in California, now in fourth place, and HP’s 4.46T flops AlphaServer SC at the University of Pittsburgh, now in sixth place.

Ahead of HP’s AlphaServer SC in fifth place is a cluster at the Lawrence Livermore National Laboratory, built by Linux Networx Inc. of Sandy, Utah and Quadrics Ltd of Bristol, England. A cluster at the Forecast Systems Laboratory in the National Oceanic and Atmospheric Association in Boulder, Colo., built by High Performance Technologies of Reston, Va., is in eighth place.

Entry to the top 10 of the list now requires a performance of more than 3.2 terabyte flops. Performance is growing fast, and the number of supercomputers exceeding 1 terabyte flops has grown from 23 to 47 in just six months, the researchers said.