Wednesday, June 28, 2006

IBM are fast, VERY FAST!!

IBM systems account for 243 of the 500 most powerful supercomputers in the world and more than half the total processing power according to the just-released Top 500 Supercomputer Sites list. IBM's Blue Gene/L at Lawrence Livermore tops the list with an unprecedented sustained performance of 280.6 Teraflops, or trillions of floating point calculations per second.

Bevan Lock, IBM South African Systems Evangelist, says IBM's industry-leading performance was propelled by its strength across diverse computing platforms: growth in the number of Blue Gene systems (from 19 to 25, compared with the previous list), Opteron clusters (from 8 to 33), and System p-based machines (from 54 to 58), including the debut on the Top 500 list of the first announced BladeCenter JS21-based supercomputer - the 15 teraflop system at Indiana University.

Joining Blue Gene/L in the TOP500 list's top three slots are IBM's own Blue Gene Watson system at 91.29 Teraflops, and the recently unveiled ASC Purple supercomputer at Lawrence Livermore National Laboratory with 75.76 Teraflops.

In related news, on June 22 the National Nuclear Security Administration (NNSA) and Lawrence Livermore National Laboratory announced that their scientists had used the world's most powerful computer, Blue Gene/L, to run a scientific code at a sustained performance level of 207 teraflops - the highest performing application ever run in the history of computing.

The application is used by Livermore scientists to understand the complex interactions of metals at the subatomic level and is a key element in the NNSA's ASC mission to protect and maintain the safety and efficacy of the United States' nuclear stockpile.

Other key indicators of IBM's supercomputing leadership include:

· IBM Leads the list with world's #1 supercomputer (BlueGene/L for US Department of Energy/NNSA/LLNL - 280.6 TFlops)

· #2 BlueGene/W at Watson Research

· #3 ASC Purple at LLNL

· Leads list with 243 entries (48.6%)

· Leads installed aggregate throughput with over 1,514 out of 2,790 Teraflops (54.3%)

· Most systems in Top 10 by any single vendor (4)

· Most systems in Top 20 with 11 systems (55%)

· Most systems in Top 100 systems with 46 (46%)

· Most cluster systems with 179 of 360 (49.7%)

The 'Top 500 Supercomputer Sites' is compiled and published by supercomputing experts Jack Dongarra from the University of Tennessee, Erich Strohmaier and Horst Simon of NERSC/Lawrence Berkeley National Laboratory and Hans Meuer of the University of Mannheim (Germany). The entire list can be viewed at

No comments: