$arr_19 ), array( 3, false, $arr_20, $arr_24 ), array( 2, false, "\" />", $arr_25 ) ) ); ?> $arr_27 ), array( 3, false, $arr_28, $arr_30 ), array( 2, false, "\" />\n\n", $arr_31 ) ) ); ?> array( 2, false, false, $arr_9 ), array( 4, $arr_10, "if", $arr_245, $arr_248 ), array( 2, false, "\n", $arr_249 ) ) ); ?> rr_466 ), array( 4, $arr_467, "if", $arr_482, $arr_484 ), array( 2, false, "\n", $arr_485 ) ) ); ?> Jaguar Passes Roadrunner in Top500 (Gallery) » Linux Magazine
 

Jaguar Passes Roadrunner in Top500 (Gallery)

Nov 17, 2009

The Petaflop pioneer Roadrunner has been relegated to second place after the 225,000 processor cores of the Jaguar XT5. Europe's largest supercomputer, Bluegene/P, has meanwhile slipped in the ranks.

The current world champion of supercomputers, the Jaguar XT5, which since end of October underwent a processor upgrade from four-core to six-core CPUs, at 2.3 Petaflops/s, has trumped second place Roadrunner by nearly 100%. Petaflops/s refers to quadrillion calculations per second.

The Jaguar is a Cray XT5 system with 37,376 six-core Opterons, nearly 300 TeraBytes of RAM and 10,000 TeraBytes of hard drives. The researchers at the National Center for Computational Sciences (NCCS) use the Jaguar at the Oak Ridge National Laboratory (ORNL) for simulation and computations for environmental, chemical and material science, nuclear energy, astrophysics and particle physics (see Gallery). All the Jaguar computer nodes somewhere run SUSE Linux, whose kernel Cray reduced to essential services to avoid interrupting any scientific applications.

The National Institute for Computational Sciences (NICS) at the University of Tennessee, the NCCS partner, took third place with its Kraken supercomputer, with 832 Teraflops/s, 100,000 processor cores and 130 TeraBytes of RAM. The machine is again from Cray, and another XT5 model, to be specific. Bluegen/P from the Juelich research center in Germany this time had to be satisfied with fourth place: with 825 Teraflops/s it meant stepping down from third place it still had in June 2009.

Nearly 90% (446 machines) from the Top500 systems run on Linux, among them SUSE, Red Hat and CentOS. Second place went to UNIX at 5% (25 machines), third to Windows with 1% (5 machines). Although the research runs over the fastest supercomputers, together with the academic realm it amounts to only 35% of the worldwide giants with more than 20 Teraflops/s. Industrial investors are banking on High Computing in that it blocks off almost 63% of all supercomputers. About 55% are deployed in the U.S., with England a distant second at 10% and Germany at 5.4%. HP took over the lead from IBM as manufacturer, even though IBM is still assembling the most powerful computers worldwide.

Gallery (8 images)

Comments

  • Flops = Floating-point operations per second

    Hey guys, "Petaflops/s" is wrong.

    1.0 Petaflops means 1.0 x 10^15 floating-point operations PER SECOND.

  • Don'f forget Debian ;)

    Among the top500 are also Debian (e.g. on the world's most efficient (linpack based) Gbit ethernet based cluster) based systems happy
comments powered by Disqus

Issue 152/2013

Tag Cloud

Bruce Byfield Gnome KDE Kernel Linux Linux Pro Magazine Open Source Projects Red Hat Ubuntu events foss free software google

News