November 16, 2009
The Cray XT5 supercomputer known as Jaguar has won the title of world's fastest computer, ending the 18-month reign of IBM's Roadrunner.
An upgrade from quad-core to six-core Opteron processors from Advanced Micro Devices boosted Jaguar's performance to 1.75 petaflop per second, easily surpassing Roadrunner's 1.04 petaflop/s, according to the biannual Top500 list released Monday at the SC09 Conference in Portland, Ore. A petaflop/s refers to 1 quadrillion calculations per second. Jaguar, located at the the Department of Energy's Oak Ridge Leadership Computing Facility, came close to beating Roadrunner in the two previous Top500 lists. This time, however, Roadrunner's performance fell from 1.105 petaflop/s in June due to a repartitioning of the system. Roadrunner, which was crowned No. 1 in June 2008 after becoming the first supercomputer to break one petaflop/s, is located at the Los Alamos National Laboratory. Jaguar, which has nearly a quarter of a million CPU cores, has a theoretical peak capability of 2.3 petaflop/s. The Linpack benchmark is used in determining the rankings of the Top500 list compiled by researchers at the University of Mannheim, Germany, NERSC/Lawrence Berkeley National Laboratory, and the University of Tennessee, Knoxville. Coming in at No. 3 in the latest Top500 list was another upgraded Cray XT5 system. Known as Kraken, the supercomputer, located at the National Institute for Computational Sciences/University of Tennessee, achieved a speed of 832 teraflop/s. A teraflop/s is a trillion calculations per second. No. 4 was the IBM BlueGene/P supercomputer located at the Forschungszentrum Juelich in Germany. The BlueGene/P, the fastest computer outside of the United States, achieved 825.5 teraflop/s. The system was No. 3 in the previous Top500 list. Rounding out the top five was the the new Tianhe-1, which means River in Sky, at the National Super Computer Center in Tianjin, China. The system is used for research in petroleum exploration and simulation of large aircraft designs. It is the highest ranked Chinese system ever and is a hybrid design with Intel Xeon and AMD graphics processors used as accelerators. Each node consists of two AMD GPUs attached to two Intel Xeon chips. The only other new system among the top 10 supercomputers is a Sun Microsystems blade system at the Sandia National Laboratories. Called Red Sky, the Sun Blade system reached 423 teraflop/s. Overall, the world's supercomputers are getting faster. The entry level of the latest list moved up to 20 teraflop/s from 17.1 teraflop/s. The last system would have ranked No. 336 in the previous Top500 list six months ago. The number of systems using Intel processors increased slightly to 402 from 399 in June. IBM Power processors are the second most commonly used processor family among the Top500 with 52 systems, down from 55 in the previous list. AMD's Opteron family is third with 42 systems, one less than in June. IBM and Hewlett-Packard continue to sell the bulk of systems on the list. HP accounts for 210 system versus IBM's 186 system. HP was also the No. 1 seller in June. In terms of speed, however, IBM remained the clear leader with 35.1% of installed total performance, down from 39.4% six months ago. HP is second with 23%, down from 25.1%. Other vendors with more than 5% installed total performance include Cray, 15.9%, and SGI, 6.6%. The United States is the leading consumer of supercomputers with 277 of the Top500 systems, down from 291 in June. Europe's share rose to 153 from 145. Asia accounts for 50 systems, up from 49. The latest Top500 list is the 34th edition of the closely watched ranking. The list will be formally presented Tuesday at the SC09 Conference.
InformationWeek Analytics has published an independent analysis on application delivery. Download the report here (registration required).
About the Author(s)
You May Also Like