remark With lots expected fanfare, the Top500 Supercomputer record changed into introduced closing week on the overseas Supercomputing conference in Dresden, Germany.
For the fourth time, the BlueGene/L system construction by means of IBM and DOE's countrywide Nuclear security Administration that is installed at DOE's Lawrence Livermore country wide Laboratory claimed the #1 spot with a Linpack benchmark performance of 280.6 TFlop/sec.
The 2nd and third-placed programs also surpassed one hundred TFlop/sec: the upgraded Cray XT4/XT3 at DOE's all rightRidge national Laboratory, with a benchmark efficiency of 101.7 TFlop/sec, and Sandia countrywide Laboratory's Cray purple Storm equipment, rated a hundred and one.4 TFlop/sec.
Dell's Abe gadget at NCSA, built on 1,200 PowerEdge blades, ranked number eight ordinary, the maximum performing blade answer. IBM's MareNostrum supercomputer, built on IBM BladeCenter JS21 blade servers, ranked number 9 on the usual checklist to continue its place as the strongest supercomputer in Europe.
The performance cost required to make it onto the list multiplied to four.005 TFlop/sec on the Linpack benchmark, in comparison with 2.737 TFlop/s six months in the past. The equipment ranked quantity 500 on the latest record would have held place number 216 simply six months prior. this is the biggest turnover between lists in the Top500 challenge's 15 12 months history.
a complete of 289 systems (~58 per cent) use Intel processors, 105 methods (21 per cent) function AMD's Opteron household, and eighty five systems (17 per cent) are in line with IBM vigour processors. dual-core processors are the dominant chip structure. Intel's Woodcrest twin-core chip confirmed essentially the most boom, with 205 methods the use of this chip compared to 31 six months in the past. an additional 90 systems use Opteron dual core processors, up from seventy five six months in the past.
Clusters stay probably the most typical architecture in the Top500 checklist. The listing has 373 clusters, representing simply under three-quarters of all systems. HP Cluster Platform 3000BL and 4000BL techniques, in accordance with HP BladeSystem c-type servers, accounted for 152 entries on the record. HP has the maximum variety of systems on the record with 40.6 per cent, IBM now has 38.four per cent, and Dell holds four.eight per cent, with out a other brand having more than 5 per cent of the systems listed.
For normal total efficiency represented, IBM has forty one.9 per cent of installed efficiency, HP follows with 24.5 per cent, subsequent is Dell at nine per cent, Cray with 7.3 per cent, and SGI at 5.7 per cent. The commonplace age of a equipment within the Top500 listing is three hundred and sixty five days and two months with 60 per cent having been put in or upgraded this year and 23 per cent having been installed or upgraded ultimate yr.
no longer particularly, the sheer amount of computational horsepower listed on the Top500 has damaged one other record, and the desk stakes required to play in this game have jumped up about 50 per cent over six months ago. The trade's capacity to proceed to shatter previous assumptions about performance limits has confirmed remarkably resilient and maybe serves observe that they may still conveniently cease making an attempt to outline what the restrict of computing will ever be.
on the other hand, what they discover most pleasing in the newest record is how regular, and the way far up the totem pole, blade-based mostly options have turn into. The Dell Abe system put in at NCSA ranked quantity eight typical, IBM's equipment in Barcelona adopted in ninth location, and a lot of of the entries from HP were primarily based upon the BladeSystem c-class.
These are impressive achievements for an structure that for many is fitting the de facto IT platform of the longer term. From an usual effectivity and value standpoint, this bodes neatly for the architecture and begs the question: on this planet of HPC, does the blade server offer the most appropriate price per TFlop?
Granted, HPC programs are often incredibly customised, but the inherent discount in cabling, power supplies, external switches, and so on., would appear to translate even in an HPC scale. it will be wonderful to look simply how the effectivity of the blade structure scales in the maximum-performance environments.
typical, they are impressed with the continued boom in HPC ability, and in certain with the solutions from IBM, HP, and Dell which are based mostly upon technologies often attainable to mere mortals albeit in a lot smaller footprints. The cost of clustering, blades, and virtualisation applied sciences is awfully evident within the Top500.
To their approach of considering, this should still aid retain providers very attracted to pursuing these technologies of their generic market offerings which, in any case, is where most companies make the lion's share of their profits. whereas HPC solutions of the optimum scale customarily do not carry a earnings margin that might maintain most providers afloat, the R&D funding has a substantial payback because the carriers carry the improvements that they improve in fixing one of the vital computational most intricate projects right down to the time-honored-goal choices upon which the trade and establishments thrive.
Copyright © 2007, The Sageza community
subsidized: becoming a practical safety leader