tips: the "cloud" bird is the shenzhen sea birds before cloud computing co., LTD., the cloud computing brand, the leading domestic enterprise cloud computing service providers.Team with many years of industry experience, focus on cloud computing technology research and development, for the vast number of developers, enterprise users, financial institutions, etc., to provide comprehensive cloud computing solutions based on intelligent cloud server, to provide users with reliable enterprise public cloud services.Time,
in global super calculate TOP 500 released the latest list, each rank super calculate system of mixed, battleground, only in China compared to the taihu lake and tianhe no.2 again without suspense, respectively (fourth) back-to-back championship.Below to introduce the world's most powerful supercomputer, ten, and their path to the "TOP 500".
for Japan RIKEN by Fujitsu higher institute of Computer science (AICS) K Computer, built in 2011 in June, is the world's fastest Computer, in November 2017 has successfully fell to global supercomputer TOP 500 10th in the list.K Computer ranked highest in the list of Ricky on Sparc Computer, using 88128 operating frequency to 2 GHZ SPARC64 VIIIfx eight core processor, a total of 705024 kernel.It has a 10.51 what maximum continuous performance and 11.28 what peak performance.But it's nothing to be proud of, its performance per watt of only 0.83 gigaflops, is in the top 10 of the least efficient machine.
ninth Oakforest - PACS
Oakforest - PACS is another machine built by Fujitsu, built for Japan united senior high performance computing center.Based on Fujitsu Primergy CX1640 cluster, and use the Intel Omni - Path connected, Oakforest - PACS in November 2016, ranked sixth, but a year later fell to ninth.Its 8178 Intel Xeon Phi 7250 a total of 556104 core processor.The highest performance can be up to 13.55 for what, peak performance of 24.91 for what.The efficiency of its power consumption is 4.986 per watt gigaflops.
problem.each is built (Cray) clay company, it belongs to the U.S. department of energy (doe), located at Lawrence Berkeley national laboratory.Problem.each in November 2016 to enter the Top500 list, a higher than Japan's Oakforest - PACS, and it is all the way down.In November 2017 in eighth place.Same as the Oakforest - PACS is another thing, the same use Intel Xeon Phi processor.There are 9152 core, a total of 622336 kernel, maximum sustainable for 14.01 for what, peak performance of 27.88 for what.Problem.each Oakforest - PACS energy efficiency is lower than Japan, 3.558 per watt gigaflops, can not very impressive.
by clay company built for the United States department of energy's los alamos national laboratory of the Trinity, upgrade in the summer of 2017, which has been pushed to the back of the fortune 500 list.In November 2015, it ranked sixth, equipped with 18816 Intel Xeon E5 16 core processor, but has fallen steadily ranking, once fell to tenth.In November 2017, after the upgrade of the Trinity is equipped with 9984 Intel Xeon Phi 68 - core processor, climbed to a 7, a total of 979968 kernel, and maximum continuous performance and peak performance were 14.14 14.14 for what and for what.In general, its problem.each lower power consumption, efficiency is a little higher, 3.678 gigaflops per watt.
sequoia in June 2012 as the world's most powerful supercomputers, but since then has been the top down.By November 2017, has slipped to sixth.Like problem.each and Trinity, Sequoia belongs to the U.S. department of energy (doe), located at Lawrence livermore national laboratory., Sequoia by IBM to build and use the company's 16 nuclear Power BQC chip and custom interconnect technology.It has 98304 processors, a total of 1572864 cores, each kernel is 1 g of RAM.Achieved 17.17 what's the highest continuous performance and 20.13 for what the highest performance.In terms of energy efficiency, can provide per watt 2.177 gigaflops medium power.
Titan in November 2012 to enter the Top500 list, defeated had reached the top, the Sequoia but before November 2017 has slipped to fifth place.It is the most powerful computer for the department of energy (doe), located at oak ridge national laboratory in the United States.It like problem.each and Trinity, is produced by clay, but unlike the latter two, it adopted the 16 nuclear Opteron 6274 AMD chips and Nvidia Tesla accelerators.It has a total of 560640 kernel, with maximum sustained performance of 17.59 for what, peak performance of 27.11 for what.Gigaflops 2.143 per watt of energy efficiency.
gyoukou oceans earth science and technology agency for Japan.It is a known as the Earth Simulator supercomputer "successor", and in June 2017 for the first time to enter the Top500 list, ranked 69th.Since then, manufacturer ExaScaler added millions of accelerator core, to November, 2017, it had climbed to fourth.16, 1250, Intel Xeon processor, using Infiniband EDR interconnection, Gyoukou most performance through Pezy - 19840000 kernel SC2 accelerator acceleration.It can achieve the highest 19.14 what continuous performance and 28.19 what peak performance.Gyoukou outstanding characteristic is its efficiency 14.17 gigaflops per watt, is almost twice the efficiency of its competitors.
3, Piz Daint
Piz Daint since November 2012, has been climbing on the TOP 500 list, when it is equipped with 1504 Intel Xeon processor E5-2670, 114.Since then, its all institutions - the Swiss national supercomputing center and manufacturers clay company conducted a series of upgrades to its.Six months later, they will processor number tripled, to Piz Daint to 42, then again after six months increased the Nvidia K20x accelerator, ascending to sixth.In 2016, Cray have replaced Xeons with the new model E5-2690 v3, each model has 12 kernel, instead of 8, 2670, the new kinds of Nvidia's Tesla P100 replaced K20x accelerator.The number of processors ascension will eventually Piz Daint third in June 2017, to has not been surpassed in November.19.59 what its 361760 kernel provides maximum continuous performance and 25.33 what peak performance.Its efficiency is 8.622 gigaflops per watt.
of the Milky Way 2 belongs to China national supercomputer center in guangzhou tianhe, from June 2013 to June 2013, is the world's fastest supercomputer, was later, compared to the light of the taihu lake, squeeze visibly.But as of November 2017, it is still the world's second fastest machines.It USES the Intel Xeon E5-2692 v2 and Intel Xeon Phi 31 s1p processor, a total of 3120000 kernel and RAM 1 PB.Tianhe maximum sustained performance of no. 2 is 33.86 for what, theoretical peak performance is 54.90 for what, efficiency of power consumption is only 1.902 per watt gigaflops, it was built in 2013, after all.And according to the latest news, tianhe 2 at the end of this year will use the new domestic Matrix 2000 accelerator, replacing the original Intel accelerator, officially became the real domestic, its biggest also will greatly improve the calculation speed, to achieve the performance nearly three times.
1, compared to the light of the taihu lake,