Global supercomputer rankings TOP10 China occupies two seats

At last week's SC11 conference in Seattle, the global http://www.aliyun.com/zixun/aggregation/13840.html "> Super computer TOP500 rankings," The Japanese "Beijing" Computer) continues to occupy the top spot with computing power spanning 100 million times per second. At the same time, in the top ten systems, two sets of supercomputers are from China, from the "Tianhe", which is deployed in Tianjin, and the "Dawning Nebula" efficient computing system deployed in Shenzhen. Let's take a look at the most powerful computing systems on the planet.

1.K Computer first supercomputer that spans billions of billions of times

The fastest supercomputer in the world "Jing" (K Computer) is a joint project of the RIKEN Institute of Advanced Computing Sciences (AICS) and Fujitsu in Japan. "Beijing" (K Computer) does not use GPU acceleration, but is completely based on traditional processors. The maximum performance of "Jing" (K Computer) is four times times the number of "Tianhe" in the second place. The current "Beijing" (K Computer) is equipped with 88128 Fujitsu SPARC64 Viiifx 2.0GHz Eight core processors, with a total core of 705,024, maximum computational performance 10.51petaflop/s, Peak performance 11.28038 petaflop/ S, at the same time efficiency up to 93.2%, total power consumption of 12659.9 kw.

2. The King of Tianhe

The "Tianhe System", located at the National Supercomputer Center in Tianjin, China, ranks second in the latest rankings. The computational ability reaches 2.57 petaflop/s. Last year, the Tianhe was also ranked top of the TOP500 rankings. The Tianhe one uses a CPU+GPU hybrid architecture. Equipped with 14336 Intel Xeon X5670 2.93GHz Six core processors, 7168 nvidia Tesla M2050 High-performance computing cards, and 2048 China self-developed FT-1000 Eight core processors, totaling more than 200,000 processor cores, It is also equipped with a proprietary internet. The cost is more than 600 million RMB.

3.JAGUAR XT5 for civilian supercomputers

"JAGUAR" Super computer system is part of the U.S. Department of Energy, located in Oak Ridge National Laboratory. The "JAGUAR XT5" system is funded by the National Science Foundation, the Cray Company, the University of Tennessee and the National Institute of Computational Science. It was ranked first in the TOP500 rankings in June 2010. "JAGUAR" is a civil computer, the use of AMD Magny-cours core Opteron processor, its maximum computational capacity of 1.75 petaflop/s. It is mainly used to simulate climate change, energy generation and other basic science studies.

4. The Dawning Nebula uses the HPP architecture of its own design

"Nebula" is located in China's Shenzhen National Supercomputer Center. "Nebula" System operation Peak value of 3 petaflop/s, the largest computational performance 1.271 petaflop/s, and is China's first, the world's third to achieve double precision floating-point computing super million times the super computer, and its unit energy consumption provides a performance of 498 million times/watts.

"Nebula" supercomputer uses the HPP architecture which is designed by itself, which consists of 4,640 computational units, and uses highly efficient heterogeneous collaborative computing technology, which consists of 9280 universal CPUs and 4640 dedicated GPU. The computing network adopts unidirectional 40Gbps QDR InfiniBand technology, and the core storage adopts the Parastor high speed IO system.

5.Tsubame 2.0 Adding a variety of new hardware technologies

The Tsubame 2.0 supercomputer was launched jointly by the Tokyo University of Technology and NEC and HP. The speed of Tsubame 2.0 is 30 times times that of Tsubame 1.0, this supercomputer is configured with 2816 Intel Westmere processors and 4224 NVIDIA Tesla M2050 GPU, and added new hardware products such as DRAM and SSD and adopted Linux and Windows dual systems. Now the Tokyo University of Technology is working on a plan for Tsubame 3.0.

6.Cielo provides support for national laboratory calculations

Cielo in Spanish means "sky". Cielo now provides computational support for Los Alamos, Sandia and Livermore three national laboratories. The Cielo supercomputer is based on Cray's next-generation Baker architecture and uses AMD's newest Magny-cours architecture processor to compute node-side Cielo connected to each other by Gemini.

7.Pleiades Unique Lustre File system

The Pleiades supercomputer provides computational support for the NASA Ames Research Center in California, USA. The Pleiades is composed of 4 nuclear harpertown and 6 core Westmere, with a core number of 81,920. The network connection architecture also uses the InfiniBand method. Pleiades's maximum computational power reaches 1.09 petaflop/s. and uses the lustre file system peculiar to the supercomputer.

8.HOPPER Super Computer named Navy Major General

Hopper supercomputer is named after the U.S. Navy Major General Grace Hopper, Grace Hopper is also a pioneer in software and programming languages. Today Hopper is located at the National Energy Research Computing Center at the Lawrence Berkeley National Laboratory. Hopper employs the Cray XE6 system unit, which contains the AMD 12 core Opteron 6172 processors, with a total of 153,408 cores, with a maximum computational power of 1.05petaflop/s.

9.tera-100 the strongest supercomputer in continental Europe.

TERA-100 is currently the most computationally powerful supercomputer in Europe. The theoretical peak performance of Tera-100 system is 1.25petaflop/s, and the maximum computational performance is about 1petaflop/s. TERA-100 completed construction last March, May 26, the official power to start. The Tera-100 consists of 4,300 BULLXS series servers. The system deploys 140000 Intel Xeon7500 processing cores, internal memory capacity is 300TB, external memory capacity reaches 20PB, and global file system throughput is 500gb/s. TERA100 will be used in the French nuclear weapons simulation program.

10.Roadrunner Hybrid Super Computer

Roadrunner supercomputer, located at Los Alamos Laboratories in the United States, topped the top 500 rankings in June 2008. It is also the world's first supercomputer with more than 1 petaflop/s computing power. Roadrunner is also the world's first hybrid supercomputer with a cell processor. Three different programming languages are required to ensure efficient processing of data at different cores. Roadrunner is built mainly by the cell broadband engine of play redevelop 3 gaming machine and the AMD x86 processor. Roadrunner has 130,536 computing cores, including 116,640 PPE and SPE cores for 12960 cell engines, and 6948 dual-core AMD Opteron processors.

(Responsible editor: Lu Guang)