The Sunway TaihuLight is at present the speediest structure on the planet, as indicated by the twice-per-yearTOP500 list, with a shocking Linpack benchmark consequent result of 93 petaflops. That makes it three times speedier than the earlier champion, China’s Tianhe-2, which we’ve secured distinctive times and had sat on top of the rundown since it first went online in 2013.
What’s much likewise entrancing this time around is what’s under TaihuLight’s hood: a subtly made ShenWei processor and custom interconnect, rather than parts sourced somewhere else. The ShenWei 26010 is a 260-center, 64-bit RISC chip that surpasses 3 teraflops at most noticeable tilt, putting it keeping pace with Intel’s Knight’s Landing Xeo Phi. TaihuLight contains 40,960 ShenWei 26010s, one for every inside that also contains 32GB of RAM, indicating an aggregate of more than 10 million core interests.
The chip has four center parties, each with 64 fragments and a solitary association dealing with portion, for a sum of 65 for each social affair (to get to the 260 center aggregate). Every gathering practices a 136.5 GB/sec memory controller; there’s no word on the system advancement focus guide utilized toward make the chip. The TOP500 report said that the chip moreover does not have any standard L1-L2-L3 hold, and rather has 12KB of principle store and 64KB “neighborhood scratchpad” that works kind of like a L1 store.
The custom interconnect, called the Sunway Network, depends on upon PCIe 3.0 and goes on 16 GB/sec of crest transmission limit between focus focuses, with 1ms dormancy.
TaihuLight will be utilized for air, environment, and earth frameworks delineating; life science research; hoarding; and information examination, as appeared by TOP500’s bona fide report. The structure is expertminded at the National Supercomputing Center in Wuxi, which lounges around two hours west of Shanghai.
This is the primary experienced China has the most frameworks on the quick overview, with 167, rather than the US, which is down to 165. China in like way now has the essential two speediest frameworks as well. Europe has 105 frameworks, down two from November 2015. Cray keeps driving in firm execution offer at 19.9 percent, however that is down from 25%. China’s National Research Center of Parallel Computing Engineering and Technology, which made TiahuLight, takes the second spot here with this single machine, at 16.4% offer, while IBM conveys third with 10.7%.
Firm joined execution of every one of the 500 supercomputers has bounced basically, from 420 petaflop/s six months going before 566.7 at this point. Ninety-five frameworks on the synopsis now surpass one petaflop.
Here’s the present rundown of the 10 snappiest supercomputers on the planet:
1. TaihuLight: Sunway MPP, SW26010; National Supercomputing Center in Wuxi, China; 10.6 million centers (93.01 petaflop/s).
2. Tianhe-2: TH-IVB-FEP Cluster; National Super Computer Center in Guangzhou, China; 3.12 million centers (33.86 petaflop/s).
3. Titan: A Cray XK7 framework at the Department of Energy’s Oak Ridge National Laboratory (17.59 petaflop/s).
4. Sequoia: An IBM BlueGene/Q framework organized at the Department of Energy’s Lawrence Livermore National Lab in California, with 1.57 million core interests.
5. K Computer: A SPARC64 framework with 705k centers at RIKEN Advanced Institute for Computational Science in Japan.
6. Mira: IBM BlueGene/Q; DOE/SC/Argonne National Laboratory, US; 786k custom IBM centers.
7. Trinity: Cray XC40; DOE/NNSA/LANL/SNL, US; 301,056 Xeon E5-2698v3 core interests.
8. Piz Daint: Cray XC30 with 116k Xeon and Nvidia centers; organized at the Swiss National Computing Center in Switzerland.
9. Hazel Hen: Cray XC40; HRLS-Höchstleistungsrechenzentrum Stuttgart, Germany; 185k Xeon E5-2680v3 core interests.
10. Shaheen II: A Cray XC40 at King Abdullah’s University of Science and Technology in Saudi Arabia, which means the essential appearance of a Middle East supercomputer in the crucial 10