Today Super Computer

Ai

At any moment, there are many well-publicized supercomputers that operate at extremely high speeds relative to all other computers. The word is also sometimes applied to far slower (but still impressively fast) computer systems. The largest, most strong supercomputers are actually multiple PCs that perform parallel digesting. In general, there will be two parallel processing methods: symmetric multiprocessing (SMP) and massively parallel processing (MPP).

By June 2016, the most effective supercomputer in the world was the Sunway TaihuLight, in the city of Wixu found in China. A few figures on TaihuLight:

  • 40,960 64-bit, RISC processors with 260 cores each.
  • Peak performance of 125 petaflops (quadrillion floating point procedures per second).
  • 32GB DDR3 memory space per compute node, 1.3 PB memory in total.
  • Linux-based Sunway Raise operating system (OS).

Significant supercomputers throughout history:

  • The first commercially successful supercomputer, the CDC (Control Data Corporation) 6600 was created by Seymour Cray. Released in 1964, the CDC 6600 had an individual CPU and cost $8 million - the equivalent of $60 million today. The CDC could handle three million floating level operations per second (flops).

Cray continued to found a good supercomputer firm under his name found in 1972. Although the business has changed hands several times it really is still functioning. In September 2008, Cray and Microsoft introduced CX1, a $25,000 personal supercomputer targeted at markets such as aerospace, automotive, academic, personal services and lifestyle sciences.

IBM is a keen competitor. The business’s Roadrunner, after the top-rated supercomputer, was twice as quickly as IBM’s Blue Gene and six circumstances as fast as any of other supercomputers in those days. IBM’s Watson is famous for having adopted cognitive computing to defeat champion Ken Jennings on Jeopardy!, a favourite quiz show.

In America, some supercomputer centers are interconnected on an Internet backbone known as vBNS or NSFNet. This network is the foundation for an evolving network infrastructure known as the National Technology Grid. Internet2 is a university-led task that is part of the initiative.

At the low end of supercomputing, clustering takes even more of a build-it-yourself method of supercomputing. The Beowulf Task offers guidance on how to put along a number of off-the-shelf pc processors, using Linux operating systems, and interconnecting the processors with Fast Ethernet. Applications must be written to control the parallel processing.

10 million-core supercomputer hits 93 petaflop/s, tripling speed record

There’s a fresh world’s fastest supercomputer for the first time in three years.

A Chinese supercomputer named Sunway TaihuLight today ranks as the world’s fastest, nearly tripling the prior supercomputer quickness record with a score of 93 petaflops per second. That’s 93 quadrillion floating point functions per second (or 93 million billion).

Sunway TaihuLight surpassed another Chinese supercomputer, Tianhe-2, which had been the world’s fastest for 3 consecutive years with speeds of 33.9 petaflop/s, based on the latest Top500.org rank released today. Top500 rankings derive from the Linpack benchmark, which necessitates each cluster “to resolve a dense program of linear equations.”

“Sunway TaihuLight, with 10,649,600 processing cores comprising 40,960 nodes, is doubly fast and 3 x as reliable as Tianhe-2,” the Top500 announcement explained. Sunway TaihuLight is probably the world’s most efficient systems, with “peak power consumption under load (jogging the HPL benchmark)… at 15.37MW, or 6 Gflops/Watt.”

The machine has memory of just one 1.3PB, or 32GB for every node. This is really not much memory considering just how many cores the machine has; if it employed “a far more reasonable amount of memory because of its size,” Sunway TaihuLight will be far more power-hungry. Sunway provides more than three times as many cores as Tianhe-2, but it uses less overall vitality-15.37MW vs 17.8MW.

Tianhe-2 have been the world’s fastest in six consecutive rankings, which happen to be released twice a yr. Tianhe-2 took the most notable spot for the 1st time in June 2013, defeating the previous most effective supercomputer, which kept the position just once.

Developed by China’s National Research Center of Parallel Pc Engineering & Technology, Sunway TaihuLight is installed in the National Supercomputing Centre in Wuxi, China. While Tianhe-2 uses Intel processors, Sunway TaihuLight was created entirely with processors designed and created in China. Sunway TaihuLight runs on the custom interconnect based on PCIe 3.0 technology.

Each node in Sunway TaihuLight has one SW26010 chip, a new version of the ShenWei processor that makes speeds of 3 teraflop/s with 260 cores. This is a 1.45GHz 64-bit RISC processor, however the Leading500 announcement said “its underlying architecture is somewhat of a mystery.”

“In 3 teraflops, the brand new ShenWei silicon is on par with Intel’s ‘Knights Landing’ Xeon Phi, another many-core design, but one with a more public history,” Best500 stated. “In somewhat of related irony, it had been the united states embargo of high-end processors, including the Xeon Phi, imposed on several Chinese supercomputing centers in April 2015, which precipitated a far more concerted effort in that country to build up and manufacture such chips domestically. The embargo probably didn’t affect the TaihuLight time-line, because it had been set to receive the brand new ShenWei parts. Nonetheless it was extensively imagined that Tianhe-2 was in series to get an upgrade using Xeon Phi processors, which could have very likely raised its functionality into 100-petaflop territory prior to the Wuxi program came online.”

While the US embargo is set up because of concern about nuclear research, Top500.org said Sunway TaihuLight is planned for work with in exploration and engineering work in fields including climate, weather and Earth devices modelling, life science exploration, advanced manufacturing, and data analytics.