Supercomputer in a EXAFLOP (2018)

In games on November 24, 2010 at 7:12 pm

The San Diego Supercomputer Center (SDSC) at the University of California has launched a project aimed at designing the next generation supercomputers. The project is part of UHPC (Ubiquitous High Performance Computing), a program that is responsible for the Agency DARPA. Responsible for the project plan have ready a thousand times faster than the current supercomputer capable of performing a EXAFLOP (a trillion operations per second floating-point million) in 2018.It be achieved?

FLOPS stands for FLoating point Operations Per Second, or floating-point per second, and is used to measure the performance of a computer. The FLOPS a modern computer may make a second amount is enormous, so use the prefixes “mega” (MFLOPS, or one million FLOPS), “giga” (GFLOPS, or billion FLOPS), etc. A PC equipped with a microprocessor running a 3 GHz desktop has a power of several tens of GFLOPS.The fastest computer in the world (at the time of writing this article) is the Tianhe-1A, able to conduct more of 2,507 PFLOPS (petaflops), or equally, 1,000,000,000,000,000 floating point per second. 7.168 14.336 Processors Intel Xeon the Tianhe-1A and Nvidia Tesla M2050 graphic processing units allow you to be approximately 50 million times faster than a desktop PC. But even this power may not be sufficient for certain simulations.

Indeed, some problems require an enormous amount of calculations to be solved, and when it comes to problems must be solved “real-time” (such as those relating to the trajectory of a missile that is coming to you above), faster computer handy, you better.The Research Agency (DARPA, by Defense Advanced Research Projects Agency) Defense Advanced projects is among organizations interested in have extremely fast supercomputers as such problems falls right in your workspace. Therefore, have been targeted in the next generation of supercomputers which should be a thousand times faster than Tianhe-1A, surpassing the fantastic EXAFLOP (a million trillion floating-point per second) barrier. To achieve this, a team of San Diego Supercomputer Center (SDSC) of the University of California, with funds provided by DARPA, UHPC (Ubiquitous High Performance Computing) program has begun work on this machine than should be point-to-point 2018.

Tianhe-1A, the fastest in the world (NVIDIA)

Largest manufacturer worldwide, Intel, CPU also participates in the project.One of the problems to be solved so that this computer becomes an actual device is power consumption.Tianhe-1A machines consume more 4 megawatts, and without a major change in the architecture of circuits used to build the machine dreamed by DARPA, the energy required to make it work would be about 1000 times higher, something completely unacceptable. Resources from Intel and DARPA certainly provide a clever solution to this problem.Tianhe-1A, for example, if you only use CPUs had consumed 12 megawatts, but the amount of energy required to rely on Nvidia Tesla M2050 graphics processing chips is three times less.The early stages of the supercomputer design should be completed in 2014, and the first functional prototype would see the light in 2018.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: