News Icon

News: Gaming Console Supercomputer Can Read, Correct Input

Story by Ian GrahamSmall RSS IconSubscriptions Icon

WASHINGTON - Video games have advanced by leaps and bounds in the past few years. What once was a big black box with a bad video version of ping-pong is now a sleek, motion-capturing, high-resolution computer system capable of networking around the world.

Mark Barnell, director of high-performance computing and the Condor Cluster project at the Air Force Research Laboratory, has used that technology to create a new supercomputer.

The Condor Cluster, a heterogeneous supercomputer built from off-the-shelf commercial components -- including 1,716 Sony PlayStation 3 game consoles -- could change the supercomputing landscape, Barnell said yesterday in a "DOD Live" bloggers roundtable.

The system computes operations by the trillions per second –- called "teraflops" in the computing world. Some supercomputers can operate at a quadrillion calculations per second, or a "petaflop."

Barnell said the Condor Cluster also represents new ways for supercomputers to increase computational resources while using less energy. Barnell said it's currently the seventh-greenest computer in the world.

"This particular system is about half a petaflop, or capable of about 500 trillion calculations per second," he said. "In the current time that we can measure it, it's about the 35th- or 36th-fastest computer in the world, and with some things that are going to be changing in the next eight or nine months with some upgrades, we could boost it to maybe the 20th-fastest computer in the world, and at the same time make it, at that moment in time, the greenest computer."

The Condor Cluster isn't designed to compete with the world's largest supercomputers, he said. The Condor Cluster, which cost $2 million to build, is made for more specific tasks. The cheapest comparable supercomputers would cost $50 million to $80 million, he said. The highly advanced Cray supercomputers are in the $100 million range.

"So from a price performance, we'd probably beat all of them, but the biggest thing for us was the particular applications and the hardware we chose to build this computer with purposely matches those applications well," Barnell said. "Some of the systems that you might refer to in the top 10 in the world are more of a general-purpose computer and also run applications that we may not. We're just going to coexist and do some things that we need to get done with this particular supercomputer."

One area the Condor Cluster is being used in is neuromorphic computing, or "computational intelligence." Essentially, programmers write algorithms to "teach" the computer how to read symbols, letters, words and sentences. By programming the computer to read, in theory it can be taught to fill in gaps and "think" on its own. The idea is that the computer, when taking in millions of lines of data, could fill in gaps or rearrange the pages in case of human error.

The Condor Cluster can read 20 pages of information per second, and even with 20 to 30 percent of the characters on the page removed, can recover all of the sentences and words with about 99.9 percent accuracy, Barnell said. The discoveries this computer could lead to would change the face of computer science, he added.

"We have quite a few research and development efforts, working on those kinds of applications to do confabulation and prediction," he said. "That will open up a variety of areas which could help a lot of other efforts and a lot of the areas in which the Air Force would like to go."


Web Views
279
Downloads
0

Podcast Hits
0



Public Domain Mark
This work, Gaming Console Supercomputer Can Read, Correct Input, by Ian Graham, identified by DVIDS, is free of known copyright restrictions under U.S. copyright law.

Date Taken:12.03.2010

Date Posted:12.03.2010 17:22

Location:WASHINGTON, DC, USGlobe

Options

  • Army
  • Marines
  • Navy
  • Air Force
  • Coast Guard
  • National Guard

HOLIDAY GREETINGS

SELECT A HOLIDAY:

VIDEO ON DEMAND

  • Facebook
  • Twitter
  • Youtube
  • Flickr