::Latest Reviews

Corsair TurboFlash 1GB
Posted Aug 8, 2007 by Carl Bloor

Sennheiser PC166
Posted Aug 8, 2007 by Sheri Bloor

Cool Labs Liquid Pro & MetalPad Review
Posted Aug 8, 2007 by Carl Bloor

World in Conflict Open Beta
Posted Jul 18, 2007 by Carl Bloor

::Latest Blogs

Website Enhaaaaanced
Posted Feb 14, 2008 by Carl Bloor

Back In Action
Posted Jan 29, 2008 by Carl Bloor

A Bit of a Rant
Posted Aug 19, 2007 by Sheri Bloor

New Website Features
Posted Jul 20, 2007 by Carl Bloor

::Forum Topics

the R4(R4DS) Cartridge
By ijkl888 in General Discussions

the R4(R4DS) Cartridge
By ijkl888 in General Discussions

the R4(R4DS) Cartridge
By ijkl888 in General Discussions

the R4(R4DS) Cartridge
By ijkl888 in Application Support

::Eureka PC News

    nVidias CUDA To Open GPUs To Supercomputing May 25, 2007 - Carl Bloor - Permalink

nVidia Corporation
In a presentation at the Microprocessor Forum 2007 in San Jose, nVidia is promoting its new software CUDA (Compute Unified Device Architecture) to use nVidia GPUs for regular CPU tasks.

That isn't to say the software enables a GPU to be completely used as though it were a CPU as they simply don't work the same. GPU's are designed to be hugely capable at a specific set of tasks. CPUs on the other hand are designed to handle almost everything thrown at it. An ideal example is the recent use of the Folding@Home program to utilise ATi graphics processors to compute the protein folding task usually given to idle CPUs around the world. As modern GPUs are programmable to a degree, this task can be optimised considerably, yielding rather impressive results. The same again happened with the PS3, using its monster processing capabilities to trump even the fastest PCs donating to the Folding@Home project.

The CUDA software is limited to only the most recent nVidia GPUs from the GeForce 8600, 8800 and the Quadro FX 4600 and 5600. The beta version has been available since February and is set to go on general release in the second half of this year.

The significant parallel computing offered by GPUs used in this way could be put to use in place of financial forecasting, science computing and medical research supercomputers amongst other things. Given the right software, of course – which nVidia hopes CUDA could provide. This could see supercomputers make way for massively paralleled GPU farms taking on the tasks previously reserved for organisations blessed with multi-million dollar supercomputers. The cost/time pay off however remains to be seen.

Δ

Chat(0)
Name:
?
Message:

::Featured

::Tech Buzz


 < Back -  Save as PDF -  Print -  Email to friend -  Top↑