Nvidia CUDA

[ThirdLeft152]

Hopefully this is not a double-post.. looks like our GPU's are fairly powerful after all. in some research through google some people report that GPU's can increase some computing up to 40-150x current CPU's potential.

NVIDIA CUDA

sorry for being a little lazy at the moment and not linking more information, google for more.

 

[Blitersety]

That's very interesting... but I wonder how many programs will ever acually be able to use this technology.

 

[ThirdLeft152]

well CUDA supports C programming, and as far as i can tell anything that can be programmed or converted into C language can be executed...

"The GPU is now a powerful, programmable processor that is ideally suited for computationally-intensive applications such as seismic processing for oil and gas exploration, computing in bioscience, and financial modeling," said Andy Keane, general manager of GPU Computing at Nvidia, in a statement. "With the recent introduction of CUDA technology, the parallel computing power of the GPU is now accessible to virtually any developer through standard C programming language."

Nvidia Releases CUDA SDK ExtremeTech - Find Articles

also im sure they will come out with something later that will take advantage of other languages as well.

 

[Blitersety]

But how will a program be able to switch between the CPU and the GPU... that's what confuses me. How will the program be able to recognize that it has the option to use the power of the GPU as well? Will it be able to use both at the same time?

I find this to be an extraordinary idea, I just hope that it works. Because if you think about it, a video card is basically a mini computer, complete with processor, RAM, and video outputs.

 

[Vaanish]

Thats GPGPU stuff. (General Processing on a Graphics Processing Unit)

To the people asking the possible uses of it: Folding is possible on certain ATi cards, and also ultra fast Virus scans can be performed on Graphics cards.

 

[ThirdLeft152]

yes it is, i dunno how they got it to work. but apparently its been used heavily in the research and process simulations.

check out some of the testimonials on nvidia's website to see the kind of stuff its currently being used for (however this is the tesla GPU, probably a more powerful card than the gtx)

NVIDIA Tesla Testimonials

if all this is true, who needs a cpu anymore. just throw the workload to the gpu lol.

 

[krazyq]

nice...
has anyone tried it yet?

I'll try downloading it once I get home.

 

[ThirdLeft152]

i sure haven't, i don't know C and i don't have a g8800 :/

 

[krazyq]

so for this to work...all i have to do is download
NVIDIA CUDA
Download] CUDA Toolkit version 1.0 for Windows XP (32-bit)

[Download] CUDA SDK version 1.0 for Windows XP (32-bit)

[Download] Windows Display Driver version 162.01 for CUDA Toolkit version 1.0

[Download] CUDA 1.0 windows Release Notes

and then im set?

 

[ThirdLeft152]

so for this to work...all i have to do is download
NVIDIA CUDA
Download] CUDA Toolkit version 1.0 for Windows XP (32-bit)

[Download] CUDA SDK version 1.0 for Windows XP (32-bit)

[Download] Windows Display Driver version 162.01 for CUDA Toolkit version 1.0

[Download] CUDA 1.0 windows Release Notes

and then im set?

i don't know very much about it, i just came across it researching something for work. i think that should do it though.