Program for GPU

Is there а vertion of Prime95 for GPU??

[akruppa]

This has been discussed a few times before, see for example

http://www.mersenneforum.org/showthread.php?t=432
http://www.mersenneforum.org/showthread.php?t=3309

Alex

[xilman]

Originally Posted by akruppa

This has been discussed a few times before, see for example

http://www.mersenneforum.org/showthread.php?t=432
http://www.mersenneforum.org/showthread.php?t=3309

Alex

But there is hope for the future. I just installed a nVidia Tesla C1060. It has 30 dp FPUs as well as 240 sp engines.

I've made no attempt to run dp FFTs on it yet so can't comment on their speed vis a vis a common x86-64 cpu. It is impressively fast on some other computations that can be easily converted to 32-bit SIMD algorithms.


Paul

[jasonp]

See also here

[frmky]

Originally Posted by xilman

But there is hope for the future. I just installed a nVidia Tesla C1060. It has 30 dp FPUs as well as 240 sp engines.

I've made no attempt to run dp FFTs on it yet so can't comment on their speed vis a vis a common x86-64 cpu. It is impressively fast on some other computations that can be easily converted to 32-bit SIMD algorithms.

Paul

Note that I have (almost exclusive) access to a Tesla S1070, which is basically four C1060's in a box, if you wish to investigate using multiple GPUs in a calculation.

Greg

[xilman]

Originally Posted by frmky

Note that I have (almost exclusive) access to a Tesla S1070, which is basically four C1060's in a box, if you wish to investigate using multiple GPUs in a calculation.

Greg

Thanks, I may take you up on that offer some time.

However, I already have multiple GPUs in the system in that a Quadro 1700 had to be removed to make room for the Tesla. The remaining card is still CUDA-capable, though much less powerful than the Tesla. The SDK's multi-GPU example runs just fine.

Paul