CUDA on second GPU

Accelerated Computing CUDA CUDA Programming and Performance
#1

Hi,

I want to run CUDA on a 260GTX.(used only for computation, not for display)
I also have an integrated ATI GPU, and want to use it for the display, to keep the cuda device as free as possible.

The monitor is connected to ATI and works fine, but the 260GTX is not seen by the CUDA applications. If I set “Extend the Windows desktop” to the GTX device, then the CUDA applications run properly.
But I want to keep the GTX only for computations.

Is windows somehow disabling the 260GTX? Is there a switch which can enable it for all the time, regardless it’s used for display or not?

Thanks, Bogdan

#2

Maybe you can switch the PCI-E slot, I mean plug the ATI board on Nvidia’s one, so does the other.

It makes the MB allocate different memory loation for these two device.

And you can also report it to Nvida for them to check if its something like driver problem.

#3

Hi,

thanks for the answer.

The Ati GPU is actually onboard, so I cannot switch places. I had to set from the bios to have the monitor on the onboard GPU (default it was on NVIDIA card).

#4

Ok, let’s put it another way. Is this the expected behaviour?(if yes, at least I won’t spend time trying to figure out how to use second NVIDIA GPU (on windows XP) without extending the desktop to it.)