Dedicated video card for CUDA?

Accelerated Computing CUDA CUDA Programming and Performance
1

I’m using only one video card both for CUDA computing and video output for my monitor. Is this a correct configuration?
What should I be careful?
Is there the risk that video memory used for video output could corrupt the memory I use in CUDA computing?

2

yes, there is a risk, since the OS video display task takes priority over a CUDA kernel.

Your kernels can be killed or memory taken by force… you need to always check error result codes.

BUT in practice this is not a problem. The display uses a fixed size of RAM and therefore never needs to override a CUDA context’s allocations.

The very rare but possible exception is when a user increases screen resolution or adds new display monitors while your CUDA context is active in a different process.

3

Bugs in the kernel may even cause a BSOD.

[url=“http://forums.nvidia.com/index.php?showtopic=94524”]http://forums.nvidia.com/index.php?showtopic=94524[/url]
[url=“http://forums.nvidia.com/index.php?showtopic=91635”]http://forums.nvidia.com/index.php?showtopic=91635[/url]

Other than that, I don’t see why it should be a big problem. It won’t damage the GPU.