If one is to use 2 or 3 Tesla C1060s + a graphics card on the same motherboard, where
each of these will be connected through PCI-E x16 v2 slots that will function at x8, then
which graphics card with what GPU would be a perfect fit?
Would a GeForce 9800 GTX+ graphics card work?
Is it possible to exclude the GPU on the graphics card from the CUDA computations and how?
Or is it better to allow CUDA to use it, if necessary?
In that case, isn’t there a limit of using it for 5 seconds or something like that, while my CUDA
computations will take much longer.
So, what would be the best way to use the GPU on the graphics card in terms of helping CUDA,
given that the CUDA computations will take many minutes and possibly hours?
Does it make sense (and would it make a difference in terms of helping CUDA) for the graphics
card to be workstation quality, or would any CUDA-enabled GPU graphics card do the trick?