Hello,
I am interested in the following configuration:
Dell XPS 720 machine - has 2 PCI express x16 slots
Tesla C870 card for computation but is not intended to talk to a display
GeForce 8600 GTS card will be used for video
Woud this configuration work?
Are there any other similar configurations that anyone in your shop has tried or could otherwise recommend?
We intend to use CUDA as the programming environment and intend to install Windows XP Pro Service Pack 2.
I am using a Dell XPS 720 H2C with 2 8800GTX’s That is working supergood (after Dell replaced the CPU & Mobo because of stability problems)
Your config does not seem to differ too much, so I guess it will work ok. I am using Linux, but since XP is supported in CUDA I guess you will not have problems.
Are you using both the 8800GTX cards for graphics? Or are you using them for CUDA computations?
One is used for display, I also removed the SLI brigde.
At the moment my computations only need 1 GPU. I have not yet tested if using the second one is faster than the first one, and I am not selecting a GPU to do CUDA in my programs, so I guess that I am using the GPU that is also used for display.
Later I will be selecting the second card, so I have all memory for use, but for now I have no need yet.
Do you know the big problem on win32 platform if you have only 1 G80 card and the machine will be dead if your computing time more than 5 second ?
I have not tried it yet, so i want to know whether we can avoid it.
I have not encoutered the problem yet (I thought so in the beginning, but it was not). But I do know that the second card will not have the trouble in Liunx at least. And otherwise I can still use in commandline mode.
As far as I have read here, the second card in Windows also does not have the trouble of the 5 sec limitation.
I think you will have BIG problem spaces when you encouter the limitation though (or design flaws in your code ;))