two 9800GX2 requirements Any specific things to look out for

Hi,

I’m looking at what is needed to get a system with two 9800GX2 to be able to use CUDA.

Is there any specific motherboard chipset that I should use for optimal performance or any motherboard with two pcie 2.0 16X would satisfy?
Would say, Intel X38, X48, Intel workstation chipsets be ok? or it must be a nvidia chipset?

Also, I was wondering if SLI is a requirement to get this working or the cards should be working independently. If so, do I need the SLI bridge connector?

Can I still get video display off one of the cards when running CUDA or I need a third video card for that sole purpose.

Thanks for any input!

Please see a little bit further down the page for considerations on which mobo to get.
http://forums.nvidia.com/index.php?showtopic=74491

To answer your questions that aren’t addressed in that post:
No, you don’t need SLI.
And yes, you can use the same GPU for CUDA and display. There are a large number of posts discussing this. Here is a recent one: http://forums.nvidia.com/index.php?showtop…4104&hl=display And the FAQ discusses it http://forums.nvidia.com/index.php?showtop…6286&hl=display

Thanks for the links. Read through them and have decided to stick with Intel chipset. I noticed this in the FAQ:

What is the maximum kernel execution time?

On Windows, individual GPU program launches have a maximum run time of around 5 seconds. Exceeding this time limit usually will cause a launch failure reported through the CUDA driver or the CUDA runtime, but in some cases can hang the entire machine, requiring a hard reset.

This is caused by the Windows “watchdog” timer that causes programs using the primary graphics adapter to time out if they run longer than the maximum allowed time.

For this reason it is recommended that CUDA is run on a GPU that is NOT attached to a display and does not have the Windows desktop extended onto it. In this case, the system must contain at least one NVIDIA GPU that serves as the primary graphics adapter.

I was wondering if this is going to be a major problem, how do people avoid this if there are only two PCIe 16X slots which are used up? Would it be possible to put in another 4X or 8X basic video card to run the OS on and leave the 4 GPUs for CUDA?

Is there anywhere that shows the Intel 5400 can work with 2 9800GX2? I seem to be unable to get it working in a Lenovo D10.

I’ve never had a problem. My longest kernel execution time is ~30 milliseconds. If you need to get around the 5s timeout with a single GPU, you can run in linux without X. You can add a 2nd inexpensive GPU, too: just be sure it is supported by the CUDA drivers if you plan to run it in windows.