Can I use Nvidia 750i sli motherboard?

I am trying to set up a developer system to try out CUDA development. My primary machine has an ATI FirePro v5700 and I’ve heard not to mix ATI and Nvidia cards with Windows. I have a second machine with an Intel Core2 quad cpu but the motherboard has only one PCIe x16 slot. I’ve done some checking and I can get another Asus LGA 775 motherboard (about the same age as the motherboard in the machine – current has a P45 chip) with an Nvidia 750i chip and 2 double spaced PCIe x16 slots. I develop in Visual Studio and am intending to upgrade the machine to Windows 7 when I make the changes to be able to use Parallel Nsight. I have a GTX 465 on the way. I know that I will need to upgrade the power supply, but I can use the same processor and memory (8GB) with the Asus 750i sli motherboard. I believe I will have one of the PCI x16 slots operating at 2.0 x16 and the other at 2.0 x8.

My other choice would be to change out the motherboard, cpu, and memory for an AMD AM3 socket with an 890FX chip. That would add about $400 to the cost. Does anyone know if my 750i plan will work? I can’t find any information on using a 750i chip with CUDA. I appreciate any information.


A 750i motherboard should be fine, as long as you have a CUDA-capable card plugged into a PCIE slot. I use a 650i SLI motherboard (Abit FP-IN9) as one of my CUDA machines (I have used Tesla 10 series and 20 series GPUs in this mobo.) At the moment I have an older Quadro card plugged into one of the PCIE slots for display and a Tesla 20 series GPU plugged into the 2nd PCIE slot. (in this config I only get 1.0 x8 electrical to each slot, but it works fine.) Since your planned GTX465 card has both display and CUDA capability in one, you’ll only need one PCIE slot.

Thanks very much for the information. The board I was considering does have two PCIE x16 2.0 slots (operating at 2.0 x16 and 2.0 x8). I was planning to use Parallel Nsight under Windows 7 and read that I would need another video card for display to be able to do local debugging on the compute card, so I was planning on using the current card in the machine for that. Thanks again for your reply.