CUDA Hw requirements

Hi, I have asus p6t6 ws revolution motherboard that has PCIe 2.0. Equipped with 24 GB of RAM 1066 and i7 940 cpu. I would like to know if CUDA 3.0 cards will work on 2.0 interface, will there be performance decrease (significant), is i7 CPU that I have sufficient enough, and if not which cards exist for 2.0 interface that I can use.
In addition I have PSU 560 W but I think I would need a new one if I buy better GPU.
One more question is after I plug GPU inside will it use automatically for floating point calculation or I need to configure it (example use of Mathematica, for Excel I read somewhere that I need to adapt worksheet), same for SLI , will it be harder for implementation.
I need the most FLOPS I can get on this HW, if you think it is better to buy something new please say so.
Advise is appreciated
Thanks

I’m assuming that CUDA 3.0 cards you mean NVIDIA GPUs that are capable of PCI-Express 3.0 speeds. Cards that are PCI-E 3.0 compliant will run at PCI-E 2.0 speeds on older motherboards. The difference would be in terms of bandwidth – how fast you are able to transfer data to/from the GPU. For PCI-E 3.0 x16 vs PCI-E 2.0 x16 that would mean about a factor of 2 difference.

There are some applications that do take advantage of the GPU processing, usually by changing specific settings within the app and depending on whatever functionality is supported to be offloaded to GPUs. You should check exactly what the benefits are with Mathematica, as I do not know those offhand.

Up until now I had never of CUDA use in Excel, but it seems that through some add-on it is possible: http://blog.quantalea.net/cuda-scripting-inside-excel-part-i/.

As far as another power supply, MAXIMUM power use of a high end Kepler GPU (e.g. GTX Titan) is up to 250W (without overclocking). So if you have enough headroom in your power supply, you might not need a new one.

Your 1st Gen i7 CPU should adequate, but it depends on what you want to accomplish. Is there any CUDA application/algorithm in particular that you’re trying to work on? For example, if you’d plan on doing real time 1080p video processing, that would be one application where you’d probably desire a fast solid state drive to deal with the data flow. If you list exactly what your intentions are with a GPU, youu might get more specific advice.

Hi vacaloca, thanks for your response. It cleared a lot of things for me.
Regarding intentions we use most of the time excel to calculate some optimizations that uses floating point calculations, it has nothing to do with video, hope this gives you more information.