most effective way to get a mobile CUDA gpu

Accelerated Computing CUDA CUDA Programming and Performance
21

It’ll still work in Linux, right?

edit - btw - I agree that having NVIDIA integrated graphics might be nicer - got any motherboard flavor suggestions? Thanks!

22

FWIW - Bleh, the integrated Intel video on my new board isn’t playing too well with the NVIDIA card. I told the BIOS to use the integrated display only. The display turned black on any DVI output. Had to use the BIOS reset jumper to get things going again.

I’m not excluding the possibility of Inter GFX + NVIDIA card as GPU, but it’s not idiot proof.

23

nvidia is soon going to relase nforce 730i, with IGP for 775…

24

Ok, so by tomorrow i have to present a feasibility study on CUDA to my professor. Nothing too fancy, just a 3 pages, latex written, presentation on CUDA and its capabilities. I have yet to outline 2 details please:

a) Actually fast acess shared memory is 16 kb per gpu (-256 byte for the system). Is it the same on all GPUs? does nvidia plan to expand it?

B) Since my professor told me that i could get a 200 euro bonus (it is HUGE for our standards) for the “upgrade of needed computing devices”, i have to present a budget for the pc:

  • Msi P45 Platinum: 147 euro

I decided for an atx since G45 matx MotherBoards have many issues

  • Intel Q6600: 170 euro

  • Fancy CPU cooler: 50 euro
    For OC

  • 4 gb of 1066 mhz ddr2: 96 euro

  • Case+fans: 90 euro

  • Antec Signature 650 W PSU: 145 euro

Quality PSU for stability and OC when needed

  • Geforce 8400 512 mb: 33 euro

I choose a budget card becasue I will need another graphic card to run X, it would be a real pity to have X crash while using in remote and not being able to code anymore. I will still buy a fancy G200 based graphic card, but when i’ll have learnt enough on CUDA.

Total: 726 euro

25

16 kb of shared mem is per multiprocessor. In all versions seen so far it has been the same. And they don’t speak freely about upcoming hw…