Sorry, I’m a newbie but I was wondering if it is possible to cluster GeForce cards for some HPC gain? My CUDA application works fine with single precision on my GeForce GTX 460M and I’m pretty sure most of my processing bottleneck is just a need for more threads (it’s a relatively low memory program). So I’m thinking theoretically I could throw like, 10 GeForce cards at my program and see some nice speedup? This is assuming I can find a suitible chassis, enough CPU cores, PCIe slots, etc. I was just wondering if there was any sort of inherent property of the GeForce series that puts a limit as to how many GeForce cards I can cluster. Thank you!
The main differences between Tesla and Geforce are more double precision hardware, better reliability (ECC, more testing, better quality board components), and more memory. If none of these matter to you, nothing is stopping
you from ganging as many Geforce cards together as you want. HPC systems typically care about the above points and use Tesla.
I’ve operated 4 GeForce GTX 295 cards in one computer, and it worked fine. Of the 20 or so GeForce cards I’ve worked with over the past 5 years, I’ve only had 2 fail completely, and maybe an occasional glitch in some of the others.
Keep in mind there is nothing magical about a “cluster” of GPUs. You simply have a lot of independent CUDA devices with different degrees of connectivity (PCI-Express bus, ethernet, etc.) that you have coordinate in software. The programming model is identical between Tesla and GeForce.
Thanks for the responses! Exactly what I wanted to hear.