Hi. I wrote my first CUDA application last night! :) But have already encountered timeouts caused by the 5-sec limitation. :( I have broken the problem into pieces to ensure all is working fine, which it is (currently 9x speedup over 3.2GHz CPU), but it would be nice to not have to divide the problem. I read in the release notes that if I got a second graphics card that there would no longer be the 5-sec limitation. Do others have experience with this? I am currently running the kernel on a 8800GTX card and there are only x1 PCI-e and PCI slots left on this motherboard. The OS is win XP.
I want to hold off replacing the motherboard and getting another high-priced card until I’m comfortable with CUDA. I’ve only had these for a couple of weeks and by the time I learn CUDA NVIDIA will have likely released a new product anyway :)
In the mean time, is there a way to disable the timeout in software?
If not, would someone please suggest a graphics card that I could use as the primary leaving the 8800GTX as a CUDA only card? The graphics card has to support 1920x1200 over DVI. I looked quite a while this morning and the only PCI NVIDIA cards that I could find maxed out at 1600x1200 over DVI.