Recently, I have encountered some problems: my computer (ivybridge) uses the built-in gpu (intel igpu HD4000) for normal use of the screen, and I run CUDA programs for computing on a discrete Nvidia GT 640 on my machine. The problem was that when iGPU is displayed, CUDA will not be able to detect the nvidia card , and the nvidia drivers will not load at all.
Keep in mind that when using nvidia window drivers for display devices, there are confirmed problems (mostly around concurrency), and you also want to use CUDA. These problems can be resolved by using Intel gpu as the display (thus loading only the smallest part of the Nvidia drivers mentioned) and in many cases nvidia drivers for Linux distributions are not included.
After many troubles with poor software, such as virtu mvp, I found the solution really simple, but was not found anywhere.
I post it here in the hope that this will help. Please post any comments to make the answers more complete, even if this fails for some PC configurations, so we can clarify the instructions for a wider range of scenarios.
source
share