Are you asking why CUDA programs stop running unexpectedly when you run them with gdm running?
If you are, the reason is that the display driver includes a watchdog timer which will not allow CUDA programs to run for more than 5 seconds at a time. You have to choices to work around this: either reduce the running time of each kernel (make the code faster or reduce the amount of work per kernel launch and run the kernel more times), or use a card without a display driver attached (ie. no gdm, which you already found by yourself).
If you are asking about something else, then perhaps try asking your question in a slightly easier to understand manner, because it is extremely cryptic and hard to follow.
Your post is still very difficult to interpret. I would actually suggest you write in your native language and run it through something like google translate.
If I understand part of your question correctly, you want to know if you should have X11/gdm running in the background? I would recommend closing it.
If you want to have X11 running, use a second graphics card.
about that… on my laptop (one gpu obviously, single SM) I have this issue that my cuda programs won’t give the correct results (they do execute but the output is whatever…). This problem is fixed if I use a tty to execute (that is, temporarily take the weight of the graphics off of the GPU).
the odd about this is that the sample nvidia programs that include graphics as well (mine don’t) run just fine, executed from gnome, with no complains or obvious performance degradation…
any ideas?
ok, I know this thread is old, but I still thought I should give it a try