I am using Ubuntu 12.10 with CUDA 5.5 installed on a laptop which have an Optimus compatible GPU. So the displays are using the integrated Intel GPU and I can use the Nvidia GPU (GeForce 525M) to run CUDA programs with optirun. My question is that, how can I disable the run time limit on kernels? (I know the risk behind it, but this option is critical only if the displays using the Nvidia GPU.)
After some more research I found that this technique works:
It is working, because the X.org server set up the watchdog timer, so if the bumblee demon don't use it, than there is no watchdog present in the system.