What is the 5 second limitation in the GPU attached a monitor?

I always got the launch-time-out error in the GPU attached a monitor.(I have four GPUs)
Does the 5s limitation indicate the time of a single kernel or many kernels running continuously?
The running time of all kernels in my program is less than 100ms, shouldn’t exceed the limitation.
I tried to remove the monitor, logged in from other computer after rebooting to run the program. The program still reported the sam error.

How can i do to solve the problem?

This regards a single kernel. A loop of short kernels may go beyond 5 seconds.