screen turn to black and kernel stop coud_memcheck: 0 error

When I run my CUDA program, my GPU load get to be 100%, and my screen go black after several seconds. Then the kernel stop, and no kernel will be called any more,but host functions run as normal.

My environment is Win7, single card GTX 460. The kernel functions are not complex, each of them will not last more than 3 seconds. They are called by host functions repeatedly. For the first several round, they run ok.

When I use cuda_memcheck version 4.1 to detect the memory access error, it shows 0 error.
Do you know why?

thank you for your interest~

Does your code use up a lot of GPU memory ?
(does the amount of GPU memory being used increase with each kernel launched by a host function)

You also say the kernels stop, what error code do they return ?

All the Best

No, it only used 20% of the device memory.

The cudaGetLastError() returned an “unknown error”.