Same code works on Tesla, doesn't work on Geforce 630


I wrote a piece of cuda code. It works well on Tesla, both C2070 and K20.
But when I run it on Geforce 630, it failed.
Not exactly failed, no error was observed by cudaGetLastError(), but the result is just not correct.
Anyone get experience on this?


Can you give us the piece of code you think is failing?

If you run the X server or the windows interface on the same card as cuda, there is a limit to how long a kernel can be executed, before is killed. Which might this.