I’m having some trouble with a CUDA program I’ve written. I’m getting some strange behaviour from my CUDA program. It compiles, and sometimes it runs, but most of the time it causes the PC to trash, giving a black screen, and graphics fan turns immediately to high. I have to do a hard power-off and even when the PC is turned on again, it will not run any 3D application like a game. Starting a game will result in the same thing, black screen, and fan goes immediately to high. I got functionality back by moving the card to another PCIe port in the motherboard (and then later back again to the original). Simply removing driver and reinstall was not enough.
I worry if my program is going over some hardware limitations, causing these crashes. I launch a kernel with 1440 blocks of 64 threads, so
myKernelOS <<< 1440, 64 >>> (arguments)
I have checked the ptxas info when compiling and if I add up the values it writes im using 176 registers and about 1200bytes of cmem. The card im using is a Geforce RTX 2080. So does this exceed any hardware limitations?