I am developing a conjugate gradient (CG) solver with GPU, GeForce GTX 480. The problem which seemed to occur randomly is that the job sometimes got stuck. There is no output anymore, btu the status show “R” for this job. This has annoyed me for a long time. I don’t know wheter my codes has some bugs or not becouse my program works well at the start. After a long-time computation, the CG sover stopped giving me the output. It seems that the codes fall into an infinite loop. I have tested my solver many times, and it can give me the right answer with CG. is it possible that the GPU card is too fragile to run a long-time job ?
I (gladly) have no personal experience with this, but you might be interested in this thread.