…got another question:
A piece of code that runs fine on a G80 crashes randomly but reproducibly on the T10P. Not immediately but after a number of iterations or if I run the executable repeatedly. The “post-crash” behavior also varies from just a hanging process immune to SIGKILL with 100% cpu utilization to a complete system freeze. Always without any error in the various logs. At least one time I got a cudaError 10 (“invalid device ordinal”) from one kernel just before crashing, however I think I don’t understand what that error means. Does anybody have a helpful hint?
As for my other question, I’m using driver 177.10, tk release 2.0.1640.