I'm having an issue mining scrypt with two GPU's: every time I start cudaMiner, one of them (not always the same) seems to return results which mostly doesn't validate. This is tested on Linux, using cudaMiner master, tested both on Z14x14 and T14x32. When I mine using a single card (selected with -d) all results validate... Any idea?
this is a long shot: Would it help to downgrade the video driver to the exact version hat shipped with the CUDA 5.5 toolkit download from nVidia? Of course if your video cards are newer than this driver release, this is a no-go, as the driver would not recognize them ;-/
I'm using the NVIDIA drivers from the Debian repositories, so downgrading isn't that easy... That said, I'm currently using the 319 driver series, which seems like the earliest one supported by CUDA 5.5. I've tried upgrading to the 331 driver series, but that doesn't change a thing.
I think your bug report is the one that made my mind go
The CUDA constant memory (the c_N loop trip count, etc...) of most CUDA kernels is only initialized properly for the first GPU (use of a single static variable to mark initialization instead of a thread-specific static variable). Which explains the majority of the crashes people are seeing with multi-GPU. Thank you. The Fermi owners use a kernel that doesn't yet make use of such constants, and hence the multi-GPU support is working fine for them.
So this is also on the FIXME list for tonight.
However I think that in your case where you run two cudaminer instances this cannot be the root cause. So we will have to keep looking.
Christian
Instead of K10x16 with the last version, the new one is giving me better results with K10x28. 