Tesla C2050 bad install or something else?

We’ve recently installed a Tesla C2050 compute board on a system running Ubuntu 10.04. We proceeded to install CUDA 4.0 and the 270.41.19 drivers. Unfortunately all CUDA programs now result in a segmentation fault from cuCtxCreate as the attached test program demonstrates. A sample valgrind output is also attached. Does this sound like it could be a bad install, any ideas to try or things to look at would be greatly appreciated.


What compute mode is your card in? Try using nvidia-smi to check this out.

Also, you should check the CUresult returned for each API call. That may give a bit more info.