Cannot run my CUDA Fortran executables on new Titan Black

Hi,

I recently purchased a Titan Black. I have installed it in a system running Windows 7 64. I have installed the graphics card drivers and CUDA 5.5. I tested the card by running some of the CUDA Samples (particles, deviceQuery, etc.). The samples run with no problem.

When I try to run one of my own executable I get an error message that the program has failed and will be closed. I am compiling with PVF 13.10. Of course the executables run just fine on my laptop (GTX 660M). I have also successfully run many of my codes on K20s and K40s.

I do not have a license for PVF on the computer that has the Titan Black, so I cannot try to compile the code on that system. I have to compile on my laptop (running Windows 8 64 with CUDA 5.5) and then transfer the executable to the system with the Titan Black.

Maybe someone from Nvidia can see if they can reproduce the problem, if yes, I can send a simple CUDA Fortran code for testing.

Thank you for any help, let me know if any more information is needed.

Kirk

Are you a student/facutly of a US institution w/ an .edu address? Seems they are providing a free license of that PVF compiler at the moment for that user base: https://www.pgroup.com/purchase/freepvf.php

That would help eliminate some sort of compilation issue and allow you to debug on the local system.

I am not a student at a US institution unfortunately.

Actually, doesn’t matter if you are, you can sign up for a 30-day trial for that specific suite regardless – https://www.pgroup.com/support/trial.htm

Depending on how you compiled the application, it is likely to have dependencies on a number of runtime components that need to be distributed along with the application. These redistributable components should be in the REDIST subdirectory of your PGI toolchain installation on the machine you built the application on, for example C:\Program Files\PGI\win64\14.3\REDIST (accordingly for other versions of the PGI toolchain).