NVIDIA API mismatch error


I am testing one of the examples in the gpu package on our cluster and I get the below message while running on the GPU node.

Error: API mismatch: the NVIDIA kernel module has version 285.05.33,
but this NVIDIA driver component has version 256.40. Please make
sure that the kernel module and all NVIDIA driver components
have the same version.
Cuda driver error 100 in call at file ‘geryon/nvd_device.h’ in line 232.

– Mo Zhang

Sounds like a problem on your machine with CUDA software,
not with LAMMPS.


Steve is correct. This is a problem with your NVIDIA drivers, not LAMMPS. There are multiple components to the NVIDIA drivers: The kernel modules, the graphics libraries, the cuda libraries, etfc.

Based on your error message, your kernel module is at version 285.05.33, which is newer than the driver component, which is at 256.40. These should be the same version. This is not a LAMMPS issue, so you should contact your local system administrator and tell him your NVIDIA driver versions are out of sync.