deviceQuery return fatal error after upgrading cuda toolkit from 7.5 to 8

after upgrading toolkit and executed make command in ~/NVIDIA_CUDA-8.0_Samples, I cannot locate deviceQuery file under ~/NVIDIA_CUDA-8.0_Samples, but I can locate it under ~/NVIDIA_CUDA-8.0_Samples/1_Utilities/deviceQuery folder. When I run it, it indicate errors like below:
devicequery cudagetdeivcecount returned 30
-> unknown error
Result=fatal.

Can anybody help me? btw: when I installed toolkit 7.5 in this machine last year, every thing was ok.
MY gpu is k40c, os is rhel 6.5, HPE deep learning physical machine.

Thanks in advance.

To whom concerned:
this issue is solved by upgrading gpu driver (up-to-date version). Hint: you can use nvidia-smi to check if the version of gpu driver is consistent with that of cuda toolkit.

it will be helpful if Nvidia can list gpu driver version requirement in the cuda toolkit pre-installation checks.