Problems connecting Cuda to Tesla graphics-card

I’m trying to connect Cuda to my graphics-card, but I get all sorts of issues:

  1. On Visual C++, when I try to run the following line: cudaError_t error_id = cudaGetDeviceCount(&deviceCount); error_id gets the value cudaErrorInsufficientDriver.
  2. When I open Nsight, it informs me that the driver's version should be 343.98.
  3. When I try to install driver version 343.98, I get the following message: "This graphics driver could not find compatible graphics hardware".

Could you please help me solve the problem?

Some technical info:

  • The graphics-card is Tesla C1060.
  • The OS is Windows 7 64bit.
  • Visual C++'s version is 2010.
  • Cuda Toolkit's version is 8.0 (We need this version in order to be compatible with Python 3's Tensor-flow package).
  • The current driver installed on the PC is of version 342.00.
  • Nsight monitor's version
  • Nvidia WNI 2.18.0.
  • The new driver's full name is: "343.98_desktop_notebook_quadro_tesla_win8_winvista_win7_64bit.exe".
  • ----------------------------

    Thanks in advance,

    C1060 is too old to be useful for Tensorflow (TF).

    CUDA 8 requires a driver version of 375.xx or newer (for CUDA 8 GA2). Your 342.xx driver won’t work with CUDA 8, and that is the gist of the error message you are receiving.

    You won’t be able to update to a newer driver because driver support for cc1.x CUDA devices was dropped after 342.xx drivers. The last CUDA version that worked with that GPU was CUDA 6.5

    Furthermore, to use TF, it will require the CUDNN library. The CUDNN library is only supported by devices of compute capability cc3.0 or newer.

    If you want to use TF with CUDA GPU acceleration, you’ll need to update to a cc3.0 or newer GPU. There is no other possibility (unless you intend to rewrite TF).

    Thank you very much for the detailed answer.