According to the Nvidia sitehttp://developer.nvidia.com/cuda/cuda-gpus
the Quadro FX 3800 is compute capability 1.3, so use Tesla. (This is a misnomer; by Tesla we really mean pre-Fermi.)
In some (older?) versions of MKL there is the -lguide library. If it is not needed, just remove it. You can lookup the required MKL libraries in the MKL documentation on Intel's website.