The ability to use two Nvidia GPU's connected together in SLI mode is not possible with the Parallel Computing Toolbox 6.0 (R2012a). The reason for this is because the CUDA engine still sees the two GPU's as separate devices even though they are connected via SLI. This is documented in section 3.2.11.3 of the CUDA programming guide.
If you want to use these two cards simultaneously, you can use MATLABPOOL to create 2 MATLAB workers, call GPUDEVICE to have each worker select a separate GPU, and then perform each set of calculations independently.
