@OP: please, clarify the question.
Ganado's guess sounds plausible. It is very typical for some laptops to have two GPU's; relatively weak IGP in the CPU and stronger discrete GPU. The default is to use IGP. That produces less heat and consumes less power (battery). However, for the heavier graphics one wants to use the discrete GPU.
I do believe that OS X does this automatically.
I've seen NVidia's Optimus used in Linux. See
https://wiki.archlinux.org/index.php/Bumblebee
Thing is, the application either uses one GPU for graphics, or another GPU for graphics. There is nothing special, as far as the program is conserned. The OS allocates
some hardware for the process.
jonnin did mention GPGPU (General Purpose computing in GPU). OpenCL and NVidia's CUDA are frameworks to generate code for compute operations that will execute in GPU, rather than in CPU. (PhysX was such as well.) All algorithms do not fit into GPUs, at all or not with benefit.
Thing is, application has to be specifically written to do GPGPU. If Windows offers "option", then either the application already has necessary features, or the option is not about GPGPU.