You can't. First requirement is that you have the parallel toolbox. Second requirement is that you need a nVidia graphics card. There is a long list of nVidia graphics cards that are supported in the parallel toolbox documentation. Lastly, if you have a nVidia graphics card, which you stated you do not, you need to load the CUDA driver from nVidia for that graphics card.
There is currently requests to the Mathworks parallel toolbox development team to add support for other graphics cards, like amd, especially since newer professional computers, like the ne Mac Pro are coming out with these.
If you have a windows PC, and really need this function, looks like your only option would be to swap out the amd for an nVidia
