How can I use my GPU as a multi-core CPU?

I have a basically serial function which needs to be evaluated. Now I use parallel computing toolbox to create workers on my multi-core CPU to do the jobs. If I want to make full use of my Geforece 970 GPU in the way that I send a copy of the function to the cores of the GPU just like how I do with the CPU, how do I write the script? I have tried gpuArray and etc.., the speed is much slower. But if each of the over 1000 stream cores can be fed with a job simultaneously, the overall speed-up would be OK. Thanks for any attention and help to my question.

[Edric Ellis]

The main means of giving a GPU enough work to do in an efficient manner is through vectorisation. If gpuArray is slower than the CPU, it's often because your code isn't sufficiently vectorised. Could you give us a simple self-contained example of what you're trying to speed up?