cornerSubPix in OpenCV: purpose and application

I tried to understand the cornerSubPix API in OpenCV for its idea behind and usefulness. I read the explanation in the link and can't understand how it works and how it is useful. Can somebody explain me how it works and how it is useful at corner refinement? I checked my corner detection application for without cornerSubPix and with cornerSubPix. Not much difference in output image, but it takes time to process. Thanks