As I understand it there are good reasons the SD is often more useful. For instance, if you are dealing with a large number of samples from your distribution, and want to apply the central limit theorem, it's the SD, not the MAD, that you need to know.