MDS reduces dimensions keeping the distance between the points. PCA reduces dimensions reducing variance as well. They are different approaches, bade for different things. You may want to use PCA when you work with linear models, for example.