RF is now a standard to effectively analyze a large number of variables, of many different types, with no previous variable selection process. It is not parametric, and in particular for survival target it does not assume the proportional risks assumption.