Conventional PID controllers as well as linear model predictive controllers often do not perform satisfactorily for dual composition control of, distillation columns because of severe loop interactions and a high degree of nonlinearity. Attention has recently been diverted to the development of non-linear controllers for such difficult-to-control systems. Only limited experimental applications have been reported in the literature. The performance of the non-linear analytical model predictive control (NAMPC) has been compared with that of the generic model control (GMC), for the control of a pilot-scale distillation column for both servo and regulatory problems. Two single-loop GMC controllers with feedforward were found to perform satisfactorily when subjected to small servo and load disturbances. Multivariable GMC as well as NAMPC were successful in tracking setpoint changes as well as eliminating load disturbances. Both algorithms were robust and stable even in the presence of large disturbances.