Abstract: The generalized minimum variance (GMV), one of the most important methods for control system performance assessment, is used to introduce the theory of controller design. The theory combines the system error and the controller output based on GMV benchmarking. By this method, we will not only take into account the system tracking error, but also pay attention to the stability of the controller output when designing the controller. Reasonable choices of error weighting and control weighting are very important to minimize the tracking error variance and guarantee closed loop stability. This paper describes the principles for the configuration of two weightings in detail. The effects are demonstrated through single loop control system, traditional cascade control system and parallel cascade system, and are compared with the results calculated by PID, the superiority of GMV benchmarking for controller design is proved.