Permanent magnet motor is the mainstream motor in the industrial servo field for its high efficiency, high power density, and large torque and inertia ratio. Servo control technology is the key to give full play to the advantages of permanent magnet motor and improve the performance of servo system. At present, permanent magnet servo system mostly adopts multiloop cascaded proportionalintegral (PI) controllers. However, due to the lag effect of the integrator, the dynamic response speed of PI is slow and the antiinterference ability is poor, which is difficult to meet the dynamic and static performance requirements of highperformance servo control such as mechanical arm and precise machining. Therefore, a composite model predictive control strategy combining generalized model predict control and finite control set model perdict control is proposed. In addition, a low computational complexity implementation method of generalized model predictive control and a mechanical parameter estimation method are also proposed. The experimental results show that the proposed composite model predictive control can improve the dynamic response speed and load disturbance rejection capability of the permanent magnet servo motor.