This paper deals with a maximum torque per voltage control method of the interior permanent magnet synchronous motor, which aims at the high demands for convergence speed and torque accuracy in electric vehicles applications. First, a Newton downhill method is adopted in solving the nonlinear torque and voltage constraints within few iterations, which guarantees a high iterative convergence speed. Second, motor parameters used in the iteration are identified by a hybrid affine projection algorithm(APA) with dual sample time. In this approach, the effect of temperature ignored by traditional methods on the variation of resistance parameter is considered, which ensures the accuracy of the current set points. Finally, the effectiveness of this method is verified by simulation and experimental results on an 120kW IPMSM.