Apart from the basic voltage vectors of a inverter, virtual vectors can be generated by discrete space vector modulation (DSVM) technique. Thus, DSVM based-model predictive control (MPC) can achieve reduced torque ripples and current harmonics for a permanent magnet synchronous machine (PMSM) drive. However, the computational burden is also significantly increased due to a large number of candidate voltage vectors. In this article, a novel DSVM-based MPC is proposed, which can select the optimal voltage vector and calculate three-phase duty ratios in an efficient way without enumerating all the candidate voltage vectors, hence achieving good performance with low computational burden. The effectiveness of the proposed method is validated by the presented experimental results.