A new scheduling strategy was proposed for the downlink of MIMO channels with zero-forcing beamforming. As the number of simultaneously supportable users at the same frequency is limited by the number of transmit and receive antennas, the base station may select a subset of users in a cell to serve. Exhaustive search for the optimal user set, however, is computationally prohibitive. With revision to the distance criterion, we proposed a novel sub-optimal scheduling algorithm for MIMO downlink channels. The proposed algorithm is a complementary method of the Frobenius norm-based algorithm, the former emphasizes on the uncorrelation of the users' channels, and the later is mainly on the energy of the channels. Combining these two algorithms can achieve a tradeoff between the system throughput and the computational complexity. Simulation results show that the system throughput obtained by this algorithm is very close to that of capacity-based algorithm and outperforms that of the distance-based algorithm.