In the context of multi-user precoding, the idea behind vector perturbation (VP) lies in adding an integer vector to the data vector such that the overall transmit power is reduced, where the performance at the users end is consequently improved. In the literature, several techniques have been proposed to nd a quasi-optimum perturbing vector, where this process was represented as an integer lattice search problem. In this paper, we propose a parallel QRD-M encoder (PQRDME) that, besides attaining a quasi-optimum diversity order, leads to tremendous reduction in the latency of the vector perturbation stage. Based on the set grouping, the proposed encoder transforms the full tree-search of the conventional QRDME into partial trees that can be pipelined to increase the encoding throughput. We evaluate the proposed algorithm under several scenarios with both perfect channel state information (PCSI) and imperfect CSI (ICSI) at the transmitter side, where simulation results show robust performance when compared to the optimum encoder.