The tradeoff between sum-rate and fairness for MIMO downlink communications employing dirty paper coding (DPC) is investigated in this paper. More specifically, we first apply the luml1 -norm fairness measure to formulate the tradeoff as a certain optimization problem, which is unfortunately nonconvex and cannot be efficiently solved. To overcome the difficulty, we make use of the uplink-downlink duality to transform the problem back and forth between uplink and downlink communications. An efficient, iterative waterfilling based algorithm is then provided to yield achievable points on the sum rate-fairness tradeoff. Simulation results show surprisingly that the proposed DPC-based approach offers a significant gain in the achievable sum-rates for a wide range of fairness values, when compared to successive zero-forcing DPC-based coding schemes.