We consider a multiple-access multiple-input-multiple- output (MIMO) system with minimum mean square error successive interference cancelation (MMSE-SIC) receiver. Spatial multiplexing is employed with some of the users having target-rate requirements while other users are working with best-effort data rates. We show that the best-effort users can achieve higher rates when the data streams of any user are not restricted to be decoded in a consecutive order and when time-sharing between different decoding orders is allowed. Two scenarios are considered. In the first scenario, we assume that the power transmitted from each antenna is fixed, and hence the optimization is done over rate and decoding order only. In the second scenario, the optimization is done jointly over rate, power, and decoding order. We show that in both scenarios the problem can be formulated as a convex optimization problem, and hence efficient algorithms can be used to obtain the globally optimal solution. We also show that in some cases it is possible to reduce the power required to obtain the optimal solution, and we show that the power minimization problem is convex. We also propose an efficient algorithm to obtain a sufficient set of decoding orders (corner points) and time-sharing factors, to obtain the optimal solution.