The data rate of a multiple input, multiple output (MIMO) communication link can be improved if knowledge about the channel statistics is exploited at the transmitter. However, many realistic system scenarios require computationally prohibitive Monte Carlo methods in order to optimize the transmit covariance for maximization of the exact mutual information. This paper instead considers an approximative approach to maximize the performance of a communication link where covariance feedback is available at the transmitter. The algorithm presented is based on an asymptotic expression of the mutual information that allows for correlation at both the transmit and the receive side of the system. From simulations we demonstrate significant gain over beamforming and pure divversity schemes in many realistic system scenarios.