In this letter, uplink sum rate of large-scale multiuser MIMO systems with maximum ratio combining receivers is analyzed. To be practical, no perfect channel state information is assumed, and both large scale path loss and small scale Rayleigh fading are considered. The large scale path losses experienced by multiple users are also assumed to be non-identical due to different locations of the users. A tight upper bound is derived in closed-form. Its tightness is verified by simulations. It is shown that the proposed upper bound is tighter than other bounds and the tightness is further improved when the system becomes larger, i.e., more antennas are deployed at the receiver. It thus provides a solid foundation for system design and optimization.