In Multi-user, Multiple-Input, Multiple-Output (MU-MIMO) wireless cellular where Quality of Service (QoS) is interference-limited, a variety of power control, precoding and decoding non-linear problems exist. Even though the sum capacity of centralized schemes (either at transmission or reception) has already been computed, the search for efficient practical schemes which can achieve this capacity is still ongoing. This paper examines the precoder and decoder joint design from a new perspective and proposes a new non-iterative architecture, which is novel, practical, and low in complexity. This new perspective is referred to as embedded power loading. The goal is to obtain a practical design for the power and bit loading, which maximizes the sum rate with user Bit Error Rate (BER) requirements in the Broadcast Channel (BC). By means of simulation results, this work shows that, although heuristically designed, our proposal achieves approximately the sum rate with a substantial reduction of complexity.