In this paper, we consider the energy-efficient power allocation in the downlink of a large-scale multiple-input multiple-output (MIMO) system under both quality of service (QoS) constraints and maximum transmit power constraints. With these two constraints satisfied, the resource allocation strategies are proposed, which aim to maximize the energy-efficiency defined as the successfully delivered information bits per Joule energy consumed (in bits/Joule). Moreover, the power consumption consists of both transmit power and circuit power. In our work, the original problem which is nonconvex in nature is first transformed into a new convex problem. Then it is demonstrated the existence of a unique globally optimal power allocation using Lagrange multiplier method. Finally, numerical simulation results show that the number of transmit antennas plays an important role in increasing the energy efficiency, namely, we can obtain an optimal value of transmit antennas. It is revealed that the number of users and the QoS requirement have an effect on the energy efficiency.