This paper presents an energy-efficient optimal power control policy with finite discrete inputs in the presence of statistical quality of service (QoS) limitations. In particular, an optimization problem is formulated to maximize the energy efficiency (EE) of the system, which is defined as the ratio of the effective capacity achieved with finite discrete inputs to the total power consumption. The circuit power consumption is explicitly considered in the total power expenditure. Subsequently, the optimal power control is derived using fractional programming. The resulting optimal power control strategy is further analyzed in two limiting cases, namely, the extremely stringent QoS constraints and looser QoS constraints. Through numerical results, the EE attained with the proposed optimal power control policy and constant power scheme is compared. Also, the effects of QoS constraints and signaling distribution on the maximum achievable EE are evaluated.