The energy efficiency of cellular networks is explored for different network deployment strategies and traffic conditions. The total network power consumption and the ratios Watts/kbps, Watts/user are used as metrics to quantify the performance and it is shown that efficiency depends on the deployment strategy and the state of operation of the network (underutilized-overutilized). As a general conclusion it is shown that a microcell based network, comprising a large number of low power stations, is the most efficient strategy but it does not present traffic proportional characteristics. For the purpose of the investigation, the paper presents a pre-processing of the database 3D ray tracing algorithm that is enhanced with an image test procedure and multiple slope diffraction mechanisms to achieve fast and accurate predictions. The algorithm is used for channel estimations over a real urban environment described in a vector format. In addition, a power control algorithm is developed that explicitly considers power levels of neighbor base stations and is used for the network power consumption estimation. Finally, network planning is achieved through an evolutionary optimization technique that combines the above mentioned algorithms.