We consider the problem of broadcasting messages in wire- less sensor networks (WSN) in an energy-efficient manner. The problem is central for many application, as WSNs often consist in autonomous battery powered devices that use broadcast for many purposes (e.g. synchronization, data collection, etc.). A number of algorithms have been proposed to solve this problem, focusing in particular on node that are able to reduce their communication range, enabling to lower energy consumption. One of the best known such centralized algorithm is the Broadcast Incremental Power (BIP). Then, several distributed algorithms have been proposed, such as Localized BIP, Dynamic Localized BIP, and Broadcast Oriented Protocols (RBOP and LBOP). Those distributed approaches aim to reach the performance of BIP without assuming that the nodes have the knowledge of the whole graph. In this paper we answer the open question left by those previous work: how do they perform (energy-wise) with realistic devices placed in a realistic environment? Unlike previous works that consider an ideal MAC layer (with no collisions) and a simple energy consumption model (that assumes that only transmitting messages consumes energy), we use simulated MAC layers (ContikiMac and 802.15.4 MAC layers) that take into account signal propagation and the possibility of collisions, and realistic battery and energy consumption models, that consider all relevant energy aspects of sensor node hardware. It turns out that our findings are significantly different from the aforementioned theoretical studies. Among our findings, we show that the hierarchy of the routing protocols (based on their performance) is not preserved (compared with the theoretical studies), which means that wireless interference impact them in different ways. Also, we found that the MAC layer plays an important role on the performance of the upper layer protocols, and does not impact all routing protocols in the same way.