This paper analyzes the behavior of the Incremental Conductance (IC) maximum power point tracking algorithm used in photovoltaic modules. Three variants of the algorithm are compared under both simulations and real world tests with a DSP 1104 card control. The results obtained show that the variant whose simulated behavior is the best, actually is the worst in real world experiments, indicating that it is mandatory to test this kind of algorithms in real facilities. In this case, the facilities are placed in the laboratories of the Faculty of Engineering of Vitoria-Gasteiz (UPV/EHU, Spain), and the photovoltaic modules are on its roof.