The aim of this paper is to show an inter-comparison between three main precipitation products (GPCP, 3B42, GSMaP). Their biases will be analyzed according to their estimation algorithm design and to the relative weight of primary information (infra-red or microwave satellite, rain gauges network). The comparison focuses on the accumulated precipitations and few rainy events between September and November 2010 in Maghreb, as well as on the event corresponding to the hydrological Algiers catastrophe in November 2001. It appears that, while microwave sensors are a valuable tool for precipitation measurements in most situations, they fail to properly retrieve the rainfall associated with orographic effects. These products do not allow a good estimation of rainfall for extreme events with low spatial extension.