The Mediterranean region is identified as one of the two main hot-spots of climate change and also known to have the highest concentration of cyclones in the world. These atmospheric features contribute significantly to the regional climate but they are not reproduced by the Atmosphere-Ocean General Circulation Models (AOGCM), due to their coarse horizontal resolution, which have recently been run in the frame of the 5th Climate Model Intercomparison Project. This article investigates the benefit of dynamically downscaling the Institut Pierre Simon Laplace (IPSL) AOGCM (IPSL-CM5) historical simulation by the weather and research forecasting (WRF) for the representation of the Mediterranean surface winds and cyclonic activity. Indeed, when considering IPSL-CM5 atmospheric fields, the dramatic underestimation of the cyclonic activity in the most cyclogenetic region of the world jeopardizes our ability to investigate in-depth the Mediterranean regional climate and trend in the context of global change. The WRF model shows remarkable skill to reproduce regional cyclogenesis. Indeed, cyclones occurrence is quasi-absent in IPSL-CM5 data but when applying dynamical downscaling their spatial-temporal variability is very close to the re-analysis. This is a clear benefit of dynamical downscaling in regions of strong topographic forcing. This "steady" source of forcing allows the production of lee cyclogenesis and the development of strong cyclones, whatever the quality of the large-scale circulation provided at the WRF's boundaries by IPSL-CM5. However, dynamical downscaling still presents disadvantages as for instance the fact that large-scale inaccurate features of the IPSL-CM5 regional circulation are replicated by WRF due to the boundary controlled (small domain) simulation. The advantages and disadvantages of dynamical downscaling are thoroughly discussed in this paper revealing its importance for climate research, especially in the context of future scenarios and wind impacts.