GEAR is a new fully parallel chemodynamical Tree/SPH code designed to study the formation and evolution of galaxies. GEAR treats the complex physics of baryons, such as radiative cooling and feedback, star formation, and chemical evolution. It has been fully characterized by performing hundreds of simulations, including convergence tests with the spatial resolution varying from 100 pc down to 12 pc. We demonstrate here the performance of GEAR in the context of the evolution of the dwarf spheroidal galaxies. Our models, combined with the stellar chemical abundances derived from high resolution spectroscopy in a sample of dSphs (The ESO Large Programme DART), bring crucial clues on otherwise poorly constrained physical parameters, such as the supernova feedback, star formation efficiency or galaxy mass profiles. Moreover, these new simulations allow to predict the number of stars formed in a given dark matter halo mass, providing a crucial test for cosmology.