The ability of MC codes to predict the preferential sputtering of compound targets is investigated. The DYNA and TRIDYN codes are run for 3 keVAr+bombardment of aSiGebinary target. The preferential sputtering ofSiandGe, the depth dependence of the sputter cross-section and the relocation operators are calculated. Difficulties arise in trying to reproduce the experimentally reported absence of preferentiality in the sputtered flux. The models used for the surface barriers, as well as the barrier heights, influences strongly the predicted quantities. A spherical surface barrier predicts much closer to stoichiometric fluxes than a planar barrier. Different codes give different collisional diffusivities for the target species in the bulk. The need for further experiments is stressed if some guidance in the choice of input parameters in the codes is desired.