The concepts of chloride penetration front and critical chloride content in reinforced concrete are not easily quantifiable and the disparity of results in the literature attests a great variability of the causes. The aim of this study is to propose part of the disparity interpretation on a diffusion test by examining the geometrical influence of gravel distribution on the variability of chloride concentration at a given depth and of depth at a given concentration, in an uncracked cover concrete. For this, concrete is considered as a two-phase material with mortar (assumed homogeneous) as one phase and gravels as another. 3D composite specimens are modeled by using a mix of mortar and randomly distributed spheres of different sizes (gravels) to reproduce particle size distribution. An isothermal diffusion model based on Fick’s second law along with Langmuir and Freundlich types interaction is considered. Freundlich parameters are calculated from C3A content of the ciment. Langmuir parameter values are taken from the literature. The finite element method is used for the composition. Simulations are compared with experimental results obtained with an OPC concrete from the BHP2000 French National project.The study shows good agreement between the proposed pure diffusion finite element model and experimental results. It also shows that depth and concentration dispersions are not negligible, especially at 3month of exposure. A part of concentration isovalue location can be 3mm deeper than the mean value location, on the studied concrete.