During 2002 and 2003 a series of experimental studies in model-scale were performed to investigate compartment flame spread and fire growth in compartments and fire spread between compartments. The extensive library of measurement data was used in evaluating the flame spread modelling capabilities of the popular CFD code FDS. Input data to the material model was derived using an iterative approach, comparing stand alone predictions with experimental data on surface temperature and mass loss rate obtained from different external heat fluxes. The initial side length of the computational grid was chosen related to the dimensionless initial fire size. This choice proved to be optimal in terms of numerical agreement with the experimental data on parameters such as gas temperature, velocity and fire growth rate. Furthermore, this was shown to hold in all the tested configurations. Strictly speaking, a numerical simulation must show identical results regardless of timestep and meshing but no grid independence could be obtained, making any arbitrary simulation difficult to interpret unless the results are known beforehand. The mechanism of thermal degradation of a combustible material is rather well investigated and documented. From a flame-spread modelling point of view the modelling of the convective and the radiative heat fluxes need further research and development.