The simulation of sediment transport by three-dimensional (3D) modelling is linked with the question of how accurate such models are. The current paper provides a test where detailed field measurements of velocities and bed elevation changes over a 3-month period from a prototype reservoir are compared with simulation results. The SSIIM software was used to compute water flow and sediment transport in the Iffezheim reservoir. The numerical model solved the Navier–Stokes equations on a 3D unstructured grid with dominantly hexahedral cells. The k-epsilon turbulence model was used, together with the SIMPLE method to find the pressure. The 3D convection–diffusion equation for suspended sediments was solved for nine sediment fractions. The computed velocity pattern showed good correspondence with the measurements. Grid sensitivity tests showed that the main flow features were computed in different grid sizes, but more accurately so with the finest grid. The sediment deposits were reasonably well computed in location and magnitude. A sensitivity test revealed that the computed bed elevation changes were most sensitive to the fall velocities of the finest cohesive sediment particles and sediment cohesion. The sediment deposition computations were also to some degree sensitive to the sediment discharge formula, bed roughness, discretization scheme and the grid resolution.