Calculations of pressure loss and separation efficiency of a uniflow cyclone were performed on the basis of CFD simulations and compared with experimental data. It is shown that simplifications which are often made to reduce computational effort significantly limit the prediction quality. Turbulence modeling has the major influence on gas phase velocity and pressure distributions and thus also affects the calculation of solid particle trajectories. Errors in separation efficiency calculation are diminished by co-simulating the flow in the swirl generator in order to determine proper boundary conditions for cyclone flow. Although CFD for cyclones is momentarily not able to completely substitute experiments, it can reduce experimental costs for design and optimization.