An accurate prediction of the non-linear roll damping is required in order to calculate the resonant roll motion of moored FPSO’s. Traditionally, the roll damping is obtained with model tests using decays or forced roll oscillation tests. Calculation methods based on potential flow are not capable of predicting this hydrodynamic damping accurately as it originates from the viscous nature of the fluid and the complex vortical flow structures around a rolling vessel. In recent years Computational Fluid Dynamics (CFD) has advanced such that accurate predictions for the roll damping can be obtained. In this paper CFD is employed to predict the roll damping for a barge-type FPSO. The objectives of the paper are to investigate the capability and accuracy of CFD to determine roll damping of an FPSO and to investigate whether two-dimensional calculations can be used to estimate the roll damping of a three-dimensional FPSO geometry. To meet these objectives, extensive numerical sensitivity studies are carried out for a 2D hull section mimicking the midsection of the FPSO. The numerical uncertainty for the added mass and damping coefficients were found to be 0.5% and 2%, respectively. The influence of the turbulence model was found to be significant for the damping coefficient with differences up to 14%. The 2D CFD results are compared to results from two-dimensional model tests. The calculated roll damping using the k-ω SST 2003 turbulence model matches the value from the experiments within 2%. The influence of various physical parameters on the damping was investigated through additional 2D calculations by changing the scale ratio, the roll amplitude, the roll period, the water depth, the origin of rotation and the bilge keel height. Lastly, three-dimensional calculations are carried out with the complete FPSO geometry. The 3D results agree with the 2D results except for the largest roll amplitude calculated, i.e. for 15 degrees, where the damping coefficient was found to be 7% smaller. For this amplitude end-effects from the ends of the bilge keels seem to have a small influence on the flow field around the bilge keels. This indicates that the 2D approach is a cost-effective method to determine the roll damping of a barge-type FPSO, but for large roll amplitudes or for different vessel geometries the 2D approach may not be valid due to 3D effects.