Installing an energy saving device such as a pre-swirl duct (PSD) is a major investment for a ship owner and prior to an order a reliable prediction of the energy savings is required. Currently there is no standard for how such a prediction is to be carried out, possible alternatives are both model-scale tests in towing tanks with associated scaling procedures, as well as methods based on computational fluid dynamics (CFD). This paper summarizes a CFD benchmark study comparing industrial state-of-the-art ship-scale CFD predictions of the power reduction through installation of a PSD, where the objective was to both obtain an indication on the reliability in this kind of prediction and to gain insight into how the computational procedure affects the results. It is a blind study, the KVLCC2, which the PSD is mounted on, has never been built and hence there is no ship-scale data available. The 10 participants conducted in total 22 different predictions of the power reduction with respect to a baseline case without PSD. The predicted power reductions are both positive and negative, on average 0.4%, with a standard deviation of 1.6%-units, when not considering two predictions based on model-scale CFD and two outliers associated with large uncertainties in the results. Among the variations present in computational procedure, two were found to significantly influence the predictions. First, a geometrically resolved propeller model applying sliding mesh interfaces is in average predicting a higher power reduction with the PSD compared to simplified propeller models. The second factor with notable influence on the power reduction prediction is the wake field prediction, which, besides numerical configuration, is affected by how hull roughness is considered. © 2022 The Authors