In the last fifty years many semi-empirical models to predict surface pressure fluctuations beneath turbulent boundary layers (TBL) have been developed for a large variety of test conditions. Nowadays, the relevance of the TBL as a source of cabin interior noise is steadily increasing, due to quieter aircraft engines. The possibility of predicting surface pressure auto-spectra with the various publicly available semi-empirical models at several positions on the fuselage of DLR's Advanced Technology Research Aircraft (ATRA) is investigated. A large validation database was used, involving in-flight measurements at different flight levels (FL) and Mach numbers, applying two different sensor types. Predictions were performed based on semi-empirically estimated TBL parameters (partly included in the different models) and, additionally, based on TBL properties that were extracted from CFD simulations. This procedure served to identify different sources of error in the prediction. Overall, it is shown that today's models provide a large (> 10 dB) scatter among the predicted spectra. Even the most suitable approaches are not generally applicable to all relevant positions at the fuselage. Particularly in regions with strong pressure gradients and high turbulence kinetic energy measured auto-spectra cannot be reproduced with sufficient accuracy. This indicates the need for more universally applicable CFD-and CAA-based surface pressure prediction methods.