Applying DEM to prediction of tumbling mill performance is challenging because several different modes of breakage are active in the process. As our knowledge of breakage and computational capacity improve, it is worth revisiting previous simulation tasks which did not produce a satisfactorily realistic description of a measured process. At SAG 2006, we reported the simulated and measured outcomes of treating a well characterised ore in a 1.2 m diameter mill. This well instrumented, pilot scale mill at the University of KwaZulu-Natal has been combined with some new approaches to ore testing to allow different modes of breakage to be tested. The mill design allows the rate of generation of fine material to be measured in close to real time for autogenous and SAG mill charges. The updated simulation model estimates mass loss with particle evolution and embeds the cumulative damage and breakage. To more clearly delineate the effects of different types of breakage, surface wear has not been included in these simulations. The simulated results are compared with the measured results. The cumulative damage model is more realistic for SAG operation but is inadequate for AG. Hence, any realistic model must incorporate several breakage mechanisms. Issues such as relating modelling inputs to particle breakage characterisation data and the accuracy of the predictions of the models are discussed. Crown Copyright (C) 2013 Published by Elsevier Ltd. All rights reserved.