Realistic granular systems consist of particles of differing sizes, shapes, and material properties. One method commonly used when examining granular systems is the discrete element method (DEM). Here we seek to examine the effect of polydispersity, particularly wide particle size distributions (PSDs), on the computational expense of a DEM model. In particular, three factors can significantly increase the computational expense when simulating a wide PSD, either for a binary mixture or continuous distribution. First, the number of particles required increases significantly when compared to an equivalent monodisperse system. Second, the time step required to simulate the system is typically very small in order to properly resolve collisions involving the smallest particles. Finally, polydispersity results in additional computation and communication costs specific to running on a parallel architecture compared to simulating a monodisperse system. Together, these three factors make extremely wide PSDs prohibitively long and expensive to simulate with modern computing technology. Alternative methods, such as continuum models, are needed to simulate such systems. (C) 2014 Elsevier B.V. All rights reserved.