Details of particle shape, including surface roughness or texture, curvature or sharpness of surface features and edges, are known to have an impact on particle packing structure. In different computer models, shape is represented using different approaches to varying degrees of precision. Thus, it is to be expected that different approaches, or rather their precision of shape representation, can affect not only how computationally demanding the simulation model is but also how accurate the model prediction can be. This paper examines two different approaches to shape representation, and their effects on the accuracy of model predictions, in the context of non-spherical particle packing. To this end, two commercially available discrete element method (DEM) based software packages, EDEM and DigiDEM, are used. The former, referred to here as sphere-composite approach, represents one extreme where a shape is typically coarsely represented by clumping together a small number of primary spheres. The latter, known as voxel-based approach, represents the other extreme where a shape is typically finely represented by a huge number of voxels (3D pixels). Both are used to simulate packing of cylinders -the most common shape of catalyst pellets in packed column reactors widely used in chemical, oil refinery and process industries. Previously reported X-ray CT scan of a packed bed provides the experimental measurements for both to compare with, in terms of bulk packing fraction, axial and radial packing fraction profiles, and pellet orientation distributions. Eight sphere-composite representations of the same cylindrical pellet were tested. Two of them gave results that quantitatively (i.e., within 5% margin of error) follow experimental measurements. A range of factors that in theory could affect accuracy of the simulation results have been examined in detail, including edge roundedness, surface roughness and restitutional behaviour as a function of sphere-composite representations. The conclusion therefore is that, for packing at least, matching the object's overall shape and dimensions is not enough, only when a high enough resolution is applied to corners and edges, could the sphere-composite approach possibly 2 match the experimental data quantitatively.