Primary radiation damage in displacement cascades in metals has been studied extensively by atomistic simulation during the last decade. The variety of defect types observed in cascade simulation is not entirely consistent with experimental data. For example, experiments on copper show a very effective production of stacking fault tetrahedra (SFTs) but this was not observed systematically in cascade simulation. To clarify this and related issues, extensive simulation of displacement cascades in copper have been performed using two different interatomic potentials, a short-range many-body potential and a long-range pair potential. We have studied the damage created by primary knock-on-atoms of energy up to 20keV, i.e. below the energy range for formation of subcascades, at temperatures 100 and 600K. Special attention was paid to cascade statistics and the accuracy of simulation in the collision stage. The former required many simulations for each temperature whereas the latter involved a modification of the simulation method. The results on variety of clusters observed, e.g. SFTs, glissile and sessile interstitial clusters, and faulted and perfect interstitial dislocation loops, lead to conclusions on the effect of the potentials and the significant variation of the number of Frenkel pairs and clustering effects produced in different cascades under the same conditions.