The energetics, structures and segregation of Pd-Au nanoalloys (all compositions for 34- and 38-atoms) have been studied using a genetic algorithm global optimization technique with the Gupta empirical potential. Three modifications of the Pd-Au parameters have been studied: parameter set I in which all parameters (A, xi, p, q and r(0)) in the Gupta potential are weighted in a symmetrical fashion; parameter set II (symmetric weighting of only the pair and many-body energy scaling parameters A and xi); and parameter set III (antisymmetric weighting of A and xi). Structural analysis reveals competition between a range of structural families; decahedra, polyicosahedra and truncated octahedra (for 34 atoms) and incomplete-icosahedra-Mackay, decahedra, polyicosahedra (low-symmetry), six-fold-polyicosahedra and a mixed octahedron-icosahedron (Oh-Ih) structure (for 38 atoms). It is shown that, by finely tuning the Gupta potential, it is possible to qualitatively reproduce the results observed at higher levels of theory (e.g. Density Functional Theory). There are four main types of chemical ordering which are observed: core-shell; spherical cap; ball-and-cup; and mixed. It is shown that the chemical ordering and the proportion of Pd-Au heteronuclear bonds in these clusters are strongly dependent on the potential parameters. Comparison of the results from parameter set III and two previously fitted potentials shows that the DFT-fit potential gives rise to similar results for energies, and lowest energy structures and homotops to those for parameter set III with w(a) = 0.8, but the exp-fit potential gives rise to qualitatively different results.