Due to their large variety of applications in the PSE area, complex optimisation problems are of high interest for the scientific community. As a consequence, a great effort is made for developing efficient solution techniques. The choice of the relevant technique for the treatment of a given problem has already been studied for batch plant design issues. However, most works reported in the dedicated literature classically considered item sizes as continuous variables. In a view of realism, a similar approach is proposed in this paper, with discrete variables representing equipment capacities. The numerical results enable to evaluate the performances of two mathematical programming (MP) solvers embedded within the GAMS package and a genetic algorithm (GA), on a set of seven increasing complexity examples. The necessarily huge number of runs for the GA could be performed within a computational framework based on a grid infrastructure; however, since the MP methods were tackled through single-computer computations, the CPU time comparison are reported for this one-PC working mode. On the one hand, the high combinatorial effect induced by the new discrete variables heavily penalizes the GAMS modules, DICOPT + + and SBB. On the other hand, the Genetic Algorithm proves its superiority, providing quality solutions within acceptable computational times, whatever the considered example.