Comprehensive on-line two-dimensional liquid chromatography is an attractive technique to enhance peak capacity and deal with complex samples. Yet, its optimization is not straightforward and requires to take into account a huge number of parameters to finally obtain a good compromise between different criteria which are often conflicting. In this study, we propose a procedure for RPLC × RPLC separations able to define, for a given analysis time, the optimized LC × LC parameters for achieving the best compromise between high peak capacity and low dilution. This procedure makes use of both a prediction tool based on theoretical relationships and a Pareto-optimality approach to optimize these two criteria. The possibility to split the flow-rate of the first dimension before the interface is taken into account as well as the focusing effect in the second dimension. This optimization procedure is applied to the prediction of optimized conditions for the separation of peptides. The need for very short and efficient columns (typically 30 mm as column length and 1.3 μm as particle diameter) in the second dimension is clearly established whatever the geometry of the first dimension column. Furthermore it is shown that several sets of quite different conditions in the first dimension can lead to similar results and hence that additional quality descriptors are required to make a decision. RPLC × RPLC separations of a protein digest were carried out and the obtained results were found in very good accordance with the predicted ones. Our results were compared to the top results reported in the literature for 1D-RPLC separations of peptides. More than 4-fold increase in peak capacity (5100 vs 1150) at 200 min has been obtained. Furthermore a 40-fold gain in analysis time (1 h vs 40 h) was obtained for achieving 1800 as effective peak capacity which is close to the highest peak capacity reported in 1D-RPLC for the separation of peptides.