An innovative control strategy, which is based on the Robust Model Predictive Control (MPC) methodology, was developed with the purpose of optimizing the engine thermal management; the proposed control strategy adjusts the coolant flow rate by means of an electric pump, in order to bring the cooling system to operate around the onset of nucleate boiling. In the present paper, the advantages of the proposed cooling approach are evaluated along the NEDC homologation cycle, which was both simulated and replicated by means of laboratory tests; the latter include coolant, lubricant and wall temperature measurements. Special attention was reserved to the warm-up period. The case considered herein is that of a Spark Ignition engine, about 1.2 dm3 displacement, and a comparison with standard crankshaft driven pump is included. The proposed strategy makes use of a dynamic model of the cooling system of an ICE that is able to predict the heat transfer both under single-phase forced convection and in the presence of nucleate or saturated boiling. The model, which was widely validated by experimental tests, also defines a metrics for establishing the heat transfer mechanism inside the engine and for estimating the distance of the engine thermal state from the onset of nucleate boiling. Results show that the developed MPC algorithm is robust in terms of disturbance rejections, respects the defined system constraints and it is effective in decreasing the warm-up time and in reducing the coolant flow rate under fully warmed conditions as compared to the standard mechanical pump.