The increasing power density of electronic components, the demand for higher overall system efficiencies and new manufacturing methods lead to a continuous expansion of thermal management boundaries. The aim of the presented work is to gather the current boundaries for optimizing the shape of a pin fin heat sink in terms of thermal resistance, pressure drop and coefficient of performance (COP, fraction of thermal power by pumping power). These new shapes can be fabricated by additive manufacturing (AM), which opens up significantly more design freedom in comparison to common methods like milling or extrusion. The optimization was performed by a Genetic Algorithm (GA) based on either a computational fluid dynamics (CFD) simulation or a metamodel approximating these results. The goal was to investigate the Pareto frontier of the physical limits. A reduction in computational effort was achieved by subdividing a reference, state of the art cooling system, into a smaller, re presentative subsystem. Depending on the pin shape and with the reference boundary conditions, an increase of 55% of COP was verified by measurement.