With customer demand for more comfort and safety features, the amount of software and electronics in a car is steadily growing. In order to cope with the corresponding increased power consumption of the E/E (electric/electronic) architecture, various power saving techniques were developed to turn off functions, electronic control units (ECUs), or even whole bus systems when they are temporarily not needed. However, the potential of these techniques is highly dependent on the mapping of the vehicle functions to the distributed network of ECUs. This paper presents a mixed-integer linear programming (MILP) approach that supports the designer in an early phase of the automotive E/E development process to find an optimal function partitioning with respect to energy-efficiency. While various MILP-based task to processor mapping concepts have been investigated before, the contribution of this paper brings forward an automotive domain specific MILP model with a cost function that is based on the power consumption for computation and communication of the overall system. Furthermore, the schedulability of the partitions is here taken into account. The approach was integrated into an existing E/E design space exploration framework and evaluated for feasibility and scalability. The results reflect the potential gain of a power-aware function partitioning.