Research of finding more energy efficient hydraulic systems than the conventional load sensing system for construction machines such as loader cranes has been carried out for many years. Optimizing system designs for energy efficiency is done based on some definition of how the machine is operated. Another way of reducing the energy consumption is to change how the machine is operated and optimizing the movements based on the current system design. For an automated system, the drive pattern to perform a task can be optimized with the objective of minimal energy consumption for a given productivity level. In this study the semi-automatic motion of a loader crane known as Crane Tip Control (CTC) is considered. The operator provides a velocity input in a 2D space that the crane tip should follow. The boom system of the crane is a kinematically redundant system with three actuators for performing a motion in 2D. There can thus be different strategies on how the control system should operate the three actuators, the current solution is optimized for lifting capacity rather than energy efficiency. The CTC strategy must be decided in real time and optimal solutions are unapplicable both due to the computational load and that the complete trajectory is unknown. To be able to benchmark future more energy efficient, real-time solutions, near-optimal solutions for specific trajectories can be found using offline dynamic programming. This paper presents such an approach to the problem of moving the crane actuators in the most energy efficient way while following a pre-defined tip trajectory in space and time. The algorithm uses a novel model of the energy demand of the hydraulic system including Gaussian process models of the cylinder chamber pressures trained on experimental crane data. As a proof of concept, the dynamic programming solutions for different crane tip trajectories are fed to a real crane, and the energy consumption is measured and compared to the performance of the commercially available CTC as well as to a drive cycle representing a human operator. The results show that the optimized solution reduces the energy consumption by 18.5% on average on the tested trajectories, compared to the commercially available CTC. Compared to the drive cycle, the average energy reduction is 14%. These results indicate that optimized automated motions can significantly reduce the energy consumption of crane operation.