A cantilever with an end mass is one of the most popular designs for piezoelectric MEMS vibration energy harvesting. The inclusion of a proof mass near the free end of a micro-cantilever can significantly enhances the power responsiveness of a vibration energy harvester per unit acceleration. However, the accommodation of the proof mass comes at the expense of the active piezoelectric area. This paper numerically and experimentally investigates this compromise, and explores the optimal proof-mass-to-cantilever-length ratio for power maximization. It was found that an end mass occupying about 60%-70% of the total cantilever length is optimal within linear response, and they notably outperform comparable cantilevers with 40% and 50% of end mass. In addition, nonlinear squeeze film air damping within the chip package was found to adversely affect the cantilevers with larger mass more significantly. A harvester prototype with 70% of the length covered by end mass (5 mm 3 ) was able to generate 1.78 μW at 0.6 ms -2 and up to 20.5 μW at 2.7 ms -2 and 210 Hz when not limited by nonlinear damping. This result outperforms the previously reported counterparts in the literature by nearly an order of a magnitude in terms of power density normalized against acceleration squared.