For the purpose of acquiring precise underwater 3D spatial orientation, a ‘fitness-wheel-shaped’ MEMS vector hydrophone (FWVH) is developed, which achieves monitoring of a tri-axial acoustic pressure differential via appropriate constructional design and strain readout. Compared to the previously reported T-structure-combined vector hydrophone (TSVH) which has a nonlinear relationship between output and external force, using the FWVH it is feasible in principle to achieve a tri-axial match, which make it promising in accurate acoustic orientation. A theoretic model for sensitivity is developed, where torque moment is taken into consideration to match the simulation and practical results. Superior to 3D force sensors, bandwidth is taken as the key index for the FWVH and is designed in detail with a compromise of sensitivity. In view of the bandwidth, the parameters are determined for detecting the signal in the range 20–300 Hz. A vibration test demonstrates tri-axial resonance frequencies. Finally, the frequency response curve of the FWVH shows that tri-axial sensitivities are consistent, agreeing with the analysis and simulation results, which reach −174 dB (0 dB referring to 1 V µPa−1) at 300 Hz.