Motion tracking systems have become increasingly popular in industrial automation, providing natural human-machine interfaces that allow human-robot collaboration, of which the safety of the human operator is of utmost importance. These systems demand robustness, high precision, and low latency of motion tracking systems. Wearable motion tracking systems that deploy IMU sensors represent a suitable alternative to occlusion-prone camera-based systems and bring the advantage of their portability, low cost, and low energy footprint. State-of-the-art techniques rely on advanced signal processing and optimization algorithms to obtain the orientation estimate from the IMUs. However, the performance of IMU-based motion-tracking solutions has yet to be studied extensively. We identify that the root cause is the lack of ground-truth measurements of the sensor orientation. In this study, we synthesize a dataset providing ground truth sensor orientation and the corresponding IMU measurements. Subsequently, we compare the performance of the three state-of-the-art real-time orientation estimation algorithms regarding accuracy, convergence speed, and stability. Evaluation results demonstrate that the available gradient descent algorithms have trade-offs between stability, accuracy, and convergence speed depending on the adjustable filter gain.