This study conducts a probabilistic tsunami hazard assessment (PTHA) and compares two approaches to representing earthquake source variability in the PTHA. The target region is the coast of Zihuatanejo in the State of Guerrero, Mexico. First, numerous synthetic fault slip distributions are generated using a stochastic random-phase process. The moment magnitude ranges from 7.8 to 8.6. A numerical tsunami simulation is implemented for each earthquake fault slip. The result of the Monte Carlo simulation indicates the tsunami heights at the nearshore of city areas tend to be higher. Then, the exceedance probabilities of tsunami height are estimated and compared using two different PTHA approaches: the random phase approach and the logic tree approach. The logic tree can generally incorporate many types of uncertainty, but this study focuses on the earthquake source uncertainty for comparison. The comparison result indicates significant differences between the two tsunami hazard models. Additionally, the logic tree approach is used to investigate the possible ranges in tsunami heights for extreme events by assuming that a sizable epistemic uncertainty exists in a given region. The tsunami heights for a 1,000-year event vary significantly when the weighting values for the paths in the logic tree are changed.