Rupture complexity, typically in the form of heterogeneous slip distribution pattern, significantly affects the local tsunami wave field. However, the effect of rupture complexity is not commonly considered in any form of tsunami hazard assessment. Taking rupture complexity into account significantly increases the computational load, particularly in regional-scaled probabilistic tsunami hazard assessments (PTHAs) that usually require a large number of simulations based on synthetic scenarios. In this study, we investigate how the heterogeneous slip distribution affects the regional-scaled PTHA by taking the South China Sea (SCS) as an example. By doing this, we update PTHA for the SCS by incorporating the best available information of seismic tsunamigenic sources along the Manila megathrust. We integrate a stochastic source model into a Monte Carlo-type simulation, in which a broad range of slip distribution patterns is generated for large numbers of synthetic earthquake events. Green's function technique is employed to efficiently calculate the nearshore tsunami wave amplitude along the SCS coastlines. Our result suggests that for a relatively small and confined region like the SCS, the commonly used approach based on the uniform slip model significantly underestimates tsunami hazard not only in the near-source region like west Luzon, as expected, but also in the relative far field, such as south China and central Vietnam. Additionally, our sensitivity test of the patch size effects suggests that large patch size is unable to adequately resolve the details of heterogeneous seafloor deformation, and such approaches considerably underestimate the potential tsunami hazard for the SCS coasts.