Realistic representation of detailed rainfall characteristics on local scale by current state-of-the-art climate models remains a key challenge, especially on sub-daily timescales. In this research, the convection-permitting Weather Research and Forecasting (WRF) model configured with 1.5 km grid spacing is used to simulate precipitation on sub-daily timescales over the Yangtze River Delta Region of China for continuous 10 years (2005–2014). The simulations are compared to rain gauge observations, reanalysis data, and the simulations of a lower resolution WRF with 9 km grid spacing that has a parameterization of convection. The results show that precipitation over the region can be well captured by using the convection-permitting model (CPM). Furthermore, the intensity, duration and coverage of these precipitation events can be more accurately described by the CPM. On the convection timescales of 1–4 hours, especially for heavy rainfall events, the CPM is more accurate than the convection-parameterized model in capturing the short-duration events, which may be related to its better account of physical processes related with the convection on the convection-permitting scale. In addition, the extreme events which are more localized and with short-duration can be represented better by the CPM while the convection-parameterized model tends to produce widespread precipitation events covering more grid cells than observed. Biases of the simulation by the 9-km mesh convection-parameterized mode appear to be related to the deficiencies in the representation of convections.