Convection-permitting models (CPMs) have been proven successful in simulating extreme precipitation statistics. However, when such models are used to study climate change, contrasting sensitivities with respect to resolution (CPM vs. models with parameterized convection) are found for different parts of the world. In this study, we explore to which extent this contrasting sensitivity is due to the specific characteristics of the model or due to the characteristics of the region. Therefore, we examine the results of 360 years of climate model data from two different climate models (COSMO-CLM driven by EC-EARTH and ALARO-0 driven by CNRM ARPEGE) both at convection-permitting scale (CPS, ~ 3 km resolution) and non-convection-permitting scale (non-CPS, 12.5 km resolution) over two distinct regions (flatland vs. hilly region) in Belgium. We found that both models show an overall consistent scale-dependency of the future increase in hourly extreme precipitation for day-time. More specifically, both models yield a larger discrepancy in the day-time climate change signal between CPS and non-CPS for extreme precipitation over flatland (Flanders) than for orographically induced extreme precipitation (Ardennes). This result is interesting, since both RCMs are very different (e.g., in terms of model physics and driving GCM) and use very different ways to represent deep convection processes. Despite those model differences, the scale-dependency of projected precipitation extremes is surprisingly similar in both models, suggesting that the this scale-dependency is more dependent on the characteristics of the region, than on the model used.