This study first evaluates the performance of three model experiments in representing rainfall over part of Vietnam and the Lower Mekong Basin for the historical period 1986-2005. The three experiments include the Coupled Model Intercomparison Project Phase 5 (CMIP5) EC-EARTH Global Climate Model (GCM) and two downscaling runs based on a regional climate model at 25km resolution with the GCM forcing (RCM-25km) and at 5km resolution with the RCM-25km forcing (RCM-5km). Verifications against observations show that the experiments generally capture the spatial distribution of climatological rainfall. While the GCM well represents the observed average rainfall cycles, its coarse resolution limits its capability in reproducing extreme rainfall values. The downscaling experiments do not clearly show their advantage in simulating average rainfall but exhibit significant added values when representing extreme rainfall in the study region. The RCM-5km does not outperform its driving 25km experiment in representing the mean and extreme rainfall values, suggesting that having a better resolution may not compensate for having a good model configuration with appropriate physical schemes. Analysis of climate projection for the far future period 2080-2099 under two representative concentration pathways (RCP) scenarios, RCP4.5 and RCP8.5, reveals that the downscaling experiments can modify the change direction of future rainfall obtained with the GCM. While the EC-EARTH GCM generally projects wetter tendencies of up to 50%, the downscaling experiments project a general decrease of down to -50% under both scenarios over the study domain. Regarding extreme rainfall, the annual maximum 1-day rainfall amount (RX1day) is projected to increase for the three experiments. The simple daily intensity index (SDII) future changes follow those of the annual rainfall values.