With increasing computational power, simulations of regional climate are now becoming possible on convective-resolving grids, thus eliminating the need for a convective parameterization. In the present study, a series of seasonal calculations using the Weather Research and Forecasting (WRF) model are computed at 4-km grid spacing, which reasonably resolves most convective systems. Simulations are computed for both the DJF and MAM seasons as averaged over 2005-2008, with a model domain covering the majority of the Amazon Basin and the adjacent South American coastline. Precipitation statistics are computed and compared to satellite rainfall retrieval data from the 13-year Tropical Rainfall Measuring Mission (TRMM) record. For comparison, a set of companion simulations with 12-km grid spacing are also computed, using the Kain-Fritsch convective parameterization. As compared to the 12-km runs, the 4-km simulations show significant improvement in the overall mean rain rate, the rain rate probability distributions, and the diurnal evolution and timing of precipitation. Both the 4-km and 12-km cases capture the coastal propagating signal and the interior basin-wide diurnal oscillation; however, the 4-km case shows better timing and evolution statistics. Compared to TRMM, the 4-km case rains too infrequently, but is more likely to produce rain events at high rain rates, thus resulting in a similar overall average rain rate. Overall, the present calculations show significant promise for computing regional rainfall patterns on convective-resolving grids.