There are major sources of uncertainty in simulating the impact of climate change on crop yield including general circulation models (GCMs), Representative Concentration Pathways (RCPs), and future time periods. This study focused on quantifying the different sources of uncertainty associated with climate change effects on rainfed wheat grain yield in five provinces of Iran categorized into arid cold, semi-arid cold, and semi-arid cool climate classes. To do this, future climate of the provinces were projected based on 29 GCMs, two RCPs (RCP4.5 and RCP8.5), and three future time periods (2030s, 2050s, and 2090s). The Agricultural Production Systems sIMulator (APSIM) model was used to simulate wheat growth, development, and grain yield. The analysis of variance (ANOVA) was applied to quantify the major sources of uncertainty in predicting wheat grain yield for the future climate change. Based on simulation results, rainfed wheat grain yield is projected to increase by 14% in 2030s largely due to rising CO2 concentrations. The increase ranged from 11.7% under RCP4.5 in the 2030s to 20.2% under RCP8.5 in the 2090s. Around 97% of variations in rainfed grain yields were justified by climate class according to the ANOVA results. Accordingly, the share of the uncertainty sources were analyzed for each climate class separately. Most of the variations arose from the scenarios (RCP × future time period; 77%) for semi-arid climate classes, while most of the variations in arid climate class made by GCMs (54%). Overall, it was suggested that the ensemble use of a wide range of GCMs should be considered to reduce the uncertainty when simulating rainfed wheat grain yield under climate change situations, especially for arid and semiarid regions with large fluctuations in rainfall and temperature.