Abstract: Marine heatwaves (MHWs) have become a frequent extreme climate phenomenon in recent years and exert significant impacts on both ecological environment and social economy. In this study, we use ocean heat content (OHC) to assess the short-term prediction skill of the First Institute of Oceanography-Climate Prediction System version 2.0 (FIO-CPS v2.0) on the MHWs in upper ocean, and further employ heat budget analysis to investigate the intricacies of the MHWs prediction. Our findings reveal that the prediction skill of FIO-CPS v2.0 on MHWs exhibits regional and seasonal variations. The model shows better prediction skill over the tropical eastern Pacific, tropical western Pacific, and northeastern Pacific. It is worth noting that the prediction system suffers from the spring prediction barrier in the tropical eastern Pacific, but performs well in autumn in the tropical western Pacific and northeastern Pacific. Further analysis shows that the prediction skill is closely related to the underlying driving mechanisms of the MHWs. In the tropical eastern Pacific, the prediction skill of the MHWs primarily derived from its capability to capture the El Niño-Southern Oscillation (ENSO). In northeastern Pacific, the surface and subsurface prediction skills are governed by surface heat flux and horizontal advection, respectively. By contrast, vertical advection dominates MHWs predictability in the tropical western Pacific. Additionally, FIO-CPS v2.0 demonstrates reliable ability in predicting the changes of the MHWs under global warming. This study not only promotes our understanding on the prediction skill of short-term climate prediction system FIO-CPS v2.0 but also offers a valuable reference for the short-term prediction of the MHWs in upper ocean.