Biodiversity significantly affects ecosystem multifunctionality, particularly in the context of changing precipitation patterns. However, the relative importance of multiple dimensions of diversity (e.g., taxonomic, functional, and phylogenetic diversity) on ecosystem multifunctionality under different precipitation patterns remains unclear. In this study, we conducted a seven-year field manipulation precipitation experiment in the semi-arid steppe to examine the effects of taxonomic, functional, and phylogenetic diversity on total ecosystem multifunctionality (EMF), aboveground ecosystem multifunctionality (AGEMF), and belowground ecosystem multifunctionality (BGEMF) under changed precipitation in different periods. The results showed that decreased precipitation in the early (DEP), late (DLP), and entire growing season (DP) reduced EMF by 6.5%, 8.4%, and 11.6%, respectively, whereas increased precipitation in different periods had no effect on EMF. DEP, DLP, and DP decreased AGEMF by 10.0%, 9.7%, and 19.0%, respectively. In contrast, increased precipitation in the early (IEP) and entire (IP) growing season enhanced AGEMF by 5.1% and 6.9%, respectively. However, increased precipitation during the late growing season (ILP) did not affect AGEMF. DLP and DP decreased BGEMF by 13.3% and 11.2%, respectively, whereas other precipitation treatments had no effect on BGEMF. Species dominance was the major driver of AGEMF and BGEMF under precipitation decreases while soil moisture was the best predictor of AGEMF under precipitation increases. In contrast, functional and phylogenetic diversity had no effect on ecosystem multifunctionality. The greater sensitivity of EMF, AGEMF, and BGEMF to precipitation decreases than to precipitation increases indicates an asymmetric response of ecosystem multifunctionality to precipitation changes, highlighting that decreased precipitation during the late growing season weakens ecosystem multifunctionality by reducing the roles of dominant species in the semi-arid steppe. The findings are crucial for informing effective grassland management policies and maintaining stable ecosystem provisions.