Mineral dust is a key component in aerosol–climate modeling due to its strong effects on radiation, clouds, and biogeochemical cycles. Dust emissions are strongly driven by convective systems, requiring convection-resolving simulations. Recent studies have identified significant uncertainties in simulated dust emissions associated with the representation of moist convection, particularly when contrasting convection-resolving and parameterized approaches. This study extends such analyzes from Northern Africa to the major global dust source regions, including the dust belt from North Africa to Central Asia, as well as East Asia and Australia. We compare dust uplift potential (DUP) derived from convection-resolving ICON simulations (DYAMOND, ∼2.5 km) with coarse-resolution CMIP6 climate models (∼200 km). DUP, based solely on threshold-exceeding near-surface winds, isolates the role of wind representation by excluding surface effects. Systematic discrepancies between the model classes emerge consistently across regions and seasons. Four key deficiencies in coarse-resolution results are identified: (1) underestimation of afternoon and nighttime convective dust events; (2) overestimation of daytime winds linked to low-level10 jet breakdown; (3) overlapping drivers causing compensating biases; and (4) under-representation of strong, persistent local wind systems. These biases arise primarily from the treatment of moist convection rather than grid spacing alone. Overall, the results demonstrate that explicitly resolving convection is essential for realistic dust emission estimates, with implications for aerosol–radiation interactions, satellite interpretation, and future high-resolution aerosol–climate modeling.