Global simulations of tropospheric sulfur are performed in the Goddard Institute for Space Studies (GISS) general circulation model (GCM) and used to calculate anthropogenic sulfate direct radiative forcing. Prognostic species are in-cloud oxidant H2O2, dimethylsulfide (DMS), methanesulfonic acid (MSA), SO2 and sulfate. Compared with most previous models (except others with prognostic H2O2), this model has relatively high anthropogenic SO2 and sulfate burden. We show that this is due partly to the depletion of the prognostic H2O2 and that moist convection delivers significant levels of SO2 to the free troposphere in polluted regions. Model agreement with surface observations is not remarkably different from previous studies. Following some previous studies, we propose that an additional in-cloud or heterogeneous oxidant is likely to improve the simulation near the surface. Our DMS source is lower than sources in previous studies, and sulfur values in remote regions are generally lower than those observed. Because of the high flux of SO2 to the free troposphere and the relatively low natural source, our model indicates a larger global anthropogenic contribution to the sulfate burden (77%) than was estimated by previous global models. Additional high-altitude observations of the sulfur species are needed for model validation and resolution of this issue. Direct radiative forcing calculations give an annual average anthropogenic sulfate forcing of -0.67 W/m2. We compare the radiative forcings due to online (hourly varying) versus offline (monthly average) sulfate and find little difference on a global average, but we do find differences as great as 10% in some regions. Thus, for example, over some polluted continental regions the forcing due to offline sulfate exceeds that of online sulfate, while over some oceanic regions the online sulfate forcing is larger. We show that these patterns are probably related to the correlation between clouds and sulfate, with positive correlations occuring over some polluted continental regions and negative correlations over high-latitude oceanic regions. ¿ 1999 American Geophysical Union