Appropriate benchmarking is essential for evaluating the efficiency with which crops use water and for identifying constraints, other than water, to crop yield. No benchmark exists for sunflower. Boundary and simulation analyses were used to quantify the water productivity of sunflower crops grown in the Western Pampas (semiarid central Argentina). The approach involved the use of a large database ( n = 169) collected in farmers’ fields over a period of 4 years, and the application of a crop simulation model in combination with actual weather and soil data. Using field data, an upper bound of 8 kg grain ha −1 mm −1 for water productivity, with an apparent seasonal soil evaporation of 75 mm, was defined. Seasonal water supply exceeded maximum expected seasonal crop requirements ( ca. 630 mm) for many crops, and a majority of crops with <630 mm of available water during the season had water productivities considerably lower than the upper bound. The field data-based upper bound was indistinguishable from that obtained using yields for a set of 47 simulations using observed initial values for soil water and nitrogen profiles. Simulation confirmed the main features of the boundary-analysis applied to field data, and many simulated crops had yields that fell below the boundary function, even when simulated yield was plotted against simulated seasonal evapotranspiration or transpiration. Long-term (33-year) simulation analyses for two sites showed that most sunflower crops in the area are subjected to episodes of transient and unavoidable water stress after floral initiation. High levels of available soil water at sowing moderate, but in most years do not eliminate, these exposures to water stress. Yield gaps with respect to the boundary function were associated with deficient or excessive rainfall during grain filling, and other, non-water related, factors such as inadequate crop nutrition, biotic stresses, low photothermal quotients during the interval close to anthesis, and lodging. A grain yield/seasonal evapotranspiration plot for a large ( n = 154) data set from experiments conducted by others in five separate environments suggests that the boundary function found for the Western Pampas is broadly applicable. Sunflower water productivity, corrected for oil-synthesis costs and seasonal vapour pressure deficit differences, approximates that of winter cereals grown in Mediterranean environments.