The production of cellulosic feedstocks for renewable fuels will increase over the coming decades. However, it is uncertain which feedstocks will be best suited for bioenergy production. A key factor dictating feedstock selection for a given region is water use efficiency (WUE), the trade-off between evapotranspiration (ET) and carbon uptake or productivity. Using an ecosystem model, two of the top candidate cellulosic feedstocks, Miscanthus × giganteus (miscanthus) and Panicum virgatum (switchgrass) were compared to Zea mays L. (maize), the existing dominant bioenergy feedstock, with 0 and 25% residue removal for the Midwest US. We determined productivity in three ways: harvested yield (HY), net ecosystem productivity (NEP) and net biome productivity (NBP). Evapotranspiration was compared against each of the three productivity metrics, respectively, to yield Harvest Water Use Efficiency (HWUE), Ecosystem Water Use Efficiency (EWUE) and Biome Water Use Efficiency (BWUE). Simulations indicated that, over the study domain, miscanthus had a significantly higher HWUE compared to switchgrass and maize, while maize and switchgrass were similar. When EWUE was compared miscanthus was higher than both maize and switchgrass, which were similar for most of the region. Biome WUE was similar for both of the perennials and higher compared to maize for most of the study domain with the exception of the driest regions where maize showed the highest BWUE. Removing 25% of maize residue slightly increased HWUE and greatly decreased BWUE throughout the domain, however only HWUE changes were statistically significant. These results indicate that the feedstock with the highest WUE varied based on the productivity metric, but BWUE for maize was consistently lower than the perennials.