We compare forecasts of the volatility of the Australian Dollar / US Dollar exchange rate to alternative measures of ex-post volatility. We develop and apply a simple test for the improvement in the ability of loss functions to distinguish between forecasts when the quality of a volatility estimator is increased. We find that both realized variance and the daily high-low range provide a significant improvement in loss function convergence relative to squared returns. We find that a model of stochastic volatility provides the best forecasts, relative to a set of GARCH models, which includes a GARCH(1,1) that is second best.