The ACCESS system has recently replaced GASP and LAPS as the operational numerical weather prediction system at the Bureau of Meteorology. This work presents an assessment of the marine surface winds from the global model, ACCESSG, using QuikSCAT scatterometer data. For the four month period between July and October 2008, it is shown that ACCESS-G provides a significant improvement to the GASP model. Wind speed scatter index is improved by 20 per cent at short lead times, with an overall gain in forecast skill of approximately 12-hour lead time. Gains persist throughout the forecast period. A negative bias has been identified in both models, with wind speed underestimated by approximately eight per cent for ACCESS-G, and three per cent for GASP. The bias is greater in the meridional direction than the zonal. Further insight into the regional variation of error for each model is gained by examining their spatial distribution. Though the overall bias for GASP is small, significant regional biases exist, with generally negative biases in the mid-latitudes being offset by positive biases in the tropics. ACCESS-G shows more consistency over the globe. This spatial structure can be expected to manifest in downstream models.