Agricultural eco-efficiency which combines both economic and environmental performance is an effective tool for agricultural sustainability analysis. Taking China's 30 provinces as case regions, We applied a SBM model of DEA to measure the agricultural eco-efficiency from 1998 to 2009. The SBM model of DEA is non-radical, non-oriented and considers the agricultural non-point source pollution as an undesirable output in the calculating process which is different from the other existing models. Fertilizer, animal, labor, machine, sown area and water consumption were considered as input indicators, provincial agricultural output value and agricultural non-point source pollution were considered as desirable output indicaor and undesirable output indicaor respectively. The main results are summarized as follows:(1)Agricultural eco-efficiency in China during 1998-2009 is 0.499 which is very low, this implies China's agricultural economic rapid growth is at the expense of natural resources destruction and agriculture environmental pollution, the coordination task of agricultural economic growth and environmental protection and resource saving is very difficult in China; (2)Agricultural eco-efficiency of Beijing, Shanghai, Chongqing, Hainan is equal to one, this demonstrates that these four provinces have efficiency of resources utilization and environment protection for their agricultural economy, but the agricultural eco-efficiency of the other 26 provinces are below one showing that most provinces in China have inefficiency of resources utilization and environment protection for their agricultural economy; (3)Moreover, in order to promote agricultural eco-efficiency in these 26 provinces, the inputs optimization and outputs optimization were designed based on optimization principles. The optimization results show that agricultural output value(agricultural desirable output) has no congestion in all provinces, so the agricultural eco-inefficiency is mostly attributed to inputs and undesirable output. China has enormous potential on the reduction of agricultural non-point source pollution and on the decrease of resource inputs to improve the agricultural eco-efficiency. The eco-efficiency optimization provides deeper insights into the causes of eco-inefficiency and gives further implications on resource and environment protection strategies in China. Our results also showed that the SBM model could be used as a useful tool to reflect the eco-efficiency in a defined region and help governments at various levels to find the most optimized solutions in improving their eco-efficiencies.