The paper analyses the differences of technical, allocative, cost and scale efficiencies of irrigated and rain-fed rice farmers in Sri Lanka in two different perspectives; first, relative to a common metafrontier, defined as the boundary of an unrestricted technology set and second relative to group frontiers defined to be the boundaries of restricted technology sets in each group. Data envelopment analysis (DEA) metafrontier and group frontier approaches are used for cross section survey data of 90 farms. Rain-fed farms perform comparably with the irrigated farms based on the group frontier results. Rain-fed farmers may be operating as technically efficient as they could, given the existing production technology. However rain-fed farms move significantly towards inefficiency compared to the irrigated farms under the metafrontier technology. Results indicate that the irrigation shifts the rice sector production frontier to a higher level. In addition, a second stage bootstrapped truncated regression shows that efficiency differences between two regions are explained by the timely availability of the water to a significant extent. We suggest that future sectoral policies should be designed to address the efficiency enhancing factors such as irrigation, quality seed, land ownership and scale and female labour participation.