Water saving technologies without reduction on grain yield, like alternate wetting and drying (AWD), have been implemented successfully in irrigated rice fields. It is now needed to identify key crop traits under more pronounced water constraint for genetic improvement and further increase in water productivity. Five contrasted genotypes were grown under three water treatments in 2006 (AWD30 with irrigation set up whenever soil water potential reached -30 kPa at 15 cm deep, AWD60, and CF as continuous flooding) while nine genotypes were grown under two water treatments in 2008 (AWD30 and CF). While water use decreased significantly for all genotypes by 29% to 37% under AWD30 and by 22% to 34% under AWD60 in 2006DS, and by 17% to 25% in 2008DS, grain yield was maintained only in rare cases which expressed high contrast among genotypes' responses: grain yield of PSBRc80 was maintained in all situations while that of IR64 was significantly and systematically affected. The response of yield components to AWD was, however, not consistent with regard to the performance: cases with stable grain yield included some with unaffected yield components (three genotypes) and others with compensation between panicle number and filled grain number per panicle (PSBRc80). Cases with reduced grain yield included some with reduced grain size (two genotypes including IR64) and others with reduced filled grain number per panicle (two genotypes). Tiller emergence rate and maximum tillering increased under AWD for all genotypes, however, effect on LAI and biomass accumulation was inconsistent. Partitioning to culm was favored under AWD to the detriment of blade during the reproductive phase of the most adapted genotypes in 2008 although this was not reported in 2006. Under CF, the root/shoot ratio, total root dry matter and root dry matter below 25 cm deep were consistently higher, but for the most three adapted genotypes only. As a response to AWD, the root/shoot ratio and total root dry