We offer an analytical solution for the optimal frame size of the non-muting version of the Basic Frame Slotted ALOHA collision resolution protocol for RFID networks. Previous investigations of RFID frame size have been empirical and have not yielded a general result. Our theoretical analysis provides a generalized result. Our solution can be used to determine the optimal frame size for any given number of RFID tags. Suboptimal selection for the frame size can result in substantially longer than minimum census delays and can unnecessarily increase energy consumption. We were able to demonstrate about 20% performance improvement in reduced census delay for a given range of values. Our results can help speed-up reader-side processing times, lower the implementation complexity of RFID readers, and increase their energy efficiency.