A new algorithm for optimizing  3G–WiFioffload and load balancing between nodes, called as New – Genetic Zone Routing Protocol (N – GZRP), is proposed in this paper. N – GZRP is a combination of outputs of Roulette Wheel selections method (RWs) and Rank Selections Method (Rs). Output from RWs is optimized by Rs where Rs is used as a threshold value of the uncertainty that occurs due to the use of RWs.  The algorithm is implemented for detecting the performance of mobile devices on several different types of networks.  The simulation results are compared with the (RWs)–GZRPand ZRP algorithm. The simulation results show that 35% of power consumption can be saved by using the proposed algorithm compared with GZRPalgorithm. Moreover,when compared with ZRP algorithm the proposed algorithm can save 60% power consumption. In addition, N – GZRP has a more stable throughput which is 3 Kbps higher than ZRP. The simulation results also show that N–GZRPhas smaller average delay than GZRP and ZRPwhich are about 20 msand 1400 ms, respectively.N – GZRP also reduces the synchronization delay about 25 ms and 45 ms compared withGZRPalgorithmand ZRP algorithm, respectively.