A new CFAR detector based on Grubbs statistic called G-CFAR detector is proposed and its detection architecture is given.in this paper. Also the values of the threshold in different confidence level are given. The G-CFAR is evaluated through Monte-Carlo simulation in various environments. In homogeneous background, the performance of G-CFAR shows a little worse than that of CA-CFAR. In multiple targets background, the G-CFAR performs robustly, especially when the interfering targets are not confined to a single half of the reference window, the G-CFAR keeps good performance, while the VI-CFAR can not detect the targets effectively .