As shown in FIG. 1A, the active region 106 may be divided into three rectangular portions arranged in the X direction: a middle portion 106A, a left portion 106B, and a right portion 106C, where the middle portion 106A is sandwiched by the left portion 106B and the right portion 106C. The middle portion 106A is wider in the Y direction than the left portion 106B and the right portion 106C. Therefore, from the perspective of a top view such as FIG. 1A, the middle portion 106A laterally extends beyond the gate stack 114 in the Y direction. But the four corner portions of the gate stack 114—namely, corner portions 114A, 114B, 114C, and 114D—still laterally extend beyond the active region 106 into the isolation region 104. The reason that the corner portions 114A-114D reach into the isolation region 104 is to prevent any leakage current from circumventing the gate stack 114 between the source 126 and the drain 128. In other words, if the entire gate stack 114 stays laterally within the active region 106 with extra active space, the gate stack 114 may no longer effectively control the channel region 124 since a leakage current may flow between the source and drain outside the gate stack 114. Two edge portions of the gate stack 114—namely, edge portion 114E sandwiched between corner portions 114A and 114B, and edge portion 114F sandwiched between corner portions 114C and 114D—do not laterally extend into the isolation region 104. From the top view perspective, edges of the active region 106 laterally “intersect” edges of the gate stack 114 (although they have different vertical positions as shown in FIGS. 1B and 1C).