We present a new design in which two branch prediction mechanisms are used in conjunction. We show that the combination of these mechanisms will reduce branch penalty, while also reducing chip area. © 1992.