A technique for intelligently evicting cache lines within an inclusive cache architecture. More particularly, embodiments of the invention relate to a technique to evict cache lines within an inclusive cache hierarchy based on the cache coherency traffic generated between an upper level cache and lower level caches.