There are two possible approaches to leverage 3D stacked memory for future microprocessor designs. One is to stack memory on top of logic [15], [19], which requires a close collaboration between microprocessor vendors and memory vendors, and thermal management is a big challenge. The other is to place 3D stacked memory (such as Micron’s Hybrid Memory Cube) with CPUs on a silicon inter poser [20] so that the CPU design and 3D stacked memory design can be decoupling, and the thermal management is much easier than the first approach.