Emergent Mind

Abstract

Due to increasing cache sizes and large leakage consumption of SRAM device, conventional SRAM caches contribute significantly to the processor power consumption. Recently researchers have used non-volatile memory devices to design caches, since they provide high density, comparable read latency and low leakage power dissipation. However, their high write latency may increase the execution time and hence, leakage energy consumption. Also, since their write endurance is small, a conventional energy saving technique may further aggravate the problem of write-variations, thus reducing their lifetime. In this paper, we present a cache energy saving technique for non-volatile caches, which also attempts to improve their lifetime by making writes equally distributed to the cache. Our technique uses dynamic cache reconfiguration to adjust the cache size to meet program requirement and turns off the remaining cache to save energy. Microarchitectural simulations performed using an x86-64 simulator, SPEC2006 benchmarks and a resistive-RAM LLC (last level cache) show that over an 8MB baseline cache, our technique saves 17.55% memory subsystem (last level cache + main memory) energy and improves the lifetime by 1.33X. Over the same resistive-RAM baseline, an SRAM of similar area with no cache reconfiguration leads to an energy loss of 186.13%.

We're not able to analyze this paper right now due to high demand.

Please check back later (sorry!).

Generate a summary of this paper on our Pro plan:

We ran into a problem analyzing this paper.

Newsletter

Get summaries of trending comp sci papers delivered straight to your inbox:

Unsubscribe anytime.