Abstract:
The increased capacity of multi-level cells (MLC) in emerging nonvolatile memory (NVM) technologies comes at the cost of higher cell write energies and lower cell enduran...Show MoreMetadata
Abstract:
The increased capacity of multi-level cells (MLC) in emerging nonvolatile memory (NVM) technologies comes at the cost of higher cell write energies and lower cell endurance. In this paper, we describe MFNW, a Flip-N-Write encoding solution that effectively reduces the average write energy and improves endurance of MLC NVMs. Two MFNW modes are proposed and analyzed: cell Hamming distance (CHD) mode and energy Hamming distance (EHD) mode. For both modes, we derive a probabilistic model to approximate the statistical behavior of MFNW. For negligible error, the probabilistic model predicts the average number of cell writes per memory word, which is proportional to energy consumption. This enables word length optimization to maximize energy reduction subject to memory overhead constraints. We also estimate the hardware and delay overheads to integrate MFNW into a phase change memory prototype. MFNW is compared to two state-of-the-art techniques in the literature using traces of SPEC2006 benchmarks. Simulation results show an average energy reduction of 19% over the state-of-the-art techniques. Furthermore, we investigate the sensitivity of MFNW to the choice of word length, and our findings suggest a tradeoff between word length and energy reduction.
Published in: Proceedings of the 2015 IEEE/ACM International Symposium on Nanoscale Architectures (NANOARCH´15)
Date of Conference: 08-10 July 2015
Date Added to IEEE Xplore: 06 August 2015
Electronic ISBN:978-1-4673-7849-9