An Aging-Aware Early Cache Eviction Strategy to Enhance Static Random-Access Memory Cells’ Lifetime

Document Type : Research Article

Author

Department of Computer Engineering, Faculty of Engineering, Golestan University, Gorgan, Iran

10.24200/sci.2025.65890.9721

Abstract

This paper presents a comprehensive analysis of the impact of early cache eviction on the aging of cache cells. It highlights that, in addition to the previously identified factors contributing to static noise margin (SNM) degradation in cache cells, the state of cached data plays a critical role in this process. The analysis reveals that the uneven distribution of clean and dirty data blocks across the lines of a cache set can also be a significant factor in SNM degradation. To address this issue, this study proposes an early cache eviction strategy aimed at balancing the distribution of dirty and clean data blocks over cache lines, thereby mitigating SNM degradation. To achieve this, the decision tree of the Pseudo-LRU replacement policy is redesigned to incorporate cache line states and address conflict miss types. Experimental results demonstrate that the enhanced cache improves the hold and read SNM by approximately 10% and 12%, respectively, while incurring negligible cache hit reduction and minimal area and energy overheads.

Keywords

Main Subjects



Articles in Press, Accepted Manuscript
Available Online from 23 September 2025
  • Receive Date: 11 December 2024
  • Revise Date: 28 May 2025
  • Accept Date: 17 August 2025