DMTJ-Based Non-Volatile Ternary Content Addressable Memory for Energy-Efficient High-Performance Systems

Kevin Vicuna, Luis Miguel Procel, Lionel Trojman, Ramiro Taco

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

2 Scopus citations

Abstract

This paper explores performance of non-volatile ternary content addressable memories (NV-TCAMs), exploiting double-barrier magnetic tunnel junction (DMTJ) as comparatively evaluated with respect to the single barrier MTJ (SMTJ)-based solution. The comparison is performed at the circuit-level, considering different memory words. Overall, simulation results show that the DMTJ-based NV-TCAM is a good alternative to replace SMTJ-based NV-TCAM, mainly due to the search operation improvement. In particular, for a 144-bit NV-TCAM word operating at a nominal voltage of 1.1 V, the DMTJ-based solution offers improvements in terms of energy and search error rate of 14% and 66%, respectively, while showing similar search delay as the SMTJ-based NV-TCAM.

Original languageEnglish
Title of host publication2022 IEEE 13th Latin American Symposium on Circuits and Systems, LASCAS 2022
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781665420082
DOIs
StatePublished - 2022
Externally publishedYes
Event13th IEEE Latin American Symposium on Circuits and Systems, LASCAS 2022 - Santiago, Chile
Duration: 1 Mar 20224 Mar 2022

Publication series

Name2022 IEEE 13th Latin American Symposium on Circuits and Systems, LASCAS 2022

Conference

Conference13th IEEE Latin American Symposium on Circuits and Systems, LASCAS 2022
Country/TerritoryChile
CitySantiago
Period1/03/224/03/22

Bibliographical note

Publisher Copyright:
© 2022 IEEE.

Keywords

  • Double-barrier magnetic tunnel junction
  • Ternary content-addressable memories
  • energy-efficiency

Fingerprint

Dive into the research topics of 'DMTJ-Based Non-Volatile Ternary Content Addressable Memory for Energy-Efficient High-Performance Systems'. Together they form a unique fingerprint.

Cite this