Energy-efficient Ternary Content Addressable Memory Using Ferroelectric FET

To overcome the Memory Wall problem in the traditional von Neumann machines, computing in-memory (CiM) circuits where computational tasks are performed within the memory blocks are widely investigated. Ternary content addressable memory (TCAM) is a commonly used circuit for CiM, which performs paral...

Full description

Saved in:
Bibliographic Details
Published in2021 International Conference on Intelligent Technology and Embedded Systems (ICITES) pp. 36 - 43
Main Authors Xudong, Lu, Xunzhao, Yin
Format Conference Proceeding
LanguageEnglish
Published IEEE 31.10.2021
Subjects
Associative memory
Capacitance
Computational modeling
computing-in-memory
Delays
Energy consumption
Energy efficiency
energy-efficient circuit
FeFET
TCAM
Transistors
Online AccessGet full text

Cover

More Information
Summary:To overcome the Memory Wall problem in the traditional von Neumann machines, computing in-memory (CiM) circuits where computational tasks are performed within the memory blocks are widely investigated. Ternary content addressable memory (TCAM) is a commonly used circuit for CiM, which performs parallel searches across the entire memory against a given query. TCAM has a wide range of application scenarios, such as IP routers, associative memories and advanced machine learning models, etc. Currently some non-volatile devices, for example, resistive random access memories (ReRAMs), ferroelectric transistors (FeFETs), are employed in TCAM designs for improved area efficiency. However, since TCAM often works in parallel with the computational unit, its energy consumption is one of the aspects that we are more concerned about. In this paper, we design two energy-efficient FeFET-based TCAM designs by either reducing the matching line capacitance or eliminating the pre-charging phase to reduce the search energy, and then use Spectre to validate the circuit functionality, and compare the proposed two FeFET TCAMs with several existing representative TCAMs in terms of cell size, energy consumption and delay.
DOI:10.1109/ICITES53477.2021.9637079