Abstract
Summary
Resistive ternary content addressable memories (Re‐TCAMs) are considered a potential platform for in‐memory associative processing. Re‐TCAM permits the store and process of the data in the same physical area, allowing in‐memory computing. Given the unique properties of the non‐volatile memristive device, Re‐TCAM is capable of low search energy, large‐scale storage, and massively parallel processing. However, like any other semiconductor, memristor as a nanodevice suffers from fabrication process imperfections that provoke a significant variance in their high (RH) and low (RL) resistance states. These variations might lead to the malfunction of the Re‐TCAM array and affect its performance and reliability. This paper proposes a simplified mathematical formulation of the memristor state variability for 2T2M bit‐cell‐based Re‐TCAM array. We present a comprehensive statistical design, considering different circuit parameters and variance effects, such as tolerance of memristance variation, the memristor ratio (
α=RH/RL), transistor technology, and memory width. The impact of these parameters on the performance is formulated as the probability of error is extensively studied, and a closed‐form expression for the optimal threshold voltage is derived. The utility of the presented investigation is illustrated using a design example with real device parameters.
Resistive ternary content addressable memories (ReTCAMs) are considered a potential platform for in‐memory associative processing. In this paper, we propose a closed form model to study the impact of the variability of RRAMs (i.e., memristors) on the performance of ReCAMs. The probability of error (failure) is evaluated under different scenarios and conditions.