Abstract:In-memory computing (IMC) architecture emerges as a promising paradigm, improving the energy efficiency of multiply-and-accumulate (MAC) operations within DNNs by integrating the parallel computations within the memory arrays. Various high-precision analog IMC array designs have been developed based on both SRAM and emerging non-volatile memories. These designs perform MAC operations of partial input and weight, with the corresponding partial products then fed into shift-add circuitry to produce the final MAC results. However, existing works often present intricate shift-add process for weight. The traditional digital shift-add process is limited in throughput due to time-multiplexing of ADCs, and advancing the shift-add process to the analog domain necessitates customized circuit implementations, resulting in compromises in energy and area efficiency. Furthermore, the joint optimization of the partial MAC operations and the weight shift-add process is rarely explored. In this paper, we propose novel, energy efficient dual designs of FeFET based high precision analog IMC featuring inherent shift-add capability. We introduce a FeFET based IMC paradigm that performs partial MAC in each column, and inherently integrates the shift-add process for 4-bit weights by leveraging FeFET's analog storage characteristics. This paradigm supports both 2's complement mode and non-2's complement mode MAC, thereby offering flexible support for 4-/8-bit weight data in 2's complement format. Building upon this paradigm, we propose novel FeFET based dual designs, CurFe for the current mode and ChgFe for the charge mode, to accommodate the high precision analog domain IMC <a class="link-external link-http" href="http://architecture.Evaluation" rel="external noopener nofollow">this http URL</a> results at circuit and system levels indicate that the circuit/system-level energy efficiency of the proposed FeFET-based analog IMC is 1.56$\times$/1.37$\times$ higher when compared to SOTA analog IMC designs.

A Novel Hybrid-FE-layer FeFET with Enhanced Linearity for On-chip Training of CIM Accelerator

Hybrid-FE-Layer FeFET with High Linearity and Endurance Toward On-Chip CIM by Array Demonstration

Back-end and Flexible Substrate Compatible Analog Ferroelectric Field Effect Transistors for Accurate Online Training in Deep Neural Network Accelerators

Accelerating On-Chip Training with Ferroelectric-Based Hybrid Precision Synapse

Freely Switching Between Ferroelectric and Resistive Switching in Hf0.5Zr0.5O2 Films and Its Application on High Accuracy On-Chip Deep Neural Networks

Design and Optimization of FeFET Based CiM for Neural Network Acceleration

Low Power and Temperature-Resilient Compute-In-Memory Based on Subthreshold-FeFET

Incremental Drain-Voltage-Ramping Training Method for Ferroelectric Field-Effect Transistor Synaptic Devices

Ferroelectric and Interlayer Co-optimization with In-depth Analysis for High Endurance FeFET

Efficient deep neural network accelerator using controlled ferroelectric domain dynamics

A Reliable 2 Bit MLC FeFET with High Uniformity and 109 Endurance by Gate Stack and Write Pulse Co-optimization

A Novel Non-Volatile Inverter-based CiM: Continuous Sign Weight Transition and Low Power on-Chip Training

A Novel Ambipolar Ferroelectric Tunnel FinFET Based Content Addressable Memory with Ultra-low Hardware Cost and High Energy Efficiency for Machine Learning

HAO $+$ Al $_{\text{2}}$ O $_{\text{3}}$ FeFET Gate-Stack for Overall Improvement in Operating Voltage, Endurance, and Retention

Van Der Waals Ferroelectric Transistors: the All-Round Artificial Synapses for High-Precision Neuromorphic Computing

A Hybrid FeMFET-CMOS Analog Synapse Circuit for Neural Network Training and Inference

Energy Efficient Dual Designs of FeFET-Based Analog In-Memory Computing with Inherent Shift-Add Capability

Design Guidelines of Hafnia Ferroelectrics and Gate-Stack for Multilevel-Cell FeFET

Neuromorphic Computing with Ferroelectric FinFETs in the Presence of Temperature, Process Variation, Device Aging and Flicker Noise

Laminated Ferroelectric FET With Large Memory Window and High Reliability

Optimal Weight Models for Ferroelectric Synapses Toward Neuromorphic Computing