Abstract:The smart edge nodes require efficient matrix-vector multiplications for local deep neural network (DNN) inference. Benefiting from its high density and CMOS compatibility, the eDRAM-based computing-in-memory (CIM) [1] –[4], especially with multi-level cells (MLCs) [4], attracts rising attention. However, the performance of prior MLCeDRAM CIM is severely limited by the inconsistency of weight representations during the programming and computing: weights are programmed as fixed voltages while transistor currents are used for computation. Thus, the programming of MLCs requires calibration due to the nonlinear transistor I-V, which can be extremely complicated in the presence of $V_{T H}$ variations. Furthermore, the computing precision is severely degraded by $\mathrm{V}_{\mathrm{TH}}$ variations when small computing currents are used for high parallelism. To fundamentally surmolunt this dilemma, we propose the first currentprogramming eDRAM CIM that unifies the weight programming and computing in the current domain. The enabling technique is a novel 3T1C eDRAM cell (Fig. 1, top right). It confers several key merits: 1) the cell is programmed by the weight current directly with the selfcalibrated voltage generated on the storage capacitor; it essentially stores the weight current instead of a fixed voltage, thus mitigating $V_{\text {TH}}$ variation and nonlinear transistor I-V impacts; 2) a dynamiccascoded read structure is proposed to significantly reduce the V <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">B</inf> sensitivity while not requiring any bias voltage; 3) thanks to the accurately programmed cell, it supports MLC operation (8 current levels) without any calibration, largely increasing density; 4) a voltage-current two-step programming scheme significantly boosts the sub- $\mu \mathrm{A}$ current-weight writing speed. Combining these merits, the proposed eDRAM cell is naturally immune to transistor-level nonidealities, thus allowing a small LSB weight current of only 100nA. A $4 \mathrm{~b}$ CIM cell composed of 2MLCs is developed to support 4bsigned weights. It contains 15 current levels ranging from -700nA to 700nA. Fabricated in a $65 \mathrm{~nm}$ CMOS, the prototype achieves the highest macro-level 4b-MAC energy efficiency of 233-305TOPS/W among eDRAM CIMs.

A 28-Nm 10.4-Fj/b Cryogenic Embedded DRAM with 3T1C Gain Cell and MBIST at 4-Kelvin

Cryogenic quasi-static embedded DRAM for energy-efficient compute-in-memory applications

Cryogenic Body Bias Effect in DRAM Peripheral and Buried-Channel-Array Transistor for Quantum Computing Applications

Design Of Almost-Nonvolatile Embedded Dram Using Nanoelectromechanical Relay Devices

30 Mb/mm<sup>2</sup>/layer 3D eDRAM Computing-in-Memory with Embedded BEOL Peripherals and Local Layer-wise Calibration based on First-Demonstrated Vertically-stacked CAA-IGZO 4F<sup>2</sup> 2T0C Cell

High Performance Si-MoS2 Heterogeneous Embedded DRAM

Low-Power and Scalable BEOL-Compatible IGZO TFT eDRAM-Based Charge-Domain Computing

A 1mb 28nm Stt-Mram With 2.8ns Read Access Time At 1.2v Vdd Using Single-Cap Offset-Cancelled Sense Amplifier And In-Situ Self-Write-Termination

A Logic-Based Embedded Dram With Novel Cell Structure And Dynamically Adaptive Refresh For Long Data Retention, Zero Data Availability Penalty And High Yield

A 14nm-Finfet 1mb Embedded 1t1r Rram With A 0.022 Mu M(2) Cell Size Using Self-Adaptive Delayed Termination And Multi-Cell Reference

Self-Terminating Write of Multi-Level Cell ReRAM for Efficient Neuromorphic Computing

A 1-Mb 28-Nm 1T1MTJ STT-MRAM with Single-Cap Offset-Cancelled Sense Amplifier and in Situ Self-Write-Termination

A 200-Ms/s 12-B Cryo-CMOS CS DAC for Quantum Computing

Gain Cell eDRAM Design with Stagger Hidden Refresh Scheme

A Calibration-Free 15-level/Cell eDRAM Computing-in-Memory Macro with 3T1C Current-Programmed Dynamic-Cascoded MLC achieving 233-to-304-TOPS/W 4b MAC

A 22-nm 1-Mb 1024-b Read Data-Protected STT-MRAM Macro With Near-Memory Shift-and-Rotate Functionality and 42.6-GB/s Read Bandwidth for Security-Aware Mobile Device

Transition metal dichalcogenide FET‐based dynamic random‐access memory

First demonstration of 2T0C-FeDRAM: a-ITZO FET and double gate a-ITZO/a-IGZO FeFET with a record-long multibit retention time of >4-bit and >2000 s

24.2 A 14nm-FinFET 1Mb Embedded 1T1R RRAM with a 0.022µm<sup>2</sup> Cell Size Using Self-Adaptive Delayed Termination and Multi-Cell Reference