Energy-Efficient Reliable HZO FeFET Computation-in-Memory with Local Multiply & Global Accumulate Array for Source-Follower & Charge-Sharing Voltage Sensing
Chihiro Matsui, Kasidit Toprasertpong, Shinichi Takagi, Ken Takeuchi
Abstract
Energy efficient, high throughput, noise immune, high density HZO FeFET Computation-in-Memory (CiM) is proposed. Local Multiply & Global Accumulate Array is realized by source-follower read, which multiplies neural network inputs and weights (FeFET V <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">TH</inf> ), and charge-sharing, which accumulates multiplied values. Proposed CiM is immune to read-disturb. After 10-year data-retention, 3 bit/cell FeFET is feasible. ×64 energy efficient, 66 TOPS/W, reliable multiply-accumulate (MAC) operation is achieved.
Topics & Concepts
ComputationVoltageNoise (video)ThroughputComputer scienceEnergy (signal processing)Electrical engineeringCharge sharingOptoelectronicsElectronic engineeringPhysicsEngineeringAlgorithmTelecommunicationsArtificial intelligenceQuantum mechanicsImage (mathematics)WirelessFerroelectric and Negative Capacitance DevicesAdvanced Memory and Neural ComputingSemiconductor materials and devices