logo
ResearchBunny Logo
Demonstration of 4-quadrant analog in-memory matrix multiplication in a single modulation

Engineering and Technology

Demonstration of 4-quadrant analog in-memory matrix multiplication in a single modulation

M. L. Gallo, O. Hrynkevych, et al.

Discover groundbreaking research by Manuel Le Gallo and colleagues from IBM Research Europe, demonstrating a revolutionary 4-quadrant matrix-vector multiplication in a single modulation cycle using an analog in-memory computing chip. This innovation tackles the challenges of phase-change memory accuracy, significantly enhancing throughput and energy efficiency.

00:00
00:00
Playback language: English
Abstract
This paper demonstrates a 4-quadrant matrix-vector multiplication (MVM) in a single modulation cycle using an analog in-memory computing (AIMC) chip based on phase-change memory (PCM). The approach addresses the challenge of conductance polarity dependence in PCM devices, which degrades accuracy in single-phase reading schemes. Through analog and digital calibration procedures, the researchers achieve accurate neural network inference and similarity search tasks with significantly improved throughput and energy efficiency compared to conventional four-phase methods.
Publisher
npj Unconventional Computing
Published On
Oct 03, 2024
Authors
Manuel Le Gallo, Oscar Hrynkevych, Benedikt Kersting, Geethan Karunaratne, Athanasios Vasilopoulos, Riduan Khaddam-Aljameh, Ghazi Sarwat Syed, Abu Sebastian
Tags
analog in-memory computing
phase-change memory
matrix-vector multiplication
neural network inference
energy efficiency
throughput improvement
calibration procedures
Listen, Learn & Level Up
Over 10,000 hours of research content in 25+ fields, available in 12+ languages.
No more digging through PDFs, just hit play and absorb the world's latest research in your language, on your time.
listen to research audio papers with researchbunny