SIMBA: A Skyrmionic In-Memory Binary Neural Network Accelerator

Abstract

Magnetic skyrmions are emerging as potential candidates for next generation non-volatile memories. In this paper, we propose an in-memory binary neural network (BNN) accelerator based on the non-volatile skyrmionic memory, which we call as SIMBA. SIMBA consumes 26.7 mJ of energy and 2.7 ms of latency when running an inference on a VGG-like BNN. Furthermore, we demonstrate improvements in the performance of SIMBA by optimizing material parameters such as saturation magnetization, anisotropic energy and damping ratio. Finally, we show that the inference accuracy of BNNs is robust against the possible stochastic behavior of SIMBA (88.5% +/- 1%).

References in corpus (11)

• Spin torque switching with the giant spin Hall effect of tantalum
• The design and verification of Mumax3
• Advances in the Physics of Magnetic Skyrmions and Perspective for Technology
• Magnetic skyrmion logic gates: conversion, duplication and merging of skyrmions
• Skyrmions in Magnetic Multilayers
• Field-Dependent Size and Shape of Single Magnetic Skyrmions
• A theory on skyrmion size and profile
• Boltzmann Test of Slonczewski's Theory of Spin Transfer Torque
• Isolated zero field sub-10 nm skyrmions in ultrathin Co films
• SkyLogic - A proposal for a skyrmion logic device
• Spin-wave mediated interactions for Majority Computation using Skyrmions and Spin-torque Nano-oscillators