Temperature sensitivity of analog in-memory computing using phase-change memory
Irem Boybat, Benedikt Kersting, Syed Ghazi Sarwat, X. Timoneda, Robert L. Bruce, M. BrightSky, Manuel Le Gallo, Abu Sebastian
Abstract
Can analog in-memory accelerators provide sufficient accuracy for AI applications under ambient temperature variations? Here, we answer this question by focusing on phase-change memory (PCM)-based deep learning acceleration. We investigate for the first time the impact of temperature on multi-level PCM conductance states used to store the synaptic weights. First, we characterize the temperature and drift behavior of 10,000 doped Ge <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> Sb <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> Te <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">5</inf> (GST)-based mushroom PCM. Next, we present a model which can capture this behavior and faithfully reproduce the complete time-temperature dependence of the conductance states. Finally, we experimentally study the sensitivity of various network architectures to ambient temperature variations. For this, we employ a multi-layer perceptron, a convolutional neural network and a recurrent neural network, with more than 1.1M PCM weights. We demonstrate that a simple array-level scaling could correct for the conductance shift due to temperature and drift and prevent any significant accuracy drop for all the studied networks during inference.