1S1R sub-threshold operation in Crossbar arrays for low power BNN inference computing

J. Minguet Lopez (1) , F. Rummens (2) , L. Reganaz (1) , A. Heraud (2) , T. Hirtzlin (1) , L. Grenouillet (1) , G. Navarro (1) , M. Bernard (1) , C. Carabasse (1) , N. Castellani (1) , V. Meli (1) , S. Martin (1) , T. Magis (1) , E. Vianello (1) , C. Sabbione (1) , D. Deleruyelle (3) , Marc Bocquet (4) , Jean-Michel Portal (4) , G. Molas (1) , F. Andrieu (1)

1 CEA-LETI - Commissariat à l'énergie atomique et aux énergies alternatives - Laboratoire d'Electronique et de Technologie de l'Information
2 LSTA - Laboratoire Systèmes-sur-puce et Technologies Avancées
3 INL - Institut des Nanotechnologies de Lyon
4 IM2NP - Institut des Matériaux, de Microélectronique et des Nanosciences de Provence

We experimentally validated the sub-threshold reading strategy in OxRAM+OTS crossbar arrays for low precision inference in Binarized Neural Networks. In order to optimize the 1S1R sub-threshold current margin, an experimental and theoretical statistical study on HfO$_2$-based 1S1R stacks with various OTS technologies has been performed. Impact of device features (OxRAM RHRS, OTS non-linearity and OTS threshold current) on 1S1R sub-threshold reading is elucidated. Accuracy and power consumption of a Binarized Neural Network designed in 28nm CMOS have been estimated with Monte Carlo simulations. A gain of 3 orders of magnitude in power consumption is demonstrated in comparison with conventional threshold reading strategy, while preserving the same network accuracy.

Mots clés: chalcogenide; crossbar; OTS; RRAM; BNN

Domaines: Micro et nanotechnologies/Microélectronique

J. Minguet Lopez, F. Rummens, L. Reganaz, A. Heraud, T. Hirtzlin, et al.. 1S1R sub-threshold operation in Crossbar arrays for low power BNN inference computing. IMW 2022 - IEEE International Memory Workshop, May 2022, Dresden, Germany. pp.1-4, ⟨10.1109/IMW52921.2022.9779253⟩. ⟨cea-03707392⟩