Analog In-Memory-Compute with Multi-bit Silicon Ferro FinFET Array for Improved Energy and Area Efficiency
In this work, we demonstrate an analog in-memory-compute (AIMC) using single-transistor (1T) Ferro (FE) FinFET-based multiply-compute element (MCE). The fabricated MCE has a small cell size of 8F<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> with 2bit per cell weight storage, a large memory window of 1.8V and a linear operation range of 1.2V. We experimentally demonstrate multiply-accumulation (MAC) operation at the array level with a program/erase inhibition biasing scheme. The 1T FE-FinFET based AIMC is achieves an area efficiency of 1377 TOPs/mm<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> due to small cell size and multi-level-cell operation, and an energy efficiency of 697 TOPs/W enabled by low bit-line read voltage and excellent device linearity, with close to baseline computational accuracy.
Authors
- Omkar Phadke (ORCID: https://orcid.org/0000-0001-5606-6242)
- Wei‐Chun Wang (ORCID: https://orcid.org/0000-0003-1788-2727)
- Sharadindu Gopal Kirtania (ORCID: https://orcid.org/0000-0002-8162-9155)
- Eknath Sarkar (ORCID: https://orcid.org/0000-0001-7541-2729)
- Suman Datta (ORCID: https://orcid.org/0000-0001-6044-5173)
- Dyutimoy Chakraborty (ORCID: https://orcid.org/0000-0003-1581-9564)
- Asif Islam Khan (ORCID: https://orcid.org/0000-0003-4369-106X)
- Shimeng Yu (ORCID: https://orcid.org/0000-0002-0068-3652)
- Khandker Akif Aabrar (ORCID: https://orcid.org/0000-0002-3417-2025)
- Jaewon Shin
- Shida Zhang (ORCID: https://orcid.org/0000-0003-4187-4428)
- Chengyang Zhang (ORCID: https://orcid.org/0000-0001-6361-717X)
- Emmanuel Quezada
Institutions
- Georgia Institute of Technology (US)
Publication Details
- Published
- 2025-05-18
- DOI
- https://doi.org/10.1109/imw61990.2025.11026976
- Citations
- 1
- Primary Topic
- Ferroelectric and Negative Capacitance Devices
- Type
- article
- Field-Weighted Citation Impact
- 0.73