New 2.5D neuromorphic discovery platform will enable AI-enhanced co-design 

A segmented graphic design showing six elements of Sandia’s 2.5D neuromorphic discovery platform.
AI-enhanced co-design will be enabled through Sandia’s 2.5D neuromorphic discovery platform.

Novel material and device concepts previously took years for iteration. Discoveries in this LDRD project will now allow them to be iterated on in weeks thanks to a new easy fabrication substrate platform for novel devices. The team designed and taped out a microelectronics discovery platform or a “lab-on-a-chip” that has high-speed pulsing and high-resolution sensing that permits researchers to gauge fundamental limits of atomic switching in novel microelectronic devices including ECRAM and ReRAM. Additionally, the team released an updated microelectronics co-design software, CrossSim 2.0, a GPU-accelerated, Python-based crossbar simulator that allows researchers to take experimental data from the discovery platform and immediately model system-level performance for sensor processing mission applications. A strong relationship with National/Regional partner Arizona State University and a partnership with former Sandian, Matt Marinella, is providing a talent pipeline to multiple U.S. citizen students working on microelectronics co-design. 


Sandia researchers linked to work

  • Sapan Agarwal 
  • William Wahby 
  • Mieko Hirabayashi 
  • Patrick Finnegan 
  • Nad Gilbert 
  • Matthew Marinella 

Sponsored by

Image of LDRD_Icon-01

Associated Publications

Xiao, T. P., Feinberg, B., Bennett, C. H., Agrawal, V., Saxena, P., Prabhakar, V., … Marinella, M. J. (2022). An accurate, error-tolerant, and energy-efficient neural network inference engine based on SONOS analog memory. IEEE Transactions on Circuits and Systems. I, Regular Papers: A Publication of the IEEE Circuits and Systems Society, 69(4), 1480–1493. doi:10.1109/tcsi.2021.3134313

Agrawal, V., Xiao, T. P., Bennett, C. H., Feinberg, B., Shetty, S., Ramkumar, K., … Agarwal, S. (2022, December 3). Subthreshold operation of SONOS analog memory to enable accurate low-power neural network inference. 2022 International Electron Devices Meeting (IEDM). Presented at the 2022 IEEE International Electron Devices Meeting (IEDM), San Francisco, CA, USA. doi:10.1109/iedm45625.2022.10019564 

Xiao, T. P., Feinberg, B., Bennett, C. H., Prabhakar, V., Saxena, P., Agrawal, V., … Marinella, M. J. (2022). On the accuracy of analog neural network inference accelerators [feature]. IEEE Circuits and Systems Magazine, 22(4), 26–48. doi:10.1109/mcas.2022.3214409