Abstract
This article presents an ultra-low-power parallel computing platform and its system-on-chip (SoC) embodiment, targeting a wide range of emerging near-sensor processing tasks for Internet of Things (IoT) applications. The proposed SoC achieves 193 million operations per second (MOPS) per mW at 162 MOPS (32 bits), improving the first-generation Parallel Ultra-Low-Power (PULP) architecture by 6.4 and 3.2 times in performance and energy efficiency, respectively.
Original language | English |
---|---|
Article number | 8065010 |
Pages (from-to) | 20-31 |
Number of pages | 12 |
Journal | IEEE Micro |
Volume | 37 |
Issue number | 5 |
DOIs | |
State | Published - 1 Sep 2017 |
Externally published | Yes |
Bibliographical note
Publisher Copyright:© 1981-2012 IEEE.
Funding
This work is supported by the European FP7 ERC Advanced project MULTITHERMAN (g.a. 291125) and by the Swiss National Science Foundation (SNF) project (no. 162524) “MicroLearn: Micropower Deep Learning.” We thank STMicroelectronics for chip fabrication.
Funders | Funder number |
---|---|
FP7 ERC | 291125 |
Micropower Deep Learning | |
Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung | 162524 |
Keywords
- UTBB FD-SOI
- body biasing
- energy efficiency
- parallel processing
- power management