Siracusa: A 16 nm Heterogenous RISC-V SoC for Extended Reality With At-MRAM Neural Engine
Authors:
- Arpan Suravi PrasadAPArpan Suravi Prasad
- Moritz SchererMSMoritz Scherer
- Francesco ContiFCFrancesco Conti
- Davide RossiDRDavide Rossi
- Alfio Di MauroAMAlfio Di Mauro
- Manuel EggimannMEManuel Eggimann
- Jorge Tomás GómezJGJorge Tomás Gómez
- Ziyun LiZLZiyun Li
- Syed Shakib SarwarSSSyed Shakib Sarwar
- Barbara De SalvoBSBarbara De Salvo
- Luca BeniniLBLuca Benini
- 23 May 2024
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 59 (7) , 2055-2069
Abstract
Extended reality (XR) applications are machine learning (ML)-intensive, featuring deep neural networks (DNNs) with millions of weights, tightly latency-bound (10–20 ms end-to-end), and power-constrained (low tens of mW average power). While ML performance and efficiency can be achieved by introducing neural engines within low-power systems-on-chip (SoCs), system-level power for nontrivial DNNs depends strongly on the energy of non-volatile memory (NVM) access for network weights. This work introduces Siracusa , a near-sensor heterogeneous SoC for next-generation XR devices manufactured in 16 nm CMOS. Siracusa couples an octa-core cluster of RISC-V digital signal processing (DSP) cores with a novel tightly coupled “At-Memory” integration between a state-of-the-art digital neural engine called and an on-chip NVM based on magnetoresistive random access memory (MRAM), achieving 1.7 $\times$ higher throughput and 3 $\times$ better energy efficiency than XR SoCs using NVM as background memory. The fabricated SoC prototype achieves an area efficiency of 65.2 GOp/s/mm ${}^\text{2}$ and a peak energy efficiency of 8.84 TOp/J for DNN inference while supporting complex, heterogeneous application workloads, which combine ML with conventional signal processing and control.Keywords
All Related Versions
Funding Information
- Meta Reality Labs
- Chips Joint Undertaking project TRISTAN (101095947)
- Chips Joint Undertaking project ISOLDE (101112274)
- Spoke 1 on Future HPC of the Italian Research Center on High-Performance Computing, Big Data and Quantum Computing (ICSC) funded by MUR Mission 4—Next Generation EU
- Convolve Project evaluated by the EU Horizon Europe (101070374)
- Swiss State Secretariat for Education Research and Innovation (22.00150)
This publication has 23 references indexed in Scilit:
- Marsellus: A Heterogeneous RISC-V AI-IoT End-Node SoC With 2–8 b DNN Acceleration and 30%-Boost Adaptive Body BiasingIEEE Journal of Solid-State Circuits, 2023
- On-Device Customization of Tiny Deep Learning Models for Keyword Spotting With Few ExamplesIEEE Micro, 2023
- Three-Dimensional Stacked Neural Network Accelerator Architectures for AR/VR ApplicationsIEEE Micro, 2022
- SamurAI: A Versatile IoT Node With Event-Driven Wake-Up and Embedded ML AccelerationIEEE Journal of Solid-State Circuits, 2022
- ILLIXR: An Open Testbed to Enable Extended Reality Systems ResearchIEEE Micro, 2022
- Vega: A Ten-Core SoC for IoT Endnodes With DNN Acceleration and Cognitive Wake-Up From MRAM-Based State-Retentive Sleep ModeIEEE Journal of Solid-State Circuits, 2021
- Fair and Comprehensive Benchmarking of Machine Learning Processing ChipsIEEE Design & Test, 2021
- SSDMNV2: A real time DNN-based face mask detection system using single shot multibox detector and MobileNetV2Sustainable Cities and Society, 2021
- Magnetoresistive Random Access Memory: Present and FutureIEEE Transactions on Electron Devices, 2020
- MRIMA: An MRAM-Based In-Memory AcceleratorIEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2019