⚡ 本页包含 AI 生成的分析内容,仅供参考
该论文提出了一款面向增强现实应用的3D集成原型系统芯片,采用面对面晶圆键合7nm逻辑工艺,解决了AR产品对高性能、低功耗和小面积的需求。通过集成包含1K处理单元的机器学习加速器,实现了高效处理ML、NN和ISP任务。
Tony F. Wu, Huichu Liu, H. Ekin Sumbul, Lita Yang, Dipti Baheti, Jeremy Coriell, William Koven, Anu Krishnan, Mohit Mittal, Matheus Trevisan Moreira, Max Waugaman, Laurent Ye, Edith Beigné As a case study, we augmented a standalone ML accelerator consisting of 1K processing Meta, Sunnyvale, CA Augmented reality (AR) products require energy-efficient systems-on-chip (SoCs) for machine learning (ML), neural networks (NNs), and image signal processing (ISP) applications [1]. These SoCs must be high-performance, yet low power with compact form factors. They are heavily constrained by the area footprint, while the third dimension is usually left with ample space. Moreover, frequent access to off-chip memories can be prohibitively expensive in terms of latency and energy for AR devices. Fortunately, recent advances in 3D integration allow integration of additional logic and memory into the SoC without area footprint cost. In particular, face-to-face (F2F) stacking with hybrid bonding
3D stacked SoCs designed for high performance computing, they are not significant for, those designed for AR applications (e.g., max on-die temperature of 29°C for our chip).