QUEST: A 7.49TOPS Multi-Purpose Log-Quantized DNN Inference Engine Stacked on 96MB 3D SRAM Using Inductive-Coupling Technology in 40nm CMOS

📋 论文概要

本文提出QUEST，一种基于对数量化（log-quantized）的多用途DNN推理引擎，通过电感耦合技术堆叠在96MB 3D SRAM上，解决了传统DRAM堆叠方案中延迟过大的问题，实现了7.49TOPS的高性能推理。

💡 主要创新点

• 采用对数量化（log-quantization）方法，大幅降低内存带宽需求，同时保持推理精度。
• 利用电感耦合（inductive-coupling）技术实现DNN加速器与96MB 3D SRAM的紧密堆叠，显著降低数据访问延迟。
• 设计多用途（multi-purpose）架构，支持多种DNN模型的灵活部署。

🏷 关键词

📄 原文摘要

Shinya Takamaeda-Yamazaki1, Junichiro Kadomoto2, Tomoki Miyata2, Mototsugu Hamada2, Tadahiro Kuroda2, Masato Motomura1 Hokkaido University, Sapporo, Japan Keio University, Yokohama, Japan 1 2 A key consideration for deep neural network (DNN) inference accelerators is the need for large and high-bandwidth external memories. Although an architectural concept for stacking a DNN accelerator with DRAMs has been proposed previously, long DRAM latency remains problematic and limits the performance

👥 作者与机构

Kodai Ueyoshi1, Kota Ando1, Kazutoshi Hirose1,