Neural inference at the frontier of energy, space, and time

• 成果类型: Article
• 署名作者: Modha, Dharmendra S.; Akopyan, Filipp; Andreopoulos, Alexander; Appuswamy, Rathinakumar; Arthur, John V.; Cassidy, Andrew S.; Datta, Pallab; DeBole, Michael V.; Esser, Steven K.; Otero, Carlos Ortega; Sawada, Jun; Taba, Brian; Amir, Arnon; Bablani, Deepika; Carlson, Peter J.; Flickner, Myron D.; Gandhasri, Rajamohan; Garreau, Guillaume J.; Ito, Megumi; Klamo, Jennifer L.; Kusnitz, Jeffrey A.; McClatchey, Nathaniel J.; McKinstry, Jeffrey L.; Nakamura, Yutaka; Nayak, Tapan K.; Risk, William P.; Schleupen, Kai; Shaw, Ben; Sivagnaname, Jay; Smith, Daniel F.; Terrizzano, Ignacio; Ueda, Takanori
• 署名单位: International Business Machines (IBM); IBM USA
• 刊物名称: SCIENCE
• ISSN/ISSBN: 0036-11041
• DOI: 10.1126/science.adh1174
• 发表日期: 2023-10-20
• 页码: 329-335
• 关键词: network

摘要

Computing, since its inception, has been processor-centric, with memory separated from compute. Inspired by the organic brain and optimized for inorganic silicon, NorthPole is a neural inference architecture that blurs this boundary by eliminating off-chip memory, intertwining compute with memory on-chip, and appearing externally as an active memory chip. NorthPole is a low-precision, massively parallel, densely interconnected, energy-efficient, and spatial computing architecture with a co-optimized, high-utilization programming model. On the ResNet50 benchmark image classification network, relative to a graphics processing unit (GPU) that uses a comparable 12-nanometer technology process, NorthPole achieves a 25 times higher energy metric of frames per second (FPS) per watt, a 5 times higher space metric of FPS per transistor, and a 22 times lower time metric of latency. Similar results are reported for the Yolo-v4 detection network. NorthPole outperforms all prevalent architectures, even those that use more-advanced technology processes.