Gregory K. Chen, Phil C. Knag, Carlos Tokunaga, R. Krishnamurthy
{"title":"An 8-core RISC-V Processor with Compute near Last Level Cache in Intel 4 CMOS","authors":"Gregory K. Chen, Phil C. Knag, Carlos Tokunaga, R. Krishnamurthy","doi":"10.1109/vlsitechnologyandcir46769.2022.9830518","DOIUrl":null,"url":null,"abstract":"An 8-core 64b processor extends RISC-V to perform multiply accumulate within shared last level cache. Compute Near Last Level Cache (CNC) enables high-bandwidth access and local compute with the highest-capacity on-chip SRAM. The 1.15GHz chip expands virtual addressing, coherency, and consistency to CNC, enabling Linux-capable multi-core operation. CNC reduces energy by 52× for fully connected and 29× for convolutional DNN layers. MLPerf™ Anomaly Detection latency is reduced by 4.25× to 40μs.","PeriodicalId":332454,"journal":{"name":"2022 IEEE Symposium on VLSI Technology and Circuits (VLSI Technology and Circuits)","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2022-06-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2022 IEEE Symposium on VLSI Technology and Circuits (VLSI Technology and Circuits)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/vlsitechnologyandcir46769.2022.9830518","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 2
Abstract
An 8-core 64b processor extends RISC-V to perform multiply accumulate within shared last level cache. Compute Near Last Level Cache (CNC) enables high-bandwidth access and local compute with the highest-capacity on-chip SRAM. The 1.15GHz chip expands virtual addressing, coherency, and consistency to CNC, enabling Linux-capable multi-core operation. CNC reduces energy by 52× for fully connected and 29× for convolutional DNN layers. MLPerf™ Anomaly Detection latency is reduced by 4.25× to 40μs.