24小时热门版块排行榜    

Znn3bq.jpeg
查看: 949  |  回复: 0

yexuqing

新虫 (文学泰斗)

太阳系系主任

[交流] 融合忆阻器和数字内存计算处理助力高效边缘计算

融合忆阻器和数字内存计算处理助力高效边缘计算

▲ 作者:TAI-HAO WEN, JE-MIN HUNG, WEI-HSING HUANG, CHUAN-JIA JHANG, YUN-CHEN LO, HUNG-HSI HSU, ET AL.

▲ 链接:

https://www.science.org/doi/10.1126/science.adf5538

▲ 摘要:

人工智能(AI)边缘设备更倾向于采用高容量非易失性内存计算(CIM)来实现高能效和足够准确的快速唤醒响应。大多数先前的工作要么依据基于忆阻器的CIM,但因其耐用性有限而遭受精度损失且不支持训练;要么依据数字静态随机存取存储器(SRAM)的CIM,但受限于大面积制造需求和易失性存储。

研究组报道了一种使用忆阻器-SRAM CIM融合方案的AI边缘处理器,可同时利用数字SRAM CIM的高精度和电阻式随机存取存储器忆阻器CIM的高能效和存储密度。这也使自适应本地训练能够适应个性化特征和用户环境。

该融合处理器实现了高CIM容量、短唤醒-响应延迟(392微秒)、高峰值能效(77.64 TOPS/W)和稳健的精度(精度损失<0.5%)。这项工作表明,忆阻器技术已经超越了实验室开发阶段,现已具备用于AI边缘处理器的可制造性。

▲ Abstract:

Artificial intelligence (AI) edge devices prefer employing high-capacity nonvolatile compute-in-memory (CIM) to achieve high energy efficiency and rapid wakeup-to-response with sufficient accuracy. Most previous works are based on either memristor-based CIMs, which suffer from accuracy loss and do not support training as a result of limited endurance, or digital static random-access memory (SRAM)–based CIMs, which suffer from large area requirements and volatile storage. We report an AI edge processor that uses a memristor-SRAM CIM-fusion scheme to simultaneously exploit the high accuracy of the digital SRAM CIM and the high energy-efficiency and storage density of the resistive random-access memory memristor CIM. This also enables adaptive local training to accommodate personalized characterization and user environment. The fusion processor achieved high CIM capacity, short wakeup-to-response latency (392 microseconds), high peak energy efficiency (77.64 teraoperations per second per watt), and robust accuracy (<0.5% accuracy loss). This work demonstrates that memristor technology has moved beyond in-lab development stages and now has manufacturability for AI edge processors.
回复此楼

» 猜你喜欢

已阅   回复此楼   关注TA 给TA发消息 送TA红花 TA的回帖
相关版块跳转 我要订阅楼主 yexuqing 的主题更新
普通表情 高级回复 (可上传附件)
最具人气热帖推荐 [查看全部] 作者 回/看 最后发表
[考研] 调剂求收留 +23 果然有我 2026-04-10 24/1200 2026-04-12 10:30 by zhouwenxian
[考研] 297求调剂 +17 ORCHID1 2026-04-10 18/900 2026-04-12 09:59 by zhouwenxian
[考研] 307求调剂 +10 tzq94092 2026-04-10 10/500 2026-04-12 08:18 by wise999
[考研] 359求调剂 +5 胃痉挛累了 2026-04-11 5/250 2026-04-11 19:55 by lbsjt
[考研] 一志愿京区985,085401,与本科专业一致,电子信息工程, +4 阳光开朗的男孩 2026-04-10 4/200 2026-04-10 18:27 by shenrf
[考研] 一志愿京区985,085401电子信息,本科电子信息 +3 阳光开朗的男孩 2026-04-10 3/150 2026-04-10 16:29 by sophia_93
[考研] 282,电气工程专业,求调剂,不挑专业 +9 jggshjkkm 2026-04-10 9/450 2026-04-10 14:55 by 逆水乘风
[考研] 环境专硕调剂 +16 会说话的肘子 2026-04-06 16/800 2026-04-10 10:30 by asy1wn
[考研] 材料复试求调剂 +20 xhhdjdjsjks 2026-04-09 20/1000 2026-04-10 10:25 by 孙小小12457
[考研] 调剂 +19 不逢春 2026-04-05 20/1000 2026-04-10 10:15 by may_新宇
[考研] 一志愿华工085600 331分 +6 天下ww 2026-04-09 6/300 2026-04-09 18:59 by l_paradox
[考研] 0703化学求调剂 +21 不知名的小卅 2026-04-08 21/1050 2026-04-09 18:55 by l_paradox
[考研] 材料调剂 +10 18815505510 2026-04-09 11/550 2026-04-09 17:07 by 544594351
[论文投稿] 求助文献原文 10+3 18500821399 2026-04-08 3/150 2026-04-09 16:56 by 北京莱茵润色
[考研] 一志愿华南师范大学0702物理学305调剂 +4 念常安 2026-04-07 6/300 2026-04-08 22:53 by bljnqdcc
[考研] 材料考研求调剂总分280 +30 mkjlz1 2026-04-06 35/1750 2026-04-08 21:25 by cyh—315
[考研] 一志愿吉大化学327求调剂 +12 王王白石 2026-04-06 13/650 2026-04-08 16:05 by luoyongfeng
[考研] 材料调剂 +11 一样YWY 2026-04-07 11/550 2026-04-07 15:13 by shdgaomin
[考研] 307求调剂 +3 Youth@@ 2026-04-07 3/150 2026-04-07 09:25 by 小黑不怕难
[考研] 362求调剂一志愿中国石油大学 +4 我要考大 2026-04-06 6/300 2026-04-06 14:11 by 无际的草原
信息提示
请填处理意见