加法器
通流晶体管逻辑
晶体管
逻辑门
计算机科学
晶体管计数
现场可编程门阵列
计算
GSM演进的增强数据速率
逻辑族
和大门
算术逻辑单元
电子工程
可编程逻辑器件
或门
门阵列
材料科学
逻辑综合
与非门
电气工程
可编程逻辑阵列
电路复杂度
NOR门
调制(音乐)
计算机硬件
中央处理器
电子线路
单位(环理论)
能源消耗
二极管-晶体管逻辑
阈下传导
算术
作者
T. Wang,L.C. Ji,Hao Wang,Yuan Pan,Zibin Huang,Hongyu Chen
标识
DOI:10.1002/adfm.202526772
摘要
ABSTRACT The ever‐increasing computational demands of artificial intelligence matrix calculations are driving breakthrough hardware designs for edge computing. However, this progress is significantly constrained by the high transistor consumption. Here, a novel logic architecture with an ultrahigh computation density based on a multi‐region gate transistor is proposed. Benefitting from the unique ability of multi‐region barrier modulation, the consumption of transistors in our full‐logic gate unit is less than 10 % of traditional circuits. In addition, owing to the transistor possessing exceptional electrical characteristics, e.g., a switching ratio of 2.62 × 10 7 , an ultra‐low off‐state current of 32 fA, and a minimized subthreshold swing of 92.25 mV/dec, a high‐performance arithmetic logic unit (ALU) multi‐bit edge computing system based on the 4 × 8 Field‐Programmable Gate Array (FPGA) is constructed using these multi‐region gate transistors. The FPGA not only executes 8‐bit cryptographic operations directly at the network edge near end‐users, but also can be dynamically reconfigured into two independent 2‐bit full adders with carry propagation for parallel arithmetic processing. It can effectively redistribute the computational burden away from cloud resources. This work will pave the way for designing the ultra‐compact circuit for further highly efficient Internet of Things (IoT) ecosystems with energy efficiency.
科研通智能强力驱动
Strongly Powered by AbleSci AI