Storage device and threshold voltage adjusting method and storage control method thereof

Abstract

The invention discloses a storage device and a threshold voltage adjusting method and a storage control method thereof. The memory device includes: a plurality of memory cells arranged in an array and each including a memory transistor; multiple word lines, wherein the grid electrodes of the storage transistors of the same row of storage units are connected to the same word line of the multiple word lines; a plurality of bit lines, wherein the drain electrodes of the storage transistors of the same column of storage units are connected to the same bit line of the plurality of bit lines; a plurality of first connecting lines, wherein the shallow P well regions of the storage transistors of the same column of storage units are connected to the same first connecting line; and a plurality of second connecting lines, wherein the deep N well regions of the storage transistors of the same column of storage units are connected to the same second connecting line. According to the storage device, the device structure is optimized for in-memory calculation, the number of transistors of the storage unit can be reduced so as to reduce the chip size, and the threshold voltage of the storage transistors can be accurately adjusted so as to improve the calculation accuracy and reliability.

Description

technical field [0001] The present invention relates to the technical field of semiconductors, in particular to a storage device and a threshold voltage adjustment method and a storage control method thereof. Background technique [0002] The development of deep learning technology has greatly promoted the breakthrough development of artificial intelligence, bringing about an unprecedented wave of commercialization and globalization of artificial intelligence. Data, computing power, and algorithms are the three elements of AI-related industries. Machines rely on efficient model algorithms for massive data training, which requires high-performance computing power as support. Among them, computing power is usually based on chips, and mature computing power carriers such as CPU (Central Processing Unit, central processing unit), DSP (Digital Signal Processing, digital signal processing), GPU (Graphics Processing Unit, graphics processing unit). [0003] In the field of artifi...

AI Technical Summary

Problems solved by technology

However, during the process of performing MAC calculations, the GPU frequently accesses data from the memory (DRAM, SRAM), which causes most of the energy to be consumed in memory access, and its energy consumption is dozens or even hundreds of times higher than the implementation Energy consumed by logic circuits for MAC calculations

Images