SK 海力士推出 3D FeNAND 技术 瞄准 AI 计算SK Hynix launches 3D FeNAND technology to target AI computing


AI free token/AI免费积分

IT之家 10 月 15 日消息,据韩媒 The Elec 今天报道,SK 海力士现已公布 3D FeNAND 达成模拟存内计算(Analog Compute-in-Memory,A-CiM)的研究成果,相关研究已发表于 IEEE 论文。

SK 海力士推出 3D FeNAND 技术 瞄准 AI 计算

据介绍,A-CiM 可利用模拟信号直接在存储器执行乘积累加运算(IT之家注:MAC),减少数据搬运、提高处理速度和能效。

3D FeNAND 则利用铁电材料的极化状态存储数据并控制电流。比较之下,传统 3D NAND 使用电荷俘获氮化物(CTN)结构,将电子捕获在氮化物层中。铁电材料的特征是,即使外部电场消失,其内部极化状态仍能保持。

SK 海力士表示,FeNAND 可在单个存储单元内产生多种电流状态,合适 AI 计算所需的权重数据。

本文来源:https://www.ithome.com/1/013/604.htm
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。

IT House October 15 news, according to the Korean media The Elec reported today, SK Hynix has now announced the research results of 3D FeNAND to achieve analog in-memory computing (Analog Compute-in-Memory, A-CiM), related research has been published in the IEEE paper.

According to reports, A-CiM can use analog signals to perform multiply-accumulate operations directly in memory (IT Home Note: Mac), reducing data handling and improving processing speed and energy efficiency.

SK Hynix launches 3D FeNAND technology to target AI computing

3D FeNAND uses the polarized state of the ferroelectric material to store data and control the current. In contrast, conventional 3D NAND uses charge-trapping nitride (CTN) structures to trap electrons in the nitride layer. The characteristic of ferroelectric materials is that their internal polarization state can be maintained even if the external electric field disappears.

SK Hynix said that FeNAND can generate multiple current states in a single memory cell, which is suitable for the weight data required for AI calculations.

This article is from: https://www.ithome.com/1/013/604.htm For copyright issues, please write to: 13951431913#139.com or submit a review on the right.

暂无评论,快来发表第一条评论吧!No comments yet. Be the first!

📮 需求咨询📮 Inquiry