Accelerating Innovations in the New Era of HPC, 5G and Networking with Advanced 3D Packaging Technologies

Max Min, Sylvie Kadivar
{"title":"Accelerating Innovations in the New Era of HPC, 5G and Networking with Advanced 3D Packaging Technologies","authors":"Max Min, Sylvie Kadivar","doi":"10.23919/IWLPC52010.2020.9375855","DOIUrl":null,"url":null,"abstract":"The explosion of big data along with the accelerated global socio economical transformations we are experiencing are dramatically transforming the way we live and work. These shifts are accelerating HPC, 5G, Mobile, AR/VR, IoT, Networking and AI infrastructure. Under “ More Moore ” paradigm, scaling down of new transistor and interconnect has been improved from 32nm high-K metal gate, 28nm/18nm FDSOI (Fully Depleted Silicon on Insulator), 14nm FINFET (Fin Field Effect Transistor), 7nm EUV (Extreme Ultraviolet Lithography) and down to 3nm GAA (Gate All Around). All these silicon technologies are helping to have more transistors and more functions in the system. However, developing transistors in advanced node processes is getting more challenging and costly. Consequently, further heterogeneous system integration requires solutions that go “ Beyond Moore ” paradigm. The solutions can be new system integration architecture and advanced packaging technologies [1]. For further discussion, let's discuss two memories: DRAM (Dynamic Random Access Memory) and SRAM (Static Random Access Memory). DRAM has high density per mm2 and high latency whereas SRAM has low density per mm2 and low latency. In PC and Mobile era, low-latency SRAM was integrated into a logic as cache memory whereas high-latency DRAM was separately integrated as a discrete component that was assembled on system board or package. In new AI era where hundreds or thousands of computing cores are needed, there is a strong demand for new system architecture with low-latency, high-bandwidth and/or higher density SRAM in 3D. 3D SRAM integration is helping to have dedicated low-latency SRAM memory per computing core or cluster. Along with transistor scaling for SRAM and core logic, advanced 2.5/3D packaging technologies are essential to the success of design platforms tailored to these new demands requiring more and higher bandwidth SRAM memories next to computing logic devices with lower latency and satisfying the sharp curve of technology acceleration and adoption in new era [2]–[6]. After discussing DRAM and decoupling capacitor integration, this paper introduces the first generation of 3DIC wafer-level logic packaging technology called X-Cube and demonstrates the technology through package and functional test vehicles with stacked SRAM memories on top of a logic die.","PeriodicalId":192698,"journal":{"name":"2020 International Wafer Level Packaging Conference (IWLPC)","volume":"13 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2020-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"5","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2020 International Wafer Level Packaging Conference (IWLPC)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.23919/IWLPC52010.2020.9375855","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 5

Abstract

The explosion of big data along with the accelerated global socio economical transformations we are experiencing are dramatically transforming the way we live and work. These shifts are accelerating HPC, 5G, Mobile, AR/VR, IoT, Networking and AI infrastructure. Under “ More Moore ” paradigm, scaling down of new transistor and interconnect has been improved from 32nm high-K metal gate, 28nm/18nm FDSOI (Fully Depleted Silicon on Insulator), 14nm FINFET (Fin Field Effect Transistor), 7nm EUV (Extreme Ultraviolet Lithography) and down to 3nm GAA (Gate All Around). All these silicon technologies are helping to have more transistors and more functions in the system. However, developing transistors in advanced node processes is getting more challenging and costly. Consequently, further heterogeneous system integration requires solutions that go “ Beyond Moore ” paradigm. The solutions can be new system integration architecture and advanced packaging technologies [1]. For further discussion, let's discuss two memories: DRAM (Dynamic Random Access Memory) and SRAM (Static Random Access Memory). DRAM has high density per mm2 and high latency whereas SRAM has low density per mm2 and low latency. In PC and Mobile era, low-latency SRAM was integrated into a logic as cache memory whereas high-latency DRAM was separately integrated as a discrete component that was assembled on system board or package. In new AI era where hundreds or thousands of computing cores are needed, there is a strong demand for new system architecture with low-latency, high-bandwidth and/or higher density SRAM in 3D. 3D SRAM integration is helping to have dedicated low-latency SRAM memory per computing core or cluster. Along with transistor scaling for SRAM and core logic, advanced 2.5/3D packaging technologies are essential to the success of design platforms tailored to these new demands requiring more and higher bandwidth SRAM memories next to computing logic devices with lower latency and satisfying the sharp curve of technology acceleration and adoption in new era [2]–[6]. After discussing DRAM and decoupling capacitor integration, this paper introduces the first generation of 3DIC wafer-level logic packaging technology called X-Cube and demonstrates the technology through package and functional test vehicles with stacked SRAM memories on top of a logic die.
查看原文
分享 分享
微信好友 朋友圈 QQ好友 复制链接
本刊更多论文
借助先进的3D封装技术,加速HPC、5G和网络新时代的创新
大数据的爆炸以及我们正在经历的加速的全球社会经济转型正在极大地改变我们的生活和工作方式。这些转变正在加速HPC、5G、移动、AR/VR、物联网、网络和人工智能基础设施的发展。在“More Moore”范式下,新晶体管和互连的缩小已经从32nm高k金属栅极,28nm/18nm FDSOI(绝缘体上完全耗尽硅),14nm FINFET (Fin场效应晶体管),7nm EUV(极紫外光刻)和3nm GAA (gate All Around)得到改善。所有这些硅技术都有助于在系统中拥有更多的晶体管和更多的功能。然而,在先进的节点工艺中开发晶体管变得越来越具有挑战性和昂贵。因此,进一步的异构系统集成需要“超越摩尔”范式的解决方案。解决方案可以是新的系统集成架构和先进的封装技术[1]。为了进一步讨论,让我们讨论两种存储器:动态随机存取存储器(DRAM)和静态随机存取存储器(SRAM)。DRAM具有每平方毫米高密度和高延迟,而SRAM具有每平方毫米低密度和低延迟。在PC和移动时代,低延迟的SRAM作为缓存存储器集成到逻辑中,而高延迟的DRAM则作为分立组件单独集成在系统板或封装上。在需要数百或数千个计算核心的新人工智能时代,对具有低延迟,高带宽和/或更高密度的3D SRAM的新系统架构有强烈的需求。3D SRAM集成有助于为每个计算核心或集群提供专用的低延迟SRAM内存。随着SRAM和核心逻辑的晶体管缩放,先进的2.5/3D封装技术对于满足这些新需求的设计平台的成功至关重要,这些新需求需要更多更高带宽的SRAM存储器,旁边是具有更低延迟的计算逻辑器件,并满足新时代技术加速和采用的急剧曲线[2]-[6]。在讨论了DRAM和去耦电容的集成之后,本文介绍了第一代3DIC晶圆级逻辑封装技术X-Cube,并通过在逻辑芯片上堆叠SRAM存储器的封装和功能测试车演示了该技术。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 去求助
来源期刊
自引率
0.00%
发文量
0
期刊最新文献
New X-Ray Tubes for Wafer Level Inspection Optimizing X-Ray Inspection for Advanaced Packaging Applications Design Process & Methodology for Achieving High-Volume Production Quality for FOWLP Packaging Handling of Different FOPLP Layouts on Large Area Thermal Chucks A Study About Facile Interconnect Formations Involving SB2-JET Solder Ball Stacking and Colonnade Patterning in Hybrid Package Architectures
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
现在去查看 取消
×
提示
确定
0
微信
客服QQ
Book学术公众号 扫码关注我们
反馈
×
意见反馈
请填写您的意见或建议
请填写您的手机或邮箱
已复制链接
已复制链接
快去分享给好友吧!
我知道了
×
扫码分享
扫码分享
Book学术官方微信
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术
文献互助 智能选刊 最新文献 互助须知 联系我们:info@booksci.cn
Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。
Copyright © 2023 Book学术 All rights reserved.
ghs 京公网安备 11010802042870号 京ICP备2023020795号-1