当前位置:首页 > 报告详情

DC25_SLIDES_Track6_112GSIInvestigationForImmersionCooling_Akinwale.pdf

上传人: S** 编号:1240865 2026-05-16 27页 3.50MB

1、Information Classification:GeneralWelcome to ConferenceJanuary 2830,2025Santa Clara Convention Center1ExpoJanuary 2930,2025 Information Classification:General112G Signal Integrity Investigation for Immersion-Cooling AI Server Femi Akinwale,(Intel)Pengfei Cheng(Douyin Vision),Ying Zhang(Douyin Vision

2、)Yulong Wang Hui Zou(ByteDance),Chen Shen(ByteDance)Huaguo(Andy)Yu,IEIT Kai Wang(Intel),Wenbin Tian(Intel),Zhiming Li(Intel),Xu U Zhang(Intel),Paul Xu(Intel),Xiaoguo Liang(Intel),Ken Zhang(Intel),Nishi Ahuja(Intel)2Information Classification:GeneralImageSPEAKERSFemi AkinwaleSignal Integrity Engineer

3、,IntelOluwafemi.AkinwaleIFemi Akinwale is a signal integrity engineer within Intels Data center and AI Group.He is focused on modeling and analysis of Ethernet systems across different product lines.Femi received his B.Sc.in Electrical and Computer Engineering from the University of California,San D

4、iego,and his M.Sc.degree from San Diego State University.3Information Classification:GeneralImageSPEAKERSPengfei ChengSignal Integrity Engineer,Douyin Vision Co.,LPengfei Cheng is a SI Engineer of Douyin Vision Co.,Ltd.since2021.He focuses on the high speed and power integrity in server over the las

5、t decade.Pengfei received a BS from Northeastern University(Shenyang,China)in 2009 and a MS from Tianjin University(Tianjin,China)in 2012.4Ying ZhangSignal Integrity and Power Integrity Engineer,Douyin Vision Co.,LSince 2020,Ying has been working as a SI/PI engineer in Douyin Vision Co.,Ltd.Responsi

6、ble for the self-developed server and chip SI/PI architecture evaluation,system-level simulation and cost optimization.Ying received a MS in Professional Processor Institute,Beijing Institute of Technology in 2016.Information Classification:GeneralImmersion Cooling Overview Impact on Ethernet Connec

word格式文档无特别注明外均可编辑修改,预览文件经过压缩,下载原文更清晰!
三个皮匠报告文库所有资源均是客户上传分享,仅供网友学习交流,未经上传用户书面授权,请勿作商用。
1. **研究背景**:2025年1月28-30日圣克拉拉会议中心,聚焦浸没式冷却AI服务器的112G信号完整性问题,由英特尔、字节跳动等团队联合研究。 2. **核心挑战**:浸没冷却介质(如Er=2.07的液体)导致信号劣化——眼高降低59%、眼宽减少42%、垂直眼闭合(VEC)增加78%,BER从空气中的2.2e-9升至液体中的1.1e-6。 3. **解决方案**:优化QSFP连接器设计,通过阻抗调谐(针对Er=2.07)恢复性能,仿真显示COM边际提升1.5-2dB。 4. **结论**:浸没冷却影响PAM4信号,但优化连接器设计可确保112G以太网和AI SerDes的可靠运行。
**液冷如何影响信号?** **112G信号在液冷中表现如何?** **如何优化连接器设计?**
客服
商务合作
小程序
服务号
折叠