当前位置:首页 > 报告详情

嘉实多与IEIT Systems就液冷技术建立战略合作伙伴关系——嘉实多公司发布.pdf

上传人: 明**** 编号:1011361 2025-12-21 8页 1.56MB

1、Castrol&IEIT Systems Strategic Partnershipon Liquid CoolingDavid Leung-IEIT SystemsMiguel Navarro-bp CastrolOCP EMEA Summit Day 1Challenges in Intelligent Computing Power Construction of Large ModelThe traditional engineering model has a long construction period and great delivery pressureThe comput

2、ing power and power consumption per unit space of the intelligent computing center are rising and cannot be smoothed out3-5 years3-5 years3-5 years10-15 yearsIT Life CycleDC Life CycleConstruction:3 yearsWith the popularization and development of the new AIGC business model,the business needs of lar

3、ge models,the sudden increase in IT density,and the rising energy consumption,the construction of AIDC faces many challenges.0100200300400500600700800GPU v.s CPU TDP Trend(W)NVIDIA GPUIntel CPUThe traditional method has a long construction period and cannot upgrade to the new generation of technolog

4、y in time,which leads to the dilemma of being outdated as soon as it is built.Row-level Computing Power Modules,Large-scale&Efficient DeploymentPOD level deploymentWide geographical deploymentSafety&ReliabilitySingle-row/double-row POD design,single POD with 16 cabinets for deploymentHigh reliabilit

5、y(CDU by 2N/N+1 deployment)High performance Megawatt-level CDU(the max.water temperature on the primary side is 41C)CDU extreme performance mode(solves the problem of high inlet water temperature in tropical areas)Convenient O&M(upper pipe routing,2-layer design)High reliability water collection tra

6、y,leakage monitoring system,pipeline leakage monitoringImproved safety(cutting out the cabinet liquid supply system when leakage occurs)Prefabricated Computing Power Factory Improves Deployment EfficiencyHighly integratedStandard 40-foot containers,integration of liquid cooling c

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。
根据标记内容,全文主要围绕Castrol与IEIT Systems在液体冷却方面的战略合作伙伴关系展开。以下是关键点: 1. 挑战:传统智能计算中心建设周期长,交付压力大,单位空间计算功率和能耗上升。 2. 解决方案:Row-level Computing Power Modules,支持大规模高效部署,提高可靠性和性能。 3. 预制计算工厂:提高部署效率,集成度高,交付快速,支持多箱部署。 4. 行业合作:与客户、大学和制造商合作,进行材料兼容性、火焰传播和流体稳定性测试。 5. CFD服务:提供DLC和PG25等流体动力学服务,优化散热策略。
如何缩短数据中心建设周期?" "大型模型需求下,液冷解决方案如何提升IT密度?" 液冷技术如何实现高效交付?"
客服
商务合作
小程序
服务号
折叠