当前位置:首页 > 报告详情

拥抱人工智能:基础设施和先进热解决方案.pdf

上传人: 明**** 编号:1011516 2025-12-21 16页 1.80MB

1、Embracing AI:Infrastructure and Advanced Thermal SolutionsAuras TechnologyEmbracing AI:Infrastructure and Advanced Thermal SolutionsSean LimManagerAuras TechnologyCOOLING ENVIRONMENTSOutline(Optional)21AI InfrastructureAdvanced Cooling SolutionComputational power is national strengthAI ModelsPerform

2、ance(%)#AI has become an arms race.AI Infrastructure Grow up to 2028 US$90B US$42BInvesting in GPU accelerates AI infrastructure deployments growth.The spending is expected to reach near US$90 billion by 2028.Source:IDC,2024,Worldwide Artificial Intelligence Infrastructure Forecast,20242028Server TD

3、Ps rapidly increaseA100MI-100FeynmanRubinB300B200H100H200MI-450MI-375MI-325MI-300/300XMI-200/250Gaudi 1Gaudi 3Gaudi 2Power Consumption drastically increaseMcKinsey predicts that data center power consumption will represent about 12%of total US demand in 2030.12%600kWNVL 288300kWAir CoolingDirect Liq

4、uid CoolingAdvanced Cooling SolutionAdvanced Cooling SolutionData CenterBuild coolingsolutionComputingcoolingsolution3DVC Heat sinkFanCold PlateRack ManifoldBuild ManifoldIn Row CDUQuick DisconnectRackAir CoolingDirect Liquid CoolingAdvanced Cooling SolutionThickness:0.2mmPitch:0.35mm0.4mmThickness:

5、0.15mmPitch:0.3mmThickness:0.1mmPitch:0.2mm0.25mm3N2NA14A10A7A5A32022202420262028203020322034700W945W1275W1720W2320W3130W4225WSingle PhaseTwo PhaseMaterialColdPlatePowerconsumptionPhotolithographyMIXCopper and Vapor ChamberPureCopperMIXCopper(Graphene&diamond)CoolingSolutionFlow Rate1.0LPM2.0LPM0.75

6、PM1.5LPM2.0PM3.0LPM1.0LPM 1.5LPM1LPM1.5LPM200kPA350kPA100kPA100kPA2Phase VC Cold Plate1.GrapheneCopper2.DiamondCopper500kPASiliconSingle Phase VC Cold PlateYearTransistor Density 1.8XPower Consumption Reduction 25%Computing speed promote 15%CDU output Pressure

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。
根据文章内容,以下是全文主要内容的简明概括及核心数据: 1. **AI基础设施增长迅速**:预计到2028年,AI基础设施投资将增长至近900亿美元,相比2024年的420亿美元大幅增长。 2. **服务器功耗增加**:预计到2030年,数据中心电力消耗将占总美国需求的12%。 3. **液冷技术的重要性**:液冷被视为AI时代可持续增长的关键技术,可加速AI和高性能计算(HPC)工作负载。 4. **先进冷却解决方案**:文章介绍了多种冷却解决方案,包括3DVC、冷板、风扇、冷板等,以及下一代冷板技术路线图。 5. **液冷PSU性能**:MGX液冷PSU功率密度为120 W/in³,总功耗12KW,冷却液为PG25,入口温度40℃,入口流量12LPM。 6. **液冷系统配置**:包括450kW电源架、热插拔PSU、液冷电源总线、控制板等。
"AI时代,液冷技术如何助力数据中心?" "未来数据中心,冷却解决方案哪家强?" "液冷技术,AI基础设施的强力助手!"
客服
商务合作
小程序
服务号
折叠