当前位置:首页 > 报告详情

从边缘计算到万亿代币人工智能:华硕基础设施满足各种规模的需求.pdf

上传人: 明**** 编号:1011688 2025-12-21 15页 3.68MB

1、From edge to Trillion-Token AI:ASUS Infrastructure for Every Scale2025 OCP Global SummitJerry WangProduct ManagerInfrastructure Solution GroupEvolution of AI Application LandscapePerception AIProcesses&interprets sensory dataAnomaly DetectionPredictive MaintenanceGenerative AIContent creationTextCod

2、eImageAgentic AIAutonomous systems makedecisionsReasoning/PlanningMulti-ModalityMuti-AgentPhysical AIIntegrates AI with physical systemsSelf-Driving CarsGeneral RoboticsAGIA theoretical AI system with capabilities that rival those of a humanSelf-AwarenessHuman CharacteristicMoving from Generative AI

3、 to Agentic AIMore token would be generated when reasoningProcedures to Build an AI Data CenterConsult&DesignDemand evaluationConfiguration and architecture designComputing and network topologyCooling,electricity,layout designInstallServer/switch set upCabling and labelingElectrical/UPSIn-row CDU/si

4、de carFire/safety systemValidateAcceptance test and accept reportingSystem foundation validationPort mappingRobust interconnectivityDeployRapid deploymentCentralized configurationAutomation&systemizationGenAI cloud serviceMaintain&SupportSLA complianceOperations and maintenanceContinuous monitoring

5、and managementAI Data Center Infrastructure PlanningElectricityCoolingLayoutElectrical InfrastructurePower ConsumptionFuture ScalabilityLiquid to Liquid,In-Row CDU Liquid to Air,Side Car Solution Air Conditioning PlanningCooling System Load Capacity Cable Routing Space Planning Safety Key to Success

6、 of Building AI Data Center 5SpecificationPerformanceScheduleCPU/GPU/Memory/NIC/RAID/NVME/Storage.Cores/Spine/Leaf Switch TopologyCabling&Port MappingComputing Power Evaluation Computing Networking BandwidthStorage Networking BandwidthPMOTimeline and MilestoneProject SOWSchedule WBSCoordination amon

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《From edge to Trillion-Token AI: ASUS Infrastructure for Every Scale》的内容,以下是全文关键点的概括: 1. **AI应用领域演变**:从感知AI到生成AI,再到自主系统和物理AI,AI应用不断扩展。 2. **AI数据中心建设**:包括咨询设计、安装、验证、部署、维护和支持等阶段。 3. **基础设施规划**:关注电力、冷却、布局、安全等关键因素。 4. **成功关键**:性能、时间表、资源管理和协调。 5. **ASUS解决方案**:提供从边缘部署到AI数据中心的全套解决方案和服务。 6. **硬件产品**:包括ASUS AI POD、NVIDIA GPU、RTX PRO服务器等。 7. **性能数据**:ASUS AI HPC在台湾的里程碑,如UBILINK AI HPC数据中心达到45.82 PFLOPS。
"AI数据中心建设关键要素" "ASUS AI解决方案优势" "NVIDIA GPU性能解析"
客服
商务合作
小程序
服务号
折叠