透过 Azure HPC 架构打造高效能 AI 云端运算.pdf

1、MicrosoftAccelerating Al thru Azure HPC GPU platformDavid FengAzure HPC/AI Solution Sales LeadGlobal Black BeltMicrosoftAsia#page#Why GPU on cloud vs On-prem ？Huge up-front capital costsLimited space for new hardwareHardware becomes obsolete quicklyLong lead times for shared capacityOperational cost

2、s & constraintsNo scaliing beyond your walls#page#Industry coverage with Azure HPCFinancial servicesEnergyAutomotiveSimulate all aspects of vehicleOptimize all upstream oi and gasConfidently meet regulatoryengineering cost effectively and atrequirements with an elastic andprocesses，including highly

3、complexscalewith highly secure infrastructureintelligent infrastructure forseismic and reservoir simulationsrisk modelingCrash testing simulationsReservoir simulationsAerodynamicsFRTB irpact assessmentsSeismic imaging and modelingAutomotiveEnergyLife sciencesSiliconManufacturingAccelerate insights i

4、n genomics，Achieve fine-grained control overRapidly iterate on product designs toprecision medicine，and clinical trialschip design fow by optimizingimprove quality with scalable and highlywith near-infinite,high-performanceyour chip design processsecure，on-demandinfrastructurebioinformatics infrastr

5、uctureCircuit designComputational fluid dynamicsGenomics analysisClinical trial simulationSilicon manufacturingFinite element analysisife sciencesSiliconManufacturing#page#Achieve more with Azure HPCAfull range of CPU and GPU capabilitiesPurpose-built HPCthat helpapplications scaleto80K+coresFast, s

6、ecureFast InfiniBand inter-connects as wellnetworkingasedge-to-cloud connectivityHigh performingArangeofstoragecapabilitiestosupportstoragesimple-to-complex storageneedsWorkloadEnd-to-endworkflowagilityusingknown，orchestrationfamiliartoolsandprocessesIntelligenceA.machine learning，and deep learnings