Protectionintelxeon
intelxeon 时间:2021-03-27 阅读:()
EmpoweringTransformationinaDigitalWorldAcrossanevolvingdigitalworld,disruptiveandemergingtechnologytrendsinbusiness,industry,science,andentertainmentincreasinglyimpacttheworld'seconomies.
By2020,thesuccessofhalfoftheworld'sGlobal2000companieswilldependontheirabilitiestocreatedigitallyenhancedproducts,services,andexperiences,1andlargeorganizationsexpecttoseean80percentincreaseintheirdigitalrevenues,2alldrivenbyadvancementsintechnologyandusagemodelstheyenable.
Thisglobaltransformationisrapidlyscalingthedemandsforflexiblecompute,networking,andstorage.
Futureworkloadswillnecessitateinfrastructuresthatcanseamlesslyscaletosupportimmediateresponsivenessandwidelydiverseperformancerequirements.
Theexponentialgrowthofdatagenerationandconsumption,therapidexpansionofcloud-scalecomputing,emerging5Gnetworks,andtheextensionofhigh-performancecomputing(HPC)andartificialintelligence(AI)intonewusagesrequirethattoday'sdatacentersandnetworksurgentlyevolve—orbeleftbehindinahighlycompetitiveenvironment.
Thesedemandsaredrivingthearchitectureofmodernized,future-readydatacentersandnetworksthatcanquicklyflexandscale.
TheIntelXeonScalableplatformprovidesthefoundationforapowerfuldatacenterplatformthatcreatesanevolutionaryleapinagilityandscalability.
Disruptivebydesign,thisinnovativeprocessorsetsanewlevelofplatformconvergenceandcapabilitiesacrosscompute,storage,memory,network,andsecurity.
Enterprisesandcloudandcommunicationsserviceproviderscannowdriveforwardtheirmostambitiousdigitalinitiativeswithafeature-rich,highlyversatileplatform.
EnablingGreaterEfficienciesandLowerTCOAcrossinfrastructures,fromenterprisetotechnicalcomputingapplications,theIntelXeonScalableplatformisdesignedfordatacentermodernizationtodriveoperationalefficienciesthatleadtoimprovedtotalcostofownership(TCO)andhigherproductivityforusers.
SystemsbuiltontheIntelXeonScalableplatformaredesignedtodeliveragileservicesandreduceTCOupto65percent4duetolowersoftwareandOSlicensingfees,andacquisition,maintenance,andinfrastructurecosts.
TheFuture-ForwardPlatformFoundationforAgileDigitalServicesIntelXeonScalablePlatformeNterPriSeit5X1.
5XHYbridcLoudiNfrAStructurecoMMuNicAtioNSerViceProViderStecHNicALcoMPutiNG4-SocketHammerDBOLTPDatabaseVirtualizedApplicationsDPDKL3ForwardingIntelDistributionforLINPACKE7-48704YR.
OLDSYSTEME7-8890v4PRIORGEN.
2.
7X1.
7XE5-26584YR.
OLDSYSTEME5-2658v4PRIORGEN.
4.
2X1.
5XE5-26904YR.
OLDSYSTEME5-2699v4PRIORGEN.
8.
2X2.
2XE5-26904YR.
OLDSYSTEME5-2699v4PRIORGEN.
8180818081806152E5-26904TO5YR.
OLDSYSTEMCONSOLIDATIONIntelXeonProcessor8180(SandyBridge:LaunchMarch2012)4.
2:1ComparingIntelXeonScalableplatformtobothfour-year-oldandlast-generationIntelXeonprocessor-basedsystems,thelatestgenerationIntelCPUdeliversbetterperformanceandcapabilitiesforenterprise,cloud,communications,andHPC.
5,8,9,10,11,12,22,24ReducecostsanddelivermoreservicesperserverwithhigherconsolidationofVMsperserverwiththeIntelXeonScalableplatform.
Virtualizationintheenterprisehasbeensteadilyontheriseforadecade.
Mostorganizationshavetakenadvantageofsomeformofvirtualization,creatinggreaterneedforserverstorunvirtualmachines(VMs)inthedatacenter.
IntelXeonScalableplatformenables4.
2x5moreVMsperservercomparedtoearliergenerations,allowingITtoincreasetheirconsolidationofmoreservicesonlesshardware.
2FoundationalEnhancementsHigherPer-CorePerformance:Upto28cores,deliveringhighperformanceandscalabilityforcompute-intensiveworkloadsacrosscompute,storage,andnetworkusages.
GreaterMemoryBandwidth/Capacity:50percentincreasedmemorybandwidthandcapacity.
Sixmemorychannelsversusfourmemorychannelsofpreviousgenerationformemory-intensiveworkloads.
ExpandedI/O:48lanesofPCIe*3.
0bandwidthandthroughputfordemandingI/O-intensiveworkloads.
IntelUltraPathInterconnect(IntelUPI):UptothreeIntelUPIchannelsincreasescalabilityoftheplatformtoasmanyaseightsockets,aswellasimprovesinter-CPUbandwidthforI/O-intensiveworkloadsversuspreviousgeneration3(withIntelQuickPathInterconnect).
IntelUPIofferstheperfectbalancebetweenimprovedthroughputandenergyefficiency.
IntelAdvancedVectorExtensions512(IntelAVX-512):WithdoubletheflopsperclockcyclecomparedtopreviousgenerationIntelAVX2,6IntelAVX-512boostsperformanceandthroughputforthemostdemandingcomputationaltasksinapplications,suchasmodelingandsimulation,dataanalyticsandmachinelearning,datacompression,visualization,anddigitalcontentcreation.
Securitywithoutcompromise:Near-zeroencryptionoverhead7enableshigherperformanceonallsecuredatatransactions.
InnovativeIntegrationsFirst-timeplatformintegrationsdeliverimprovementsinperformanceandlatencyacrosstheinfrastructure:IntegratedIntelOmni-PathArchitecture(IntelOPA)HostFabricInterface:End-to-endhigh-bandwidth,low-latencyfabricthatoptimizesperformanceandeasesdeploymentofHPCclustersbyeliminatingtheneedforadiscretehostfabricinterfacecard.
IntegratedintheCPUpackage.
Pervasive,BreakthroughPerformanceFromitsnewIntelMeshArchitectureandwidelyexpandedresourcestoitshardware-acceleratingandnewlyintegratedtechnologies,theIntelXeonScalableplatformenablesanewlevelofconsistent,pervasive,andbreakthroughperformance.
IntelXeonProcessorprovidesconsistentperformancefordiverseworkloadsWHATMATTERSINTELXEONSCALABLEPLATFORMADVANTAGESMinimizecomplexitywithcompatiblevirtualizationinfrastructureDeployquickly.
IntelVMscoexistw/otherInteltechnology-basedservers.
MeetstringentcustomerSLAsFastresponsetimes.
MaximizevectorfloatingperformanceandefficiencyHighperformancewithfewerservers.
EnsuredeterministicstorageresponseDeterministicperformance.
Cores,cache,memory,I/Oinasingledie.
DeliverdiverseservicesefficientlyIncredibleefficiencyandhardwareaccelerationfromaplatformwithfeaturesrequiredforapplication,control,packet,andsignalprocessing.
IntheIntelXeonScalableplatformIntelMeshArchitecturewithupto28cores,theLastLevelCache(LLC),sixmemorychannels,and48PCIe*channelsaresharedamongallthecores,givingaccesstolargeresourcesacrosstheentiredieandcreatingdynamicscalabilitywithoutsacrificingperformanceforavarietyofdeployments,suchasvirtualization.
VMsscaleeasilywithfullaccesstoallthesharedonboardresourcesthattheyneed.
EnterpriseandCloudHPCStorageCommunicationsXeonXeonXeonXeoniMCXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeoniMCXeonXeonXeonXeon3xDDR4Channels3xDDR4Channels3x16PCIeGen33xUPILinks3IntegratedIntelQuickAssistTechnology(IntelQAT):Chipset-basedhardwareaccelerationforgrowingcompressionandcryptographicworkloadsforgreaterefficiencywhiledeliveringenhanceddatatransportandprotectionacrossserver,storage,andnetworkinfrastructure.
IntegratedIntelEthernetwithscalableiWARP*RDMA*:Providesuptofour10Gbpshigh-speedEthernetportsforhighdatathroughputandlow-latencyworkloads.
Idealforsoftware-definedstoragesolutions,NVMExpress*overFabricsolutions,andvirtualmachinemigrations.
Integratedinthechipset.
Industry-LeadingStorageSupportStorageinnovationscandrivesignificantimprovementsinefficiencyandperformanceofdata-hungryworkloads.
SupportforIntelOptaneSSDsandIntel3DNANDSolidStateDrives:Deliversindustry-leadingcombinationofhighthroughput,lowlatency,highQoS,andultra-highendurancetobreakthroughdataaccessbottlenecks.
DeploynextgenerationstoragewithconfidencewithIntelVolumeManagementDevice(IntelVMD):EnableshotswappingofNVMeSSDsfromthePCIebuswithoutshuttingdownthesystem,whilestandardizedLEDmanagementhelpsprovidequickeridentificationofSSDstatus.
Thiscommonalitybringsenterprisereliability,availability,andserviceability(RAS)featurestoNVMeSSDs,enablingdeploymentofnext-generationstoragewithconfidence.
IntelIntelligentStorageAccelerationLibrary(IntelISA-L):Optimizesstorageoperations,suchasencryption,forincreasedstorageperformance.
––Upto3.
1xfastercryptographichashingofSHAalgorithmsusingIntelAVX-512.
14––Upto1.
2xalgorithmspeedupforAES-128-GCM.
14––Upto2xfasterReedSolomonErasureCodeperformanceusingIntelAVX-512.
14ComplementaryOfferingsforEvenGreaterPerformance,ScalabilityInteloffersabroadhardwareandsoftwareportfoliothatcomplementsthisnewprocessor.
TheIntelXeonPhiProcessorisanidealfoundationforhighlyparallelapplications,suchasmachinelearningtraining,simulation,andvisualization.
IntelFPGAsofferefficientacceleration15alongwiththeflexibilitythatcomeswithusingprogrammablehardwareforlow-latencyapplications,suchasvirtualswitching,networkservices,dataanalytics,andAI.
ArangeofsoftwaretoolsandlibrariesforgeneralandhighlyparallelcomputinghelpdevelopersoptimizeapplicationsforIntelArchitecture.
EnhancedPlatformTrustDataandplatformreliabilityandprotectionarekeyconcernsforenterprisesdealingwithincreasingconcernsandscrutinyregardingdatasecurityandprivacy.
IntelXeonScalableplatformhelpsbuildhighlytrustedinfrastructureswithplatformdataprotection,resiliency,anduptime.
IncreasedDataProtectionandReliabilityAcrossEveryWorkloadEnhancedIntelRunSureTechnology:NewenhancementsdeliveradvancedReliability,Availability,andServiceability(RAS)andserveruptimeforacompany'smostcriticalworkloads.
Hardware-assistedcapabilities,includingenhancedMCAandrecoveryandadaptivemulti-deviceerrorcorrection,diagnoseandrecoverfrompreviouslyfatalerrors.
And,theyhelpensuredataintegritywithinthememorysubsystem.
IntelKeyProtectionTechnology(IntelKPT)withIntegratedIntelQATandIntelPlatformTrustTechnology(IntelPTT):Deliverhardware-enhancedplatformsecuritybyprovidingefficientkeyanddataprotectionatrest,in-use,andin-flight.
IntelTrustedExecutionTechnology(IntelTXT)withOne-TouchActivation:Enhancedplatformsecurity,whileprovidingsimplifiedandscalabledeploymentforIntelTXT.
Asmoredata-richworkloadsflowthroughthedatacenter,thiscomprehensivesuiteofhardware-enhancedfeaturesbringsbetterdata-andplatform-levelprotectionmechanismsfortrustedservicesinenterpriseandcloudenvironments.
4DynamicandHighlyEfficientServiceDeliveryTheconvergenceofenhancedcompute,memory,network,andstorageperformance,combinedwithsoftwareecosystemoptimizations,makeIntelXeonScalableplatformtheidealplatformforfullyvirtualized,software-defineddatacentersthatdynamicallyself-provisionresources—on-premise,throughthenetwork,andinthepubliccloud—basedonworkloadneeds.
PowerfulToolsandTechnologiesforanAgileDataCenterIntelVT-xnewfeatures:–Modebasedexecutioncontrol(MBE)virtualization:Providesanextralayerofprotectionfrommalwareattacksinavirtualizedenvironmentbyenablinghypervisorstomorereliablyverifyandenforcetheintegrityofkernellevelcode.
–Timestampcounterscaling(TSC)virtualization:ProvidesworkloadoptimizationinhybridcloudenvironmentsbyallowingvirtualmachinestomoveacrossCPUsoperatingatdifferentbasefrequencies.
IntelNodeManager4.
0:HelpsITintelligentlymanageandoptimizepower,cooling,andcomputeresourcesinthedatacenter,maximizingefficiency,whilereducingthechancesofcostlyoverheats.
Strong,CapablePlatformsfortheHybrid-Cloud,Data-FueledEnterpriseEnterprisesarekeentoextractvaluefromtheexplodingdatastreamsbeingpresentedtothemforrapidinsightsthatcanshapetheirbusinessinitiatives.
Traditionalandemergingapplicationsintheenterprise,includingpredictiveanalytics,machinelearning,andHPC,requirenewlevelsofpowerfulcomputecapabilitiesandmassivetiereddatastoragevolumes.
ThemodernizeddatacenterisbeingarchitectedusingaconvergedandholisticapproachthatcanflexiblydelivernewservicesandimproveTCOacrossinfrastructureassetstoday,whileprovidingthemostseamlessandscalableon-ramptoaself-governing,hybriddatacenter.
Yet,organizationsrunningtheirfoundationalbusinessworkloads,suchasOLTPandwebinfrastructure,seektoreduceTCOwithhigherperforminginfrastructures.
TheIntelXeonScalableplatformdeliversnext-generationenterprisecapabilitiestobusinessesthroughafuture-readyplatformthatcanservethehybrid-cloud,data-fueledera,plusithelpsimproveday-to-dayoperationswithupto58percentmorerequests/secondonruntimeworkloads.
16Thisversatileplatformbringsdisruptivelevelsofcomputeperformance,coupledwithmemoryandI/Oadvances,tocompute-hungryandlatency-sensitiveapplications.
CombinedwithinnovativeIntelSSDDataCenterFamilytomanagelargedatavolumesacrossstorage,caching,andmemory,platformsbuiltontheIntelXeonScalableplatformarereadytohandletheintensedemandsofthedataandcloudera.
Withascalableportfolioofpackagingoptionstosuitdiverseworkloadrequirements,theIntelXeonScalableplatformisaperformanceworkhorsedesignedfordeployinghighlyefficient,virtualizedinfrastructuresforcompute,storage,andnetworking.
HighlightsforEnterpriseInnovationUpto28high-performancecores,sixmemorychannels,and48lanesofPCIe3.
0UptothreeIntelUPIchannelsSupportforIntelOptaneSSDsandIntel3DNANDSSDsEnhancedIntelRunSureTechnologyNext-GenerationPlatformforCloud-Optimized,5G-ReadyNetworks,andNext-GenerationVirtualNetworksThecomingeraof5Gwillenableentirelynewecosystemsandclassesofconsumerandenterpriseservicesalongwithmediaapplicationsonwirelessandwirelinenetworks.
Thesedata-rich,innovativeusecases,drivenbythenewInternetofThings(IoT),visualcomputing,andanalytics,representsignificantfutureopportunitiesforcommunicationsserviceproviders(CommSPs)togrowrevenue.
Thetransitionfrompurpose-built,fixedfunctioninfrastructuretoanewgenerationofopennetworksistheessentialfirststeptopreparefora5G-enabledworld.
Software-definednetworkingwithNetworkFunctionsVirtualization(NFV)isenablingnewserviceopportunitiesandoperationsefficienciesforbothcommunicationservicesprovidersandenterprisesalike.
Usingflexible,optimized,industry-standardserversandvirtualized,orchestratednetworkfunctionswillallowfuture-readyinfrastructurestobeabletodeliverinnovativeserviceswithefficiencyandease.
Suchdistributedcommunicationsnetworkscansupportextremelevelsofscalability,agility,programmability,andsecurityacrossanever-growingvolumeandvarietyofnetworkingworkloads—fromthenetworkcoretotheedge.
TheIntelXeonScalableplatformisthebasisfornext-generationplatformstobuildvirtualized,cloud-optimized,5G-readynetworks.
Itoffersanarchitecturethatscalesandadaptswitheasetohandlethedemandsofemergingapplicationsandtheconvergenceofkeyworkloads,suchasapplicationsandservices,controlplaneprocessing,high-performancepacketprocessing,andsignalprocessing.
Thisnewprocessorprovidesafoundationforagilenetworksthatcanoperatewithcloudeconomics,behighlyautomatedandresponsive,andsupportrapidandmoresecuredeliveryofnewandenhancedservicesenabledby5G.
5AccelerateNetworkTrafficwithIntelXeonScalablePlatform,DPDK,andIntelQAT2.
521.
510.
50DPDKL3PacketForwarding(Gbps)PreviousIntelXeonScalablePlatformIPSec(Gbps)SSLWebProxy(Gbps)DPDKoptimizations17:Utilizenewcachehierarchywithlargemid-levelcacheandbalancedbandwidthimprove-mentstoaddressdata,audio,andvideogrowth.
IPSec18:Providedataauthentication,integrity,andconfidentialitywithacceleratedcryptographyandcompression/decompression.
SSLWebProxy19:Acceleratedataencryption/decryptionforsecuretransmissions.
1.
77X2.
5X2.
4XHiGHeriSbetterCombinedwiththeDataPlaneDevelopmentKit(DPDK)andIntelQAT,thisnewprocessorcanboostnetworkperformance,enablingserviceproviderstoaccommodategreatertraffictogrowservices—andrevenue—andhelpingthemtopreparefor5G.
HighlightsforCommSPInnovationHardware-basedaccelerationofencryptionandcompressionusingintegratedIntelQAT.
IntegratedIntelEthernetIntelFPGAsmaximizeversatilityincommunicationsinfrastructureIntelKPTwithIntegratedIntelQATandIntelPTTIntelTXTwithOne-TouchActivationAdditionalResourcesOptimizedforCommSPsTheopensourceDataPlaneDevelopmentKit(DPDK)enablesoptimizedcommunicationsoperationsonIntelArchitecture.
DPDKhasdemonstratedabilitytoscaleperformanceasprocessorcorecountandperformanceincrease;workloads,suchasVectorPacketProcessing(VPP)IPSec,benefitfromthisenhancedperformance.
Additionally,theselibrariesprovidepre-optimizedmechanismstoallownewprocessorcapabilities(suchasIntelAVX-512andmemoryandI/Oenhancements)tobeabletoutilizethenewfunctionalityforimprovedpacketprocessingperformancewithlessdirectdevelopmenteffort.
Inteloffersprograms,suchasIntelNetworkBuildersUniversity,idealfornetworkevolutioninthe5Gera.
Withsolutionguidanceandtrainingfromtheseprograms,CommSPscandrivetheirnetworktransformationinitiativesforwardwithincreasedconfidence.
BreakthroughHPCandHigh-PerformanceDataAnalyticsInnovationToday'sscientificdiscoveriesarefueledbyinnovativealgorithms,newsourcesandvolumesofdata,andadvancesincomputeandstorage.
Benefittingfromexponentiallyexpandingvolumesandvarietyofdata,HPCclustersarealsotheengineforrunningevolvingHigh-PerformanceDataAnalytics(HPDA)workloads,leadingtoincrediblediscoveriesandinsightforbusinessandhumanunderstanding.
Machinelearning,deeplearning,andAIconvergethecapabilitiesofmassivecomputewiththefloodofdatatodrivenext-generationapplications,suchasautonomoussystemsandself-drivingvehicles.
IntelXeonScalableplatformoffersacommonplatformforAIwithhighthroughputforbothinferenceandtraining—upto18xhigherinference20andupto19xhighertraining21throughput,comparedtofour-year-oldsystems.
HPCisnolongerjustthedomainoflargescientificinstitutions.
EnterprisesareincreasinglyconsumingamassivenumberofHPCcomputecycles;someoftheworld'slargestHPCclustersareinprivateoilandgascompanies.
ResearchinpersonalizedmedicineappliesHPCforhighlyfocusedtreatmentplans.
NewHPCinstallationsareengaginginnovative,convergedarchitecturesfornon-traditionalusagesthatcombinesimulation,AI,visualization,andanalyticsinasinglesupercomputer.
6HPCplatforms—fromthesmallestclusterstolargestsupercomputers—demandabalanceacrosscompute,memory,storage,andnetwork.
TheIntelXeonScalableplatformwasdesignedtodeliverandenablesuchbalancewithmassivescalability—totensofthousandsofcores.
FromitsimprovedcorecountandmesharchitecturetonewlyintegratedtechnologiesandsupportforIntelOptanememoryandstoragedevices,theIntelXeonScalableplatformenablestheultimategoalsofHPC—tomaximizeperformanceacrosscompute,memory,storage,andnetworkwithoutinducingbottlenecksatanyintersectionofresources.
TheintegrationofIntelOmni-PathArchitecture,anend-to-endhigh-performancefabric,intotheIntelXeonScalableplatformdeliversbothincreasedperformanceandscalingtodistributed,parallelcomputingclusters.
Nearlinearscalingupto32nodesenablesbuildinglargeHPCsolutionsthatarenotinhibitedbytheinterconnect.
And,theIntelXeonScalableplatform,comparedtopreviousgenerationIntelXeonprocessors,delivers1.
63xbetteroverallperformanceacross13commonHPCworkloadsusedthroughoutthesciencesandfinancialservices.
23TheIntelXeonScalableplatformandIntelOmni-PathArchitecturecanenablenewdiscoveriesandfastersolutionsforhighlyparallelworkloadsinmanydatacenters.
HighlightsforHPCInnovationIntelUltraPathInterconnectIntelAdvancedVectorExtensions512IntegratedIntelOmni-PathArchitectureHostFabricInterfaceSupportforIntelOptaneSSDsandIntel3DNANDSolidStateDrivesIntelXeonGold6148processor3.
02.
52.
01.
51.
00.
50HigherisBetterBroadwell(E5-2697v4)WRF*HOMME*LSTCLS-DYNAExplicit*INTESPERMAS*V16MILC*GROMACS*VASP*NAMD*LAMMPS*AmberGB*BinomialOptionPricingBlack-ScholesMonteCarloEuropeanOptions1.
001.
411.
681.
411.
441.
521.
561.
581.
671.
731.
751.
421.
872.
38ManufacturingEarthSystemsModelsLifeSciencesFSIAcross13commonHPCworkloads,IntelXeonScalableplatformdeliversimprovedperformance.
AdditionalTechnologiesforHPC,HPDA,andAIThebootableIntelXeonPhiProcessor7200Series,withitsManyIntegratedCore(MIC)architecture,isanidealfoundationforhighlyparallelworkloads,suchasmachinelearningtraining,simulation,andvisualization.
Arangeofhigh-productivitysoftwaretools,optimizedlibraries,foundationalbuildingblocks,andflexibleframeworksforgeneralandhighlyparallelcomputinghelpsimplifyworkflowsandassistdeveloperstocreatecodesthatmaximizethecapabilitiesofIAforHPCandAI.
OptimizationsforpopulardeeplearningframeworksforIA,includingNeon,*Caffe,*Theano,*Torch,*andTensorFlow*offerincreasedvalueandperformancefordatascientists.
IntelParallelStudioXE2017includesperformancelibraries,suchasIntelMathKernelLibraryforDeepNeuralNetworks(IntelMKL-DNN)toacceleratedeeplearningframeworksonIA,andIntelDataAnalyticsAccelerationLibrary(IntelDAAL)tospeedbigdataanalytics.
ResourcesOptimizedforHPCTocontinuetoadvancediscoverythroughHPCintotheExascaleera,theIntelModernCodeDeveloperProgramoffersdevelopersanddatascientistseasilyaccessibleonlineandface-to-facecodemodernizationtechnicalsessionsontechniques,suchasvectorization,memoryanddatalayout,multi-threading,andmulti-nodeprogramming.
7IntelXeonScalablePlatformataGlanceIntelXeonScalableplatformoffersarangeofperformance,scalability,andfeatureoptionstomeetawidevarietyofworkloadsinthedatacenter,fromthemostadvanced(IntelXeonPlatinum8100processors)toentry-level(IntelXeonBronze3100processors).
BESTPERFORMANCE,MOSTSCALABLE,BESTBUSINESSAGILITYIntelXeonPlatinum8100ProcessorsBestchoiceformission-critical,real-timeanalytics,machinelearning,andartificialintelligenceworkloadsBestworkload-optimizedperformanceforgeneralpurposecomputeandhybrid-clouddeploymentsBestperformanceforthemostdemandingstorageandnetworkingworkloadsBestmemorybandwidthand2,4,8+socketscalabilityGREATPERFORMANCE,FASTMEMORY,ANDMOREINTERCONNECT/ACCELERATORENGINESIntelXeonGold6100ProcessorsSignificantworkload-optimizedperformanceimprovementsforgeneralpurposecomputeSignificantimprovementsfordemandingstorageandnetworkingworkloadsHighestmemoryspeed,highestmemorycapacity,andenhancedIntelAVX-512Enhanced2-4socketscalabilityandperformanceBETTERPERFORMANCE,ADVANCEDRELIABILITYIntelXeonGold5100ProcessorsImprovedperformanceforcompute-boundworkloadsAffordableAdvancedRASand4-socketscalabilitySuitableforawiderrangeofworkloadsEFFICIENTPERFORMANCEATLOWPOWERIntelXeonSilver4100ProcessorsSolidcomputecapability(Hyper-Threading,TurboBoost)Improvedmemoryspeed,energyefficiencySuitableforamoderaterangeworkloadsENTRY-LEVELPERFORMANCEANDHW-ENHANCEDSECURITYIntelXeonBronze3100ProcessorsAffordable,entry-level2-socketsupportsuitableforlight-rangeworkloadsReliableupgradeversusIntelXeonprocessorE3productfamily8IntelXeonBronzeProcessor(3100Series)IntelXeonSilverProcessor(4100Series)IntelXeonGoldProcessor(5100Series)IntelXeonGoldProcessor(6100Series)IntelXeonPlatinumProcessor(8100Series)PERVASIVEPERFORMANCEANDSECURITYHighestCoreCountSupported8cores12cores14cores22cores28coresHighestSupportedFrequency1.
7GHz(8C/85W)2.
2GHz(10C/85W)3.
6GHz(4C/105W)3.
4GHz(6C/115W)3.
6GHz(4C/105W)NumberofCPUSocketsSupportedUpto2Upto2Upto4Upto4Upto8IntelUltraPathInterconnect(UPI)22233IntelUPISpeed9.
6GT/s9.
6GT/s10.
4GT/s10.
4GT/s10.
4GT/sIntelAdvancedVectorExtensions512(AVX-512)1FMA1FMA1FMA2FMA2FMAMemorySpeedSupport(DDR4)2133MHz2400MHz2400MHz2666MHz2666MHzHighestMemoryCapacitySupportedPerSocket768GB768GB768GB768GB,1.
5TB768GB,1.
5TBIntelOmni-PathArchitecture(DiscretePCIe*card)IntelQuickAssistTechnology(Integratedinchipset)IntelQuickAssistTechnology(DiscretePCIecard)IntelOptaneTechnology-basedSSDs(3DXPoint)IntelSSDDataCenterFamily(3DNAND)PCIe3.
0(48lanes)IntelQuickDataTechnology(CBDMA)Non-TransparentBridge(NTB)IntelTurboBoostTechnology2.
0IntelHyper-ThreadingTechnologyNodeControllerSupportIntelOmni-PathArchitecture(Integrated)HIGHRELIABILITYReliability,Availability,andServiceability(RAS)CapabilityStandardStandardStandardAdvancedAdvancedIntelRunSureTechnologyAGILITY&EFFICIENCYIntelVolumeManagementDevice(IntelVMD)IntelVT-xIntelSpeedShiftTechnologyIntelNodeManager4.
0SECURITYMode-basedExecuteControlIntelKeyProtectionTechnology(KPT)w/IntegratedIntelQATIntelPlatformTrustTechnology(PTT)IntelTXTw/One-TouchActivation(OTA)FamilyFeatures9IntelXeonScalablePlatformSKUNumberingProcessornumbersfortheIntelXeonScalableplatformuseanalphanumericschemebasedonperformance,features,processorgeneration,andanyoptions,followingthebrandanditsclass.
SKULevel8=Platinum6,5=Gold4=Silver3=BronzeProcessorGeneration1=1stGen(Skylake)ProcessorSKU(ex.
20,34…)IntegrationsandOptimizations(ifapplicable)F=FabricT=HighTcase/ExtendedReliabilityMemoryCapacityNoSuffix=768GBpersocketM=1.
5TBpersocketIntelXeonPlatinumprocessorαα##18IntelXeonGoldprocessorαα##16IntelXeonGoldprocessorαα##15IntelXeonSilverprocessorαα##14IntelXeonBronzeprocessorαα##1310IntelXeonScalablePlatformSKUsSKUSOPTIMIZEDFORHIGHESTPER-COREPERFORMANCEClassSKUCoresBaseNon-AVXSpeed(GHz)TDP(W)Platinum8180a282.
5205Platinum8168242.
7205Platinum8158b123.
0150Platinum8156c43.
6105Gold6148202.
4150Gold6154183.
0200Gold6150182.
7165Gold6142a162.
6150Gold6132142.
6140Gold6146b123.
2165Gold6136b123.
0150Gold6126d122.
6125Gold6144b83.
5150Gold6134a,b83.
2130Gold6128d63.
4115Gold5122c43.
6105SKUSOPTIMIZEDFORBALANCEDENERGYEFFICIENTPERFORMANCE/WATTClassSKUCoresBaseNon-AVXSpeed(GHz)TDP(W)Platinum8176a282.
1165Platinum8170a262.
1165Platinum8164262.
0150Platinum8160a242.
1150Platinum8153e162.
0125Gold6152222.
1140Gold6138202.
0125Gold6140a182.
3140Gold6130162.
1125Gold5120142.
2105Gold5118122.
3105Gold5115102.
485Silver4116122.
185Silver4114102.
285Silver411242.
685Silver411082.
185Silver410881.
885Bronze310681.
785Bronze310461.
78511IntelXeonScalablePlatformSKUs,continuedSKUSWITHEXTENDEDLIFE(10-YEARUSE)ANDNEBS-FRIENDLYTHERMALSPECIFICATIONClassSKUCoresBaseNon-AVXSpeed(GHz)TDP(W)Platinum8160T242.
1150Gold6138T202.
0125Gold6130T162.
1125Gold6126Td122.
6125Gold5120T142.
2105Gold5119T141.
985Silver4116T122.
185Silver4114T102.
285Silver4109T82.
070SKUSWITHINTEGRATEDINTELOMNI-PATHARCHITECTUREFABRICClassSKUCoresBaseNon-AVXSpeed(GHz)TDP(W)Platinum8176F282.
1173Platinum8160F242.
1160Gold6148F202.
4160Gold6142F162.
6160Gold6138F202.
0135Gold6130F162.
1135Gold6126Fd122.
6105PRODUCTNAMESKU10Gb/1GbETHERNETPORTSCOMPRESSIONENCRYPTIONRSAIntelQuickAssistTechnologyIntelC621ChipsetLBG-1G0/4N/AN/AN/AIntelC622ChipsetLBG-22/4N/AN/AN/AIntelC624ChipsetLBG-44/4N/AN/AN/AIntelC625ChipsetLBG-E4/420Gb/s20Gb/s20KOps/sIntelC626ChipsetLBG-M4/440Gb/s40Gb/s40KOp/sIntelC627ChipsetLBG-T4/4100Gb/s100Gb/s100KOps/sIntelC628ChipsetLBG-L4/4100Gb/s100Gb/s100KOps/sa"M"SKUavailablesupporting1.
5TB/socketbNon-default24.
75MBcachecNon-default16.
5MBcachedNon-default19.
25MBcacheeCompatiblefor8-socketbuilds121Source:IDC,http://www.
idc.
com/getdoc.
jspcontainerId=prUS41888916.
2Source:GartnerGroup,http://www.
gartner.
com/newsroom/id/3142917.
3Moreinter-CPUbandwidth.
10.
4GT/sonIntelXeonScalableprocessorsvs.
9.
6GT/sonIntelXeonprocessorE5-2600v4productfamily.
4Upto65%lower4-yearTCOestimateexamplebasedonequivalentrackperformanceusingVMwareESXi*virtualizedconsolidationworkloadcomparing20installed2-socketserverswithIntelXeonprocessorE5-2690(formerly"SandyBridge-EP")runningVMwareESXi*6.
0GAusingGuestOSRHEL6.
4comparedatatotalcostof$919,362to5newIntelXeonPlatinum8180(Skylake)runningVMwareESXi6.
0U3GAusingGuestOSRHEL664bitatatotalcostof$320,879includingbasicacquisition.
ServerpricingassumptionsbasedoncurrentOEMretailpublishedpricingforBroadwellbasedservers–subjecttochangebasedonactualpricingofsystemsoffered.
TotalCostofOwnershipEstimatedServerA:2SXeonE5-2690(2.
9GHz,8C,20MB)ServerB:2SXeonPlatinum8180(2.
5GHz,28C,38.
5M)NumberofServers205ServerAcquisition$273,220$150,995OperatingSystem&SoftwareOSLicense$123,100$30,775OSMaintenance$241,360$60,340SoftwareLicense$39,800$9,950SoftwareMaintenance$51,680$12,920ServerMaintenance$134,080$33,520Infrastructure&UtilitiesPowerandCooling$42,522$18,979Rack/FloorSpace$12,400$3,100Networking$1,200$300Total$919,362$320,8795Upto4.
2xmoreVMsbasedonservervirtualizationconsolidationworkload:BasedonIntelinternalestimates1-Node,2xIntelXeonProcessorE5-2690onRomley-EPwith256GBTotalMemoryonVMwareESXi*6.
0GAusingGuestOSRHEL6.
4,glassfsh3.
1.
2.
2,postgresql9.
2.
DataSource:RequestNumber:1718,Benchmark:servervirtualizationconsolidation,Score:377.
6@21VMsvs.
1-Node,2xIntelXeonPlatinum8180ProcessoronWolfPassSKXwith768GBTotalMemoryonVMwareESXi6.
0U3GAusingGuestOSRHEL664bit.
DataSource:RequestNumber:2563,Benchmark:servervirtualizationconsolidation,Score:1580@90VMs.
Higherisbetter.
6AsmeasuredbyIntelXeonProcessorScalableFamilywithIntelAVX-512comparedtoanIntelXeonE5v4withIntelAVX2.
7BigBench,NearZeroencryptionoverhead:BigBenchqueryRuntime/second.
TestingdonebyIntel.
BASELINE:Platform8168,NODES1Mgmt+6Workers,MakeIntelCorporation,ModelS2600WFD,FormFactor2U,ProcessorIntelXeonPlatinum8168,BaseClock2.
70GHz,Corespersocket24,Hyper-ThreadingEnabled,NUMAmodeEnabled,RAM384GBDDR4,RAMType12x32GBDDR4,OSDriveIntelSSDDCS3710Series(800GB,2.
5inSATA6Gb/s,20nm,MLC),DataDrives8x–SeagateEnterprise2.
5HDDST2000NX04032TB,IntelSSDDCP3520Series(2.
0TB),TempDriveDC35202TB,NICIntelX72210GBE–DualPort,HadoopCloudera5.
11,BenchmarkTPCx-BB1.
2,OperatingSystemCentOSLinuxrelease7.
3.
1611(Core);HDFSencryp-tionturnedOFF.
vs.
NEW:Platform8168,NODES1Mgmt+6Workers,MakeIntelCorporation,ModelS2600WFD,FormFactor2U,ProcessorIntelXeonPlatinum8168,BaseClock2.
70GHz,Corespersocket24,Hyper-ThreadingEnabled,NUMAmodeEnabled,RAM384GBDDR4,RAMType12x32GBDDR4,OSDriveIntelSSDDCS3710Series(800GB,2.
5inSATA6Gb/s,20nm,MLC),DataDrives8x–SeagateEnterprise2.
5HDDST2000NX04032TB,IntelSSDDCP3520Series(2.
0TB),TempDriveDC35202TB,NICIntelX72210GBE–DualPort,HadoopCloudera5.
11,BenchmarkTPCx-BB1.
2,OperatingSystemCentOSLinuxrelease7.
3.
1611(Core);HDFSencryptionturnedON.
8Upto5xclaimbasedonOLTPWarehouseworkload:1-Node,4xIntelXeonProcessorE7-4870onEmeraldRidgewith512GBTotalMemoryonOracleLinux*6.
4usingOracle12c*running800warehouses.
DataSource:RequestNumber:56,Benchmark:HammerDB,Score:2.
46322e+006Higherisbettervs.
1-Node,4xIntelXeonPlatinum8180ProcessoronLightningRidgeSKXwith768GBTotalMemoryonRedHatEnterpriseLinux*7.
3usingOracle12.
2.
0.
1(includingdatabaseandgrid)with800warehouses.
DataSource:RequestNumber:2542,Benchmark:HammerDB,Score:1.
2423e+007Higherisbetter.
9Upto1.
52xclaimbasedonHammerDB:1-Node,4xIntelXeonProcessorE7-8890v4onBrickland(Patsburg)with1536GBTotalMemoryonOracleLinux*7.
1usingOracle12.
1.
0.
2.
0(in-cludingdatabaseandgrid)with800warehouses.
DataSource:RequestNumber:2239,Benchmark:HammerDB,Score:8.
17145e+006vs.
1-Node,4xIntelXeonPlatinum8180ProcessoronLightningRidgeSKXwith768GBTotalMemoryonRedHatEnterpriseLinux*7.
3usingOracle12.
2.
0.
1(includingdatabaseandgrid)with800warehouses.
DataSource:RequestNumber:2542,Benchmark:HammerDB,Score:1.
2423e+007Higherisbetter.
10Upto1.
5xmoreVMsbasedonvirtualizationconsolidationworkload:1-Node,2xIntelXeonProcessorE5-2699v4onGrantley-EP(Wellsburg)with512GBTotalMemoryonVMwareESXi*6.
0Update1usingGuestVM'sutilizeRHEL664bitOS.
DataSource:RequestNumber:1637,Benchmark:servervirtualizationworkload,Score:1034@58vs.
1-Node,2xIntelXeonPlatinum8180ProcessoronWolfPassSKXwith768GBTotalMemoryonVMwareESXi6.
0U3GAusingGuestVM'sutilizeRHEL664bitOS.
DataSource:RequestNumber:2563,Benchmark:servervirtualizationworkload,Score:1580@90VMsHigherisbetter.
11Upto2.
7xclaimbasedonDPDKL3PacketForwarding:IntelXeonE5-2650processors2.
00GHz,8GT/sQPI,20MBL3cache,PatsburgChipset(C0stepping),GrizzlyPassPlatform(R2216G-ZBPP),DDR31333MHz,8xdualrank4GB(total32GB),4memorychannelspersocketConfiguration,1DIMMperchannel,6xIntel82599dual-portPCI-ExpressGen2x810GbEthernetNIC,1xIntel82599dual-portGen2x8I/Oexpansionmodule10GbEthernetNIC,BIOSversionS5500.
86B.
01.
00.
0048,Operatingsystem:FedoraCore15,Kernelversion:2.
6.
38.
4,IxNetwork*6.
0.
400.
22,DPDKversion:FD5_1Score:102Gbits/spacketforwardingat256Busingcoresvs.
Gold6152:EstimatesbasedonIntelinternaltestingonIntelXeon61522.
1GHz,2xIntel,FM10420(RRC)GenDualPort100GbEEthernetcontroller(100Gbit/card)2xIntelXXV710PCIExpressGenDualPort25GbEEthernetcontroller(2x25G/card),DPDK17.
02.
Score:281Gbits/spacketforwardingat256Bpacketusingcores,IOandmemoryonasinglesocket.
12Upto1.
7xclaimbasedonDPDKL3PacketForwarding:E5-2658v4:5xIntelXL710-QDA2,DPDK16.
04.
Benchmark:DPDKl3fwdsampleapplicationScore:158Gbits/spacketforwardingat256Bpacketusingcores.
Gold6152:EstimatesbasedonIntelinternaltestingonIntelXeon61522.
1GHz,2xIntel,FM10420(RRC)GenDualPort100GbEEthernetcontroller(100Gbit/card)2xIntelXXV710PCIExpressGenDualPort25GbEEthernetcontroller(2x25G/card),DPDK17.
02.
Score:281Gbits/spacketforwardingat256Bpacketusingcores,IOandmemoryonasinglesocket.
13SystemConfguration:4NodevSAN*Cluster.
PerNodeconfiguration:Supermicro*SuperServer2028U-TN24R4T+DualIntelXeonE5-2687Wv4(12Core@3.
0Ghz),Supermicro*ServerBoard,256GBDDR4RAM,BootDrive,1xIntelSSDDCS3710Series(200GB,2.
5"),vSANIntel3DNANDCluster:VirtualSANSSDs-2DiskGroupscomprisedof2xIntelSSDDCP4600Series(1.
6TB,2.
5"SFF),8xIntelSSDDCP4500Series(4TB,2.
5"SFF),vSANIntel2DNANDCluster:VirtualSANSSDs-2DiskGroupscomprisedof2xIntelSSDDCP3700Series(800GB,2.
5"SFF),8xIntelSSDDCP3500Series(2TB,2.
5"SFF),IntelEthernetServerAdapterX540-DA2.
143.
1x,1.
2x,2xISA-Lconfiguration.
IntelXeonProcessorScalableFamily:Platinum8180Processor,28C,2.
5GHz,H0,NeonCityCRB,12x16GBDDR42666MT/sECCRDIMM,BIOSPLYCRB1.
86B.
0128.
R08.
1703242666.
IntelXeonE5-2600v4SeriesProcessor,E5-2650v4,12C,2.
2GHz,AztecCityCRB,4x8GBDDR42400MT/sECCRDIMM,BIOSGRRFCRB1.
86B.
0276.
R02.
1606020546.
OperatingSystem:RedhatEnterpriseLinux7.
3,Kernel4.
2.
3,ISA-L2.
18,BIOSConfguration,P-States:Disabled,Turbo:Disabled,SpeedStep:Disabled,C-States:Disabled,ENERGY_PERF_BIAS_CFG:PERF.
15AsmeasuredbyIntelXeonProcessorScalableFamilywithIntelFPGAoptimizedworkloadandIntelXeonProcessorScalableFamilywithoutFPGAoptimizedworkload.
16Upto1.
58xclaimbasedonGhost-NodeJSworkload:1-Node,2xIntelXeonProcessorE5-2699v4onWildcatPasswith384GBTotalMemoryonUbuntu16.
04LTSusingNode.
jsversion6.
9.
2,MySQLMariaDBversion15.
1Distrib10.
0.
30.
DataSource:RequestNumber:2687,Benchmark:Ghost-NodeJS,Score:2308Higherisbettervs.
1-Node,2xIntelXeonPlatinum8180ProcessoronWolfPassSKXwith384GBTotalMemoryonUbuntu16.
10usingNode.
jsversion6.
9.
2,MySQLMariaDBversion15.
1Distrib10.
0.
30.
DataSource:RequestNumber:2687,Benchmark:Ghost-NodeJS,Score:3647Higherisbetter.
17Upto1.
77xclaimbasedonDPDKL3Forwarding:E5-2658v4:5xIntelXL710-QDA2,DPDK16.
04.
Benchmark:DPDKl3fwdsampleapplication.
Vs.
Gold6152:EstimatesbasedonIntelinternaltestingonIntelXeon61522.
1GHz,2xIntel,FM10420(RRC)GenDualPort100GbEEthernetcontroller(100Gbit/card)2xIntelXXV710PCIExpressGenDualPort25GbEEthernetcontroller(2x25G/card),DPDK17.
02.
Score:281Gbits/spacketforwardingat256Bpacketusingcores,IOandmemoryonasinglesocket.
18Upto2.
5xclaimbasedonDPDKIPSecSeg-gwbenchmark:IntelXeonE5-2658v4,IntelPCHC612,DDR4-2400Intel895XCCbasedQuickAssistAcceleratorAdapterPCIeGen3x8links,DPDK16.
11IPSec-secgw,1420Bpacket).
IntelDPDK16.
11IPsec-secgwsampleapplication.
Cores,IO,packetbuffermemory,andprocessingcoresareonasinglesocket.
6coresusedononeSocket,Cryptoalgorithm:AES-128-CBC-HMAC-SHA1vs.
IntelXeon61522.
1GHz,3xIntelCorporation,EthernetControllerX710(4x10Gbeportspercard),LewisburgB0QuickAssistAcceleratorwithPCIeGen3x24links,IntelDPDK17.
02IPsec-secgw,IntelQAT1.
7.
Upstream.
L.
1.
0.
0-15,6coresused.
Cores,IO,packetbuffermemory,andprocessingcoresareonasinglesocket.
6coresusedononeSocket,Cryptoalgorithm:AES-128-CBC-HMAC-SHA1.
2017IntelCorporation.
Intel,theIntellogo,IntelXeon,IntelOptaneandIntelXeonPhiaretrademarksofIntelCorporationintheU.
S.
and/orothercountries.
*Othernamesandbrandsmaybeclaimedasthepropertyofothers.
PrintedinUSA0717/SC/HBD/PDFPleaseRecycle336198-00119Upto2.
4xclaimbasedonTLSWebProxyusingNGINX:IntelXeonE5-2658v4,DDR4-2133,IntelPCHC612,Intel895XCCbasedQuickAssistAcceleratorAdapterPCIeGen3x8links,OpenSSL-Async(0.
4.
9-009)+NGINX-1.
6.
2(0.
1.
0-008),QAT1.
6.
L.
2.
6.
0-60.
Cores,IO,packetbuffermemory,andprocessingcoresareonasinglesocket.
6coresusedononeSocket12Coresareused,Cryptoalgorithm:AES-128-CBC-HMAC-SHA1vs.
IntelXeon61522.
10GHz,DDR4-24003xIntelCorporationEthernetControllerX710(4x10Gbeportspercard),1xIntelCorpora-tionEthernetControllerX710(2x10Gbeportspercard),PCIex16to2x8PCIebifurcationplugincard,Lewisburg-LB1QuickAssistAcceleratorwithPCIeGen3x24links,IntelOpenSSL-1.
0.
1u+NGINX-1.
9.
6,IntelQAT1.
7.
Upstream.
L.
1.
0.
0-15.
Cores,IO,packetbuffermemory,andprocessingcoresareonasinglesocket.
6coresusedononeSocket,20Coreareused.
Cryptoalgorithm:AES-128-CBC-HMAC-SHA1.
20Inference:Platform:2SIntelXeonPlatinum8180CPU@2.
50GHz(28cores),HTdisabled,turbodisabled,scalinggovernorsetto"performance"viaintel_pstatedriver,384GBDDR4-2666ECCRAM.
CentOSLinuxrelease7.
3.
1611(Core),Linuxkernel3.
10.
0-514.
10.
2.
el7.
x86_64.
SSD:IntelSSDDCS3700Series(800GB,2.
5inSATA6Gb/s,25nm,MLC).
Performancemeasuredwith:Environmentvariables:KMP_AFFINITY='granularity=fne,compact',OMP_NUM_THREADS=56,CPUFreqsetwithcpupowerfrequency-set-d2.
5G-u3.
8G–gperformance.
DeepLearningFrameworks:Caffe:(http://github.
com/intel/caffe/),revisionf96b759f71b2281835f690af267158b82b150b5c.
Inferencemeasuredwith"caffetime--forward_only"command,trainingmeasuredwith"caffetime"command.
For"ConvNet"topologies,dummydatasetwasused.
Forothertopologies,datawasstoredonlocalstorageandcachedinmemorybeforetraining.
Topologyspecsfromhttps://github.
com/intel/caffe/tree/master/models/intel_optimized_models(GoogLeNet,AlexNet,andResNet-50),https://github.
com/intel/caffe/tree/master/models/default_vgg_19(VGG-19),andhttps://github.
com/soumith/convnet-benchmarks/tree/master/caffe/imagenet_winners(ConvNetbenchmarks;fileswereupdatedtousenewerCaffeprototxtformatbutarefunction-allyequivalent).
IntelC++compilerver.
17.
0.
220170213,IntelMKLsmalllibrariesversion2018.
0.
20170425.
Cafferunwith"numactll".
21Training:Platform:2SIntelXeonCPUE5-2697v2@2.
70GHz(12cores),HTenabled,turboenabled,scalinggovernorsetto"performance"viaintel_pstatedriver,256GBDDR3-1600ECCRAM.
CentOSLinuxrelease7.
3.
1611(Core),Linuxkernel3.
10.
0-514.
21.
1.
el7.
x86_64.
SSD:IntelSSD520Series240GB,2.
5inSATA6Gb/s,25nm,MLC.
Performancemeasuredwith:Environmentvariables:KMP_AFFINITY='granularity=fne,compact,1,0',OMP_NUM_THREADS=24,CPUFreqsetwithcpupowerfrequency-set-d2.
7G-u3.
5G–gperformance.
DeepLearningFrameworks:Caffe:(http://github.
com/intel/caffe/),revisionb0ef3236528a2c7d2988f249d347d5fdae831236.
Inferencemeasuredwith"caffetime--forward_only"command,trainingmeasuredwith"caffetime"command.
For"ConvNet"topologies,dummydatasetwasused.
Forothertopologies,datawasstoredonlocalstorageandcachedinmemorybeforetraining.
Topologyspecsfromhttps://github.
com/intel/caffe/tree/master/models/intel_optimized_models(GoogLeNet,AlexNet,andResNet-50),https://github.
com/intel/caffe/tree/master/models/default_vgg_19(VGG-19),andhttps://github.
com/soumith/convnet-benchmarks/tree/master/caffe/imagenet_winners(ConvNetbenchmarks;fleswereupdatedtousenewerCaffeprototxtformatbutarefunctionallyequivalent).
GCC4.
8.
5,IntelMKLsmalllibrariesversion2017.
0.
2.
20170110.
22Upto8.
2xclaimbasedonIntelDistributionforLINPACKBenchmark:1-Node,2xIntelXeonProcessorE5-2690onIntelServerBoardS2600CP2with32GBTotalMemoryonRedHatEnterpriseLinux*6.
0(Santiago)kernelversion2.
6.
32-504.
el6.
x86_64usingIntelDistributionforLINPACKBenchmarkusing56000problemsize.
Score:366.
0GFLOPS/svs.
1-Node,2xIntelXeonPlatinum8180ProcessoronPurley-EP(Lewisburg)with192GBTotalMemoryonUbuntu17.
04usingMKL2017Update2.
DataSource:RequestNumber:2535,Benchmark:IntelDistributionforLINPACKBenchmark,Score:3007.
8GFLOPS/sHigherisbetter.
23Upto1.
63xGainsbasedonGeomeanofWeatherResearchForecasting-Conus12Km,HOMME,LSTCLS-DYNAExplicit,INTESPERMASV16,MILC,GROMACSwater1.
5M_pme,VASP-Si256,NAMDstmv,LAMMPS,AmberGBNucleosome,Binomialoptionpricing,Black-Scholes,MonteCarloEuropeanoptions.
ResultshavebeenestimatedbasedoninternalIntelanalysisandareprovidedforinformationalpurposesonly.
Anydifferenceinsystemhardwareorsoftwaredesignorconfigurationmayaffectactualperformance.
SoftwareandworkloadsusedinperformancetestsmayhavebeenoptimizedforperformanceonlyonIntelmicroprocessors.
Performancetests,suchasSYSmarkandMobileMark,aremeasuredusingspecificcomputersystems,compo-nents,software,operationsandfunctions.
Anychangetoanyofthosefactorsmaycausetheresultstovary.
Youshouldconsultotherinformationandperformanceteststoassistyouinfullyevaluatingyourcontemplatedpurchases,includingtheperformanceofthatproductwhencombinedwithotherproducts.
Formoreinformationgotohttp://www.
intel.
com/performance/datacenter.
24Upto2.
2xclaimbasedonIntelDistributionforLINPACKBenchmark:1-Node,2xIntelXeonProcessorE5-2699v4onGrantley-EP(Wellsburg)with64GBTotalMemoryonRedHatEnterpriseLinux*7.
0kernel3.
10.
0-123usingMP_LINPACK11.
3.
1(ComposerXE2016U1).
DataSource:RequestNumber:1636,Benchmark:IntelOptimizedMPLINPACK,Score:1446.
4GFLOPS/svs.
1-Node,2xIntelXeonPlatinum8180ProcessoronPurley-EP(Lewisburg)with192GBTotalMemoryonUbuntu17.
04usingMKL2017Update2.
DataSource:RequestNumber:2535,Benchmark:IntelDistributionforLINPACKBenchmark,Score:3007.
8GFLOPS/sHigherisbetter.
By2020,thesuccessofhalfoftheworld'sGlobal2000companieswilldependontheirabilitiestocreatedigitallyenhancedproducts,services,andexperiences,1andlargeorganizationsexpecttoseean80percentincreaseintheirdigitalrevenues,2alldrivenbyadvancementsintechnologyandusagemodelstheyenable.
Thisglobaltransformationisrapidlyscalingthedemandsforflexiblecompute,networking,andstorage.
Futureworkloadswillnecessitateinfrastructuresthatcanseamlesslyscaletosupportimmediateresponsivenessandwidelydiverseperformancerequirements.
Theexponentialgrowthofdatagenerationandconsumption,therapidexpansionofcloud-scalecomputing,emerging5Gnetworks,andtheextensionofhigh-performancecomputing(HPC)andartificialintelligence(AI)intonewusagesrequirethattoday'sdatacentersandnetworksurgentlyevolve—orbeleftbehindinahighlycompetitiveenvironment.
Thesedemandsaredrivingthearchitectureofmodernized,future-readydatacentersandnetworksthatcanquicklyflexandscale.
TheIntelXeonScalableplatformprovidesthefoundationforapowerfuldatacenterplatformthatcreatesanevolutionaryleapinagilityandscalability.
Disruptivebydesign,thisinnovativeprocessorsetsanewlevelofplatformconvergenceandcapabilitiesacrosscompute,storage,memory,network,andsecurity.
Enterprisesandcloudandcommunicationsserviceproviderscannowdriveforwardtheirmostambitiousdigitalinitiativeswithafeature-rich,highlyversatileplatform.
EnablingGreaterEfficienciesandLowerTCOAcrossinfrastructures,fromenterprisetotechnicalcomputingapplications,theIntelXeonScalableplatformisdesignedfordatacentermodernizationtodriveoperationalefficienciesthatleadtoimprovedtotalcostofownership(TCO)andhigherproductivityforusers.
SystemsbuiltontheIntelXeonScalableplatformaredesignedtodeliveragileservicesandreduceTCOupto65percent4duetolowersoftwareandOSlicensingfees,andacquisition,maintenance,andinfrastructurecosts.
TheFuture-ForwardPlatformFoundationforAgileDigitalServicesIntelXeonScalablePlatformeNterPriSeit5X1.
5XHYbridcLoudiNfrAStructurecoMMuNicAtioNSerViceProViderStecHNicALcoMPutiNG4-SocketHammerDBOLTPDatabaseVirtualizedApplicationsDPDKL3ForwardingIntelDistributionforLINPACKE7-48704YR.
OLDSYSTEME7-8890v4PRIORGEN.
2.
7X1.
7XE5-26584YR.
OLDSYSTEME5-2658v4PRIORGEN.
4.
2X1.
5XE5-26904YR.
OLDSYSTEME5-2699v4PRIORGEN.
8.
2X2.
2XE5-26904YR.
OLDSYSTEME5-2699v4PRIORGEN.
8180818081806152E5-26904TO5YR.
OLDSYSTEMCONSOLIDATIONIntelXeonProcessor8180(SandyBridge:LaunchMarch2012)4.
2:1ComparingIntelXeonScalableplatformtobothfour-year-oldandlast-generationIntelXeonprocessor-basedsystems,thelatestgenerationIntelCPUdeliversbetterperformanceandcapabilitiesforenterprise,cloud,communications,andHPC.
5,8,9,10,11,12,22,24ReducecostsanddelivermoreservicesperserverwithhigherconsolidationofVMsperserverwiththeIntelXeonScalableplatform.
Virtualizationintheenterprisehasbeensteadilyontheriseforadecade.
Mostorganizationshavetakenadvantageofsomeformofvirtualization,creatinggreaterneedforserverstorunvirtualmachines(VMs)inthedatacenter.
IntelXeonScalableplatformenables4.
2x5moreVMsperservercomparedtoearliergenerations,allowingITtoincreasetheirconsolidationofmoreservicesonlesshardware.
2FoundationalEnhancementsHigherPer-CorePerformance:Upto28cores,deliveringhighperformanceandscalabilityforcompute-intensiveworkloadsacrosscompute,storage,andnetworkusages.
GreaterMemoryBandwidth/Capacity:50percentincreasedmemorybandwidthandcapacity.
Sixmemorychannelsversusfourmemorychannelsofpreviousgenerationformemory-intensiveworkloads.
ExpandedI/O:48lanesofPCIe*3.
0bandwidthandthroughputfordemandingI/O-intensiveworkloads.
IntelUltraPathInterconnect(IntelUPI):UptothreeIntelUPIchannelsincreasescalabilityoftheplatformtoasmanyaseightsockets,aswellasimprovesinter-CPUbandwidthforI/O-intensiveworkloadsversuspreviousgeneration3(withIntelQuickPathInterconnect).
IntelUPIofferstheperfectbalancebetweenimprovedthroughputandenergyefficiency.
IntelAdvancedVectorExtensions512(IntelAVX-512):WithdoubletheflopsperclockcyclecomparedtopreviousgenerationIntelAVX2,6IntelAVX-512boostsperformanceandthroughputforthemostdemandingcomputationaltasksinapplications,suchasmodelingandsimulation,dataanalyticsandmachinelearning,datacompression,visualization,anddigitalcontentcreation.
Securitywithoutcompromise:Near-zeroencryptionoverhead7enableshigherperformanceonallsecuredatatransactions.
InnovativeIntegrationsFirst-timeplatformintegrationsdeliverimprovementsinperformanceandlatencyacrosstheinfrastructure:IntegratedIntelOmni-PathArchitecture(IntelOPA)HostFabricInterface:End-to-endhigh-bandwidth,low-latencyfabricthatoptimizesperformanceandeasesdeploymentofHPCclustersbyeliminatingtheneedforadiscretehostfabricinterfacecard.
IntegratedintheCPUpackage.
Pervasive,BreakthroughPerformanceFromitsnewIntelMeshArchitectureandwidelyexpandedresourcestoitshardware-acceleratingandnewlyintegratedtechnologies,theIntelXeonScalableplatformenablesanewlevelofconsistent,pervasive,andbreakthroughperformance.
IntelXeonProcessorprovidesconsistentperformancefordiverseworkloadsWHATMATTERSINTELXEONSCALABLEPLATFORMADVANTAGESMinimizecomplexitywithcompatiblevirtualizationinfrastructureDeployquickly.
IntelVMscoexistw/otherInteltechnology-basedservers.
MeetstringentcustomerSLAsFastresponsetimes.
MaximizevectorfloatingperformanceandefficiencyHighperformancewithfewerservers.
EnsuredeterministicstorageresponseDeterministicperformance.
Cores,cache,memory,I/Oinasingledie.
DeliverdiverseservicesefficientlyIncredibleefficiencyandhardwareaccelerationfromaplatformwithfeaturesrequiredforapplication,control,packet,andsignalprocessing.
IntheIntelXeonScalableplatformIntelMeshArchitecturewithupto28cores,theLastLevelCache(LLC),sixmemorychannels,and48PCIe*channelsaresharedamongallthecores,givingaccesstolargeresourcesacrosstheentiredieandcreatingdynamicscalabilitywithoutsacrificingperformanceforavarietyofdeployments,suchasvirtualization.
VMsscaleeasilywithfullaccesstoallthesharedonboardresourcesthattheyneed.
EnterpriseandCloudHPCStorageCommunicationsXeonXeonXeonXeoniMCXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeonXeoniMCXeonXeonXeonXeon3xDDR4Channels3xDDR4Channels3x16PCIeGen33xUPILinks3IntegratedIntelQuickAssistTechnology(IntelQAT):Chipset-basedhardwareaccelerationforgrowingcompressionandcryptographicworkloadsforgreaterefficiencywhiledeliveringenhanceddatatransportandprotectionacrossserver,storage,andnetworkinfrastructure.
IntegratedIntelEthernetwithscalableiWARP*RDMA*:Providesuptofour10Gbpshigh-speedEthernetportsforhighdatathroughputandlow-latencyworkloads.
Idealforsoftware-definedstoragesolutions,NVMExpress*overFabricsolutions,andvirtualmachinemigrations.
Integratedinthechipset.
Industry-LeadingStorageSupportStorageinnovationscandrivesignificantimprovementsinefficiencyandperformanceofdata-hungryworkloads.
SupportforIntelOptaneSSDsandIntel3DNANDSolidStateDrives:Deliversindustry-leadingcombinationofhighthroughput,lowlatency,highQoS,andultra-highendurancetobreakthroughdataaccessbottlenecks.
DeploynextgenerationstoragewithconfidencewithIntelVolumeManagementDevice(IntelVMD):EnableshotswappingofNVMeSSDsfromthePCIebuswithoutshuttingdownthesystem,whilestandardizedLEDmanagementhelpsprovidequickeridentificationofSSDstatus.
Thiscommonalitybringsenterprisereliability,availability,andserviceability(RAS)featurestoNVMeSSDs,enablingdeploymentofnext-generationstoragewithconfidence.
IntelIntelligentStorageAccelerationLibrary(IntelISA-L):Optimizesstorageoperations,suchasencryption,forincreasedstorageperformance.
––Upto3.
1xfastercryptographichashingofSHAalgorithmsusingIntelAVX-512.
14––Upto1.
2xalgorithmspeedupforAES-128-GCM.
14––Upto2xfasterReedSolomonErasureCodeperformanceusingIntelAVX-512.
14ComplementaryOfferingsforEvenGreaterPerformance,ScalabilityInteloffersabroadhardwareandsoftwareportfoliothatcomplementsthisnewprocessor.
TheIntelXeonPhiProcessorisanidealfoundationforhighlyparallelapplications,suchasmachinelearningtraining,simulation,andvisualization.
IntelFPGAsofferefficientacceleration15alongwiththeflexibilitythatcomeswithusingprogrammablehardwareforlow-latencyapplications,suchasvirtualswitching,networkservices,dataanalytics,andAI.
ArangeofsoftwaretoolsandlibrariesforgeneralandhighlyparallelcomputinghelpdevelopersoptimizeapplicationsforIntelArchitecture.
EnhancedPlatformTrustDataandplatformreliabilityandprotectionarekeyconcernsforenterprisesdealingwithincreasingconcernsandscrutinyregardingdatasecurityandprivacy.
IntelXeonScalableplatformhelpsbuildhighlytrustedinfrastructureswithplatformdataprotection,resiliency,anduptime.
IncreasedDataProtectionandReliabilityAcrossEveryWorkloadEnhancedIntelRunSureTechnology:NewenhancementsdeliveradvancedReliability,Availability,andServiceability(RAS)andserveruptimeforacompany'smostcriticalworkloads.
Hardware-assistedcapabilities,includingenhancedMCAandrecoveryandadaptivemulti-deviceerrorcorrection,diagnoseandrecoverfrompreviouslyfatalerrors.
And,theyhelpensuredataintegritywithinthememorysubsystem.
IntelKeyProtectionTechnology(IntelKPT)withIntegratedIntelQATandIntelPlatformTrustTechnology(IntelPTT):Deliverhardware-enhancedplatformsecuritybyprovidingefficientkeyanddataprotectionatrest,in-use,andin-flight.
IntelTrustedExecutionTechnology(IntelTXT)withOne-TouchActivation:Enhancedplatformsecurity,whileprovidingsimplifiedandscalabledeploymentforIntelTXT.
Asmoredata-richworkloadsflowthroughthedatacenter,thiscomprehensivesuiteofhardware-enhancedfeaturesbringsbetterdata-andplatform-levelprotectionmechanismsfortrustedservicesinenterpriseandcloudenvironments.
4DynamicandHighlyEfficientServiceDeliveryTheconvergenceofenhancedcompute,memory,network,andstorageperformance,combinedwithsoftwareecosystemoptimizations,makeIntelXeonScalableplatformtheidealplatformforfullyvirtualized,software-defineddatacentersthatdynamicallyself-provisionresources—on-premise,throughthenetwork,andinthepubliccloud—basedonworkloadneeds.
PowerfulToolsandTechnologiesforanAgileDataCenterIntelVT-xnewfeatures:–Modebasedexecutioncontrol(MBE)virtualization:Providesanextralayerofprotectionfrommalwareattacksinavirtualizedenvironmentbyenablinghypervisorstomorereliablyverifyandenforcetheintegrityofkernellevelcode.
–Timestampcounterscaling(TSC)virtualization:ProvidesworkloadoptimizationinhybridcloudenvironmentsbyallowingvirtualmachinestomoveacrossCPUsoperatingatdifferentbasefrequencies.
IntelNodeManager4.
0:HelpsITintelligentlymanageandoptimizepower,cooling,andcomputeresourcesinthedatacenter,maximizingefficiency,whilereducingthechancesofcostlyoverheats.
Strong,CapablePlatformsfortheHybrid-Cloud,Data-FueledEnterpriseEnterprisesarekeentoextractvaluefromtheexplodingdatastreamsbeingpresentedtothemforrapidinsightsthatcanshapetheirbusinessinitiatives.
Traditionalandemergingapplicationsintheenterprise,includingpredictiveanalytics,machinelearning,andHPC,requirenewlevelsofpowerfulcomputecapabilitiesandmassivetiereddatastoragevolumes.
ThemodernizeddatacenterisbeingarchitectedusingaconvergedandholisticapproachthatcanflexiblydelivernewservicesandimproveTCOacrossinfrastructureassetstoday,whileprovidingthemostseamlessandscalableon-ramptoaself-governing,hybriddatacenter.
Yet,organizationsrunningtheirfoundationalbusinessworkloads,suchasOLTPandwebinfrastructure,seektoreduceTCOwithhigherperforminginfrastructures.
TheIntelXeonScalableplatformdeliversnext-generationenterprisecapabilitiestobusinessesthroughafuture-readyplatformthatcanservethehybrid-cloud,data-fueledera,plusithelpsimproveday-to-dayoperationswithupto58percentmorerequests/secondonruntimeworkloads.
16Thisversatileplatformbringsdisruptivelevelsofcomputeperformance,coupledwithmemoryandI/Oadvances,tocompute-hungryandlatency-sensitiveapplications.
CombinedwithinnovativeIntelSSDDataCenterFamilytomanagelargedatavolumesacrossstorage,caching,andmemory,platformsbuiltontheIntelXeonScalableplatformarereadytohandletheintensedemandsofthedataandcloudera.
Withascalableportfolioofpackagingoptionstosuitdiverseworkloadrequirements,theIntelXeonScalableplatformisaperformanceworkhorsedesignedfordeployinghighlyefficient,virtualizedinfrastructuresforcompute,storage,andnetworking.
HighlightsforEnterpriseInnovationUpto28high-performancecores,sixmemorychannels,and48lanesofPCIe3.
0UptothreeIntelUPIchannelsSupportforIntelOptaneSSDsandIntel3DNANDSSDsEnhancedIntelRunSureTechnologyNext-GenerationPlatformforCloud-Optimized,5G-ReadyNetworks,andNext-GenerationVirtualNetworksThecomingeraof5Gwillenableentirelynewecosystemsandclassesofconsumerandenterpriseservicesalongwithmediaapplicationsonwirelessandwirelinenetworks.
Thesedata-rich,innovativeusecases,drivenbythenewInternetofThings(IoT),visualcomputing,andanalytics,representsignificantfutureopportunitiesforcommunicationsserviceproviders(CommSPs)togrowrevenue.
Thetransitionfrompurpose-built,fixedfunctioninfrastructuretoanewgenerationofopennetworksistheessentialfirststeptopreparefora5G-enabledworld.
Software-definednetworkingwithNetworkFunctionsVirtualization(NFV)isenablingnewserviceopportunitiesandoperationsefficienciesforbothcommunicationservicesprovidersandenterprisesalike.
Usingflexible,optimized,industry-standardserversandvirtualized,orchestratednetworkfunctionswillallowfuture-readyinfrastructurestobeabletodeliverinnovativeserviceswithefficiencyandease.
Suchdistributedcommunicationsnetworkscansupportextremelevelsofscalability,agility,programmability,andsecurityacrossanever-growingvolumeandvarietyofnetworkingworkloads—fromthenetworkcoretotheedge.
TheIntelXeonScalableplatformisthebasisfornext-generationplatformstobuildvirtualized,cloud-optimized,5G-readynetworks.
Itoffersanarchitecturethatscalesandadaptswitheasetohandlethedemandsofemergingapplicationsandtheconvergenceofkeyworkloads,suchasapplicationsandservices,controlplaneprocessing,high-performancepacketprocessing,andsignalprocessing.
Thisnewprocessorprovidesafoundationforagilenetworksthatcanoperatewithcloudeconomics,behighlyautomatedandresponsive,andsupportrapidandmoresecuredeliveryofnewandenhancedservicesenabledby5G.
5AccelerateNetworkTrafficwithIntelXeonScalablePlatform,DPDK,andIntelQAT2.
521.
510.
50DPDKL3PacketForwarding(Gbps)PreviousIntelXeonScalablePlatformIPSec(Gbps)SSLWebProxy(Gbps)DPDKoptimizations17:Utilizenewcachehierarchywithlargemid-levelcacheandbalancedbandwidthimprove-mentstoaddressdata,audio,andvideogrowth.
IPSec18:Providedataauthentication,integrity,andconfidentialitywithacceleratedcryptographyandcompression/decompression.
SSLWebProxy19:Acceleratedataencryption/decryptionforsecuretransmissions.
1.
77X2.
5X2.
4XHiGHeriSbetterCombinedwiththeDataPlaneDevelopmentKit(DPDK)andIntelQAT,thisnewprocessorcanboostnetworkperformance,enablingserviceproviderstoaccommodategreatertraffictogrowservices—andrevenue—andhelpingthemtopreparefor5G.
HighlightsforCommSPInnovationHardware-basedaccelerationofencryptionandcompressionusingintegratedIntelQAT.
IntegratedIntelEthernetIntelFPGAsmaximizeversatilityincommunicationsinfrastructureIntelKPTwithIntegratedIntelQATandIntelPTTIntelTXTwithOne-TouchActivationAdditionalResourcesOptimizedforCommSPsTheopensourceDataPlaneDevelopmentKit(DPDK)enablesoptimizedcommunicationsoperationsonIntelArchitecture.
DPDKhasdemonstratedabilitytoscaleperformanceasprocessorcorecountandperformanceincrease;workloads,suchasVectorPacketProcessing(VPP)IPSec,benefitfromthisenhancedperformance.
Additionally,theselibrariesprovidepre-optimizedmechanismstoallownewprocessorcapabilities(suchasIntelAVX-512andmemoryandI/Oenhancements)tobeabletoutilizethenewfunctionalityforimprovedpacketprocessingperformancewithlessdirectdevelopmenteffort.
Inteloffersprograms,suchasIntelNetworkBuildersUniversity,idealfornetworkevolutioninthe5Gera.
Withsolutionguidanceandtrainingfromtheseprograms,CommSPscandrivetheirnetworktransformationinitiativesforwardwithincreasedconfidence.
BreakthroughHPCandHigh-PerformanceDataAnalyticsInnovationToday'sscientificdiscoveriesarefueledbyinnovativealgorithms,newsourcesandvolumesofdata,andadvancesincomputeandstorage.
Benefittingfromexponentiallyexpandingvolumesandvarietyofdata,HPCclustersarealsotheengineforrunningevolvingHigh-PerformanceDataAnalytics(HPDA)workloads,leadingtoincrediblediscoveriesandinsightforbusinessandhumanunderstanding.
Machinelearning,deeplearning,andAIconvergethecapabilitiesofmassivecomputewiththefloodofdatatodrivenext-generationapplications,suchasautonomoussystemsandself-drivingvehicles.
IntelXeonScalableplatformoffersacommonplatformforAIwithhighthroughputforbothinferenceandtraining—upto18xhigherinference20andupto19xhighertraining21throughput,comparedtofour-year-oldsystems.
HPCisnolongerjustthedomainoflargescientificinstitutions.
EnterprisesareincreasinglyconsumingamassivenumberofHPCcomputecycles;someoftheworld'slargestHPCclustersareinprivateoilandgascompanies.
ResearchinpersonalizedmedicineappliesHPCforhighlyfocusedtreatmentplans.
NewHPCinstallationsareengaginginnovative,convergedarchitecturesfornon-traditionalusagesthatcombinesimulation,AI,visualization,andanalyticsinasinglesupercomputer.
6HPCplatforms—fromthesmallestclusterstolargestsupercomputers—demandabalanceacrosscompute,memory,storage,andnetwork.
TheIntelXeonScalableplatformwasdesignedtodeliverandenablesuchbalancewithmassivescalability—totensofthousandsofcores.
FromitsimprovedcorecountandmesharchitecturetonewlyintegratedtechnologiesandsupportforIntelOptanememoryandstoragedevices,theIntelXeonScalableplatformenablestheultimategoalsofHPC—tomaximizeperformanceacrosscompute,memory,storage,andnetworkwithoutinducingbottlenecksatanyintersectionofresources.
TheintegrationofIntelOmni-PathArchitecture,anend-to-endhigh-performancefabric,intotheIntelXeonScalableplatformdeliversbothincreasedperformanceandscalingtodistributed,parallelcomputingclusters.
Nearlinearscalingupto32nodesenablesbuildinglargeHPCsolutionsthatarenotinhibitedbytheinterconnect.
And,theIntelXeonScalableplatform,comparedtopreviousgenerationIntelXeonprocessors,delivers1.
63xbetteroverallperformanceacross13commonHPCworkloadsusedthroughoutthesciencesandfinancialservices.
23TheIntelXeonScalableplatformandIntelOmni-PathArchitecturecanenablenewdiscoveriesandfastersolutionsforhighlyparallelworkloadsinmanydatacenters.
HighlightsforHPCInnovationIntelUltraPathInterconnectIntelAdvancedVectorExtensions512IntegratedIntelOmni-PathArchitectureHostFabricInterfaceSupportforIntelOptaneSSDsandIntel3DNANDSolidStateDrivesIntelXeonGold6148processor3.
02.
52.
01.
51.
00.
50HigherisBetterBroadwell(E5-2697v4)WRF*HOMME*LSTCLS-DYNAExplicit*INTESPERMAS*V16MILC*GROMACS*VASP*NAMD*LAMMPS*AmberGB*BinomialOptionPricingBlack-ScholesMonteCarloEuropeanOptions1.
001.
411.
681.
411.
441.
521.
561.
581.
671.
731.
751.
421.
872.
38ManufacturingEarthSystemsModelsLifeSciencesFSIAcross13commonHPCworkloads,IntelXeonScalableplatformdeliversimprovedperformance.
AdditionalTechnologiesforHPC,HPDA,andAIThebootableIntelXeonPhiProcessor7200Series,withitsManyIntegratedCore(MIC)architecture,isanidealfoundationforhighlyparallelworkloads,suchasmachinelearningtraining,simulation,andvisualization.
Arangeofhigh-productivitysoftwaretools,optimizedlibraries,foundationalbuildingblocks,andflexibleframeworksforgeneralandhighlyparallelcomputinghelpsimplifyworkflowsandassistdeveloperstocreatecodesthatmaximizethecapabilitiesofIAforHPCandAI.
OptimizationsforpopulardeeplearningframeworksforIA,includingNeon,*Caffe,*Theano,*Torch,*andTensorFlow*offerincreasedvalueandperformancefordatascientists.
IntelParallelStudioXE2017includesperformancelibraries,suchasIntelMathKernelLibraryforDeepNeuralNetworks(IntelMKL-DNN)toacceleratedeeplearningframeworksonIA,andIntelDataAnalyticsAccelerationLibrary(IntelDAAL)tospeedbigdataanalytics.
ResourcesOptimizedforHPCTocontinuetoadvancediscoverythroughHPCintotheExascaleera,theIntelModernCodeDeveloperProgramoffersdevelopersanddatascientistseasilyaccessibleonlineandface-to-facecodemodernizationtechnicalsessionsontechniques,suchasvectorization,memoryanddatalayout,multi-threading,andmulti-nodeprogramming.
7IntelXeonScalablePlatformataGlanceIntelXeonScalableplatformoffersarangeofperformance,scalability,andfeatureoptionstomeetawidevarietyofworkloadsinthedatacenter,fromthemostadvanced(IntelXeonPlatinum8100processors)toentry-level(IntelXeonBronze3100processors).
BESTPERFORMANCE,MOSTSCALABLE,BESTBUSINESSAGILITYIntelXeonPlatinum8100ProcessorsBestchoiceformission-critical,real-timeanalytics,machinelearning,andartificialintelligenceworkloadsBestworkload-optimizedperformanceforgeneralpurposecomputeandhybrid-clouddeploymentsBestperformanceforthemostdemandingstorageandnetworkingworkloadsBestmemorybandwidthand2,4,8+socketscalabilityGREATPERFORMANCE,FASTMEMORY,ANDMOREINTERCONNECT/ACCELERATORENGINESIntelXeonGold6100ProcessorsSignificantworkload-optimizedperformanceimprovementsforgeneralpurposecomputeSignificantimprovementsfordemandingstorageandnetworkingworkloadsHighestmemoryspeed,highestmemorycapacity,andenhancedIntelAVX-512Enhanced2-4socketscalabilityandperformanceBETTERPERFORMANCE,ADVANCEDRELIABILITYIntelXeonGold5100ProcessorsImprovedperformanceforcompute-boundworkloadsAffordableAdvancedRASand4-socketscalabilitySuitableforawiderrangeofworkloadsEFFICIENTPERFORMANCEATLOWPOWERIntelXeonSilver4100ProcessorsSolidcomputecapability(Hyper-Threading,TurboBoost)Improvedmemoryspeed,energyefficiencySuitableforamoderaterangeworkloadsENTRY-LEVELPERFORMANCEANDHW-ENHANCEDSECURITYIntelXeonBronze3100ProcessorsAffordable,entry-level2-socketsupportsuitableforlight-rangeworkloadsReliableupgradeversusIntelXeonprocessorE3productfamily8IntelXeonBronzeProcessor(3100Series)IntelXeonSilverProcessor(4100Series)IntelXeonGoldProcessor(5100Series)IntelXeonGoldProcessor(6100Series)IntelXeonPlatinumProcessor(8100Series)PERVASIVEPERFORMANCEANDSECURITYHighestCoreCountSupported8cores12cores14cores22cores28coresHighestSupportedFrequency1.
7GHz(8C/85W)2.
2GHz(10C/85W)3.
6GHz(4C/105W)3.
4GHz(6C/115W)3.
6GHz(4C/105W)NumberofCPUSocketsSupportedUpto2Upto2Upto4Upto4Upto8IntelUltraPathInterconnect(UPI)22233IntelUPISpeed9.
6GT/s9.
6GT/s10.
4GT/s10.
4GT/s10.
4GT/sIntelAdvancedVectorExtensions512(AVX-512)1FMA1FMA1FMA2FMA2FMAMemorySpeedSupport(DDR4)2133MHz2400MHz2400MHz2666MHz2666MHzHighestMemoryCapacitySupportedPerSocket768GB768GB768GB768GB,1.
5TB768GB,1.
5TBIntelOmni-PathArchitecture(DiscretePCIe*card)IntelQuickAssistTechnology(Integratedinchipset)IntelQuickAssistTechnology(DiscretePCIecard)IntelOptaneTechnology-basedSSDs(3DXPoint)IntelSSDDataCenterFamily(3DNAND)PCIe3.
0(48lanes)IntelQuickDataTechnology(CBDMA)Non-TransparentBridge(NTB)IntelTurboBoostTechnology2.
0IntelHyper-ThreadingTechnologyNodeControllerSupportIntelOmni-PathArchitecture(Integrated)HIGHRELIABILITYReliability,Availability,andServiceability(RAS)CapabilityStandardStandardStandardAdvancedAdvancedIntelRunSureTechnologyAGILITY&EFFICIENCYIntelVolumeManagementDevice(IntelVMD)IntelVT-xIntelSpeedShiftTechnologyIntelNodeManager4.
0SECURITYMode-basedExecuteControlIntelKeyProtectionTechnology(KPT)w/IntegratedIntelQATIntelPlatformTrustTechnology(PTT)IntelTXTw/One-TouchActivation(OTA)FamilyFeatures9IntelXeonScalablePlatformSKUNumberingProcessornumbersfortheIntelXeonScalableplatformuseanalphanumericschemebasedonperformance,features,processorgeneration,andanyoptions,followingthebrandanditsclass.
SKULevel8=Platinum6,5=Gold4=Silver3=BronzeProcessorGeneration1=1stGen(Skylake)ProcessorSKU(ex.
20,34…)IntegrationsandOptimizations(ifapplicable)F=FabricT=HighTcase/ExtendedReliabilityMemoryCapacityNoSuffix=768GBpersocketM=1.
5TBpersocketIntelXeonPlatinumprocessorαα##18IntelXeonGoldprocessorαα##16IntelXeonGoldprocessorαα##15IntelXeonSilverprocessorαα##14IntelXeonBronzeprocessorαα##1310IntelXeonScalablePlatformSKUsSKUSOPTIMIZEDFORHIGHESTPER-COREPERFORMANCEClassSKUCoresBaseNon-AVXSpeed(GHz)TDP(W)Platinum8180a282.
5205Platinum8168242.
7205Platinum8158b123.
0150Platinum8156c43.
6105Gold6148202.
4150Gold6154183.
0200Gold6150182.
7165Gold6142a162.
6150Gold6132142.
6140Gold6146b123.
2165Gold6136b123.
0150Gold6126d122.
6125Gold6144b83.
5150Gold6134a,b83.
2130Gold6128d63.
4115Gold5122c43.
6105SKUSOPTIMIZEDFORBALANCEDENERGYEFFICIENTPERFORMANCE/WATTClassSKUCoresBaseNon-AVXSpeed(GHz)TDP(W)Platinum8176a282.
1165Platinum8170a262.
1165Platinum8164262.
0150Platinum8160a242.
1150Platinum8153e162.
0125Gold6152222.
1140Gold6138202.
0125Gold6140a182.
3140Gold6130162.
1125Gold5120142.
2105Gold5118122.
3105Gold5115102.
485Silver4116122.
185Silver4114102.
285Silver411242.
685Silver411082.
185Silver410881.
885Bronze310681.
785Bronze310461.
78511IntelXeonScalablePlatformSKUs,continuedSKUSWITHEXTENDEDLIFE(10-YEARUSE)ANDNEBS-FRIENDLYTHERMALSPECIFICATIONClassSKUCoresBaseNon-AVXSpeed(GHz)TDP(W)Platinum8160T242.
1150Gold6138T202.
0125Gold6130T162.
1125Gold6126Td122.
6125Gold5120T142.
2105Gold5119T141.
985Silver4116T122.
185Silver4114T102.
285Silver4109T82.
070SKUSWITHINTEGRATEDINTELOMNI-PATHARCHITECTUREFABRICClassSKUCoresBaseNon-AVXSpeed(GHz)TDP(W)Platinum8176F282.
1173Platinum8160F242.
1160Gold6148F202.
4160Gold6142F162.
6160Gold6138F202.
0135Gold6130F162.
1135Gold6126Fd122.
6105PRODUCTNAMESKU10Gb/1GbETHERNETPORTSCOMPRESSIONENCRYPTIONRSAIntelQuickAssistTechnologyIntelC621ChipsetLBG-1G0/4N/AN/AN/AIntelC622ChipsetLBG-22/4N/AN/AN/AIntelC624ChipsetLBG-44/4N/AN/AN/AIntelC625ChipsetLBG-E4/420Gb/s20Gb/s20KOps/sIntelC626ChipsetLBG-M4/440Gb/s40Gb/s40KOp/sIntelC627ChipsetLBG-T4/4100Gb/s100Gb/s100KOps/sIntelC628ChipsetLBG-L4/4100Gb/s100Gb/s100KOps/sa"M"SKUavailablesupporting1.
5TB/socketbNon-default24.
75MBcachecNon-default16.
5MBcachedNon-default19.
25MBcacheeCompatiblefor8-socketbuilds121Source:IDC,http://www.
idc.
com/getdoc.
jspcontainerId=prUS41888916.
2Source:GartnerGroup,http://www.
gartner.
com/newsroom/id/3142917.
3Moreinter-CPUbandwidth.
10.
4GT/sonIntelXeonScalableprocessorsvs.
9.
6GT/sonIntelXeonprocessorE5-2600v4productfamily.
4Upto65%lower4-yearTCOestimateexamplebasedonequivalentrackperformanceusingVMwareESXi*virtualizedconsolidationworkloadcomparing20installed2-socketserverswithIntelXeonprocessorE5-2690(formerly"SandyBridge-EP")runningVMwareESXi*6.
0GAusingGuestOSRHEL6.
4comparedatatotalcostof$919,362to5newIntelXeonPlatinum8180(Skylake)runningVMwareESXi6.
0U3GAusingGuestOSRHEL664bitatatotalcostof$320,879includingbasicacquisition.
ServerpricingassumptionsbasedoncurrentOEMretailpublishedpricingforBroadwellbasedservers–subjecttochangebasedonactualpricingofsystemsoffered.
TotalCostofOwnershipEstimatedServerA:2SXeonE5-2690(2.
9GHz,8C,20MB)ServerB:2SXeonPlatinum8180(2.
5GHz,28C,38.
5M)NumberofServers205ServerAcquisition$273,220$150,995OperatingSystem&SoftwareOSLicense$123,100$30,775OSMaintenance$241,360$60,340SoftwareLicense$39,800$9,950SoftwareMaintenance$51,680$12,920ServerMaintenance$134,080$33,520Infrastructure&UtilitiesPowerandCooling$42,522$18,979Rack/FloorSpace$12,400$3,100Networking$1,200$300Total$919,362$320,8795Upto4.
2xmoreVMsbasedonservervirtualizationconsolidationworkload:BasedonIntelinternalestimates1-Node,2xIntelXeonProcessorE5-2690onRomley-EPwith256GBTotalMemoryonVMwareESXi*6.
0GAusingGuestOSRHEL6.
4,glassfsh3.
1.
2.
2,postgresql9.
2.
DataSource:RequestNumber:1718,Benchmark:servervirtualizationconsolidation,Score:377.
6@21VMsvs.
1-Node,2xIntelXeonPlatinum8180ProcessoronWolfPassSKXwith768GBTotalMemoryonVMwareESXi6.
0U3GAusingGuestOSRHEL664bit.
DataSource:RequestNumber:2563,Benchmark:servervirtualizationconsolidation,Score:1580@90VMs.
Higherisbetter.
6AsmeasuredbyIntelXeonProcessorScalableFamilywithIntelAVX-512comparedtoanIntelXeonE5v4withIntelAVX2.
7BigBench,NearZeroencryptionoverhead:BigBenchqueryRuntime/second.
TestingdonebyIntel.
BASELINE:Platform8168,NODES1Mgmt+6Workers,MakeIntelCorporation,ModelS2600WFD,FormFactor2U,ProcessorIntelXeonPlatinum8168,BaseClock2.
70GHz,Corespersocket24,Hyper-ThreadingEnabled,NUMAmodeEnabled,RAM384GBDDR4,RAMType12x32GBDDR4,OSDriveIntelSSDDCS3710Series(800GB,2.
5inSATA6Gb/s,20nm,MLC),DataDrives8x–SeagateEnterprise2.
5HDDST2000NX04032TB,IntelSSDDCP3520Series(2.
0TB),TempDriveDC35202TB,NICIntelX72210GBE–DualPort,HadoopCloudera5.
11,BenchmarkTPCx-BB1.
2,OperatingSystemCentOSLinuxrelease7.
3.
1611(Core);HDFSencryp-tionturnedOFF.
vs.
NEW:Platform8168,NODES1Mgmt+6Workers,MakeIntelCorporation,ModelS2600WFD,FormFactor2U,ProcessorIntelXeonPlatinum8168,BaseClock2.
70GHz,Corespersocket24,Hyper-ThreadingEnabled,NUMAmodeEnabled,RAM384GBDDR4,RAMType12x32GBDDR4,OSDriveIntelSSDDCS3710Series(800GB,2.
5inSATA6Gb/s,20nm,MLC),DataDrives8x–SeagateEnterprise2.
5HDDST2000NX04032TB,IntelSSDDCP3520Series(2.
0TB),TempDriveDC35202TB,NICIntelX72210GBE–DualPort,HadoopCloudera5.
11,BenchmarkTPCx-BB1.
2,OperatingSystemCentOSLinuxrelease7.
3.
1611(Core);HDFSencryptionturnedON.
8Upto5xclaimbasedonOLTPWarehouseworkload:1-Node,4xIntelXeonProcessorE7-4870onEmeraldRidgewith512GBTotalMemoryonOracleLinux*6.
4usingOracle12c*running800warehouses.
DataSource:RequestNumber:56,Benchmark:HammerDB,Score:2.
46322e+006Higherisbettervs.
1-Node,4xIntelXeonPlatinum8180ProcessoronLightningRidgeSKXwith768GBTotalMemoryonRedHatEnterpriseLinux*7.
3usingOracle12.
2.
0.
1(includingdatabaseandgrid)with800warehouses.
DataSource:RequestNumber:2542,Benchmark:HammerDB,Score:1.
2423e+007Higherisbetter.
9Upto1.
52xclaimbasedonHammerDB:1-Node,4xIntelXeonProcessorE7-8890v4onBrickland(Patsburg)with1536GBTotalMemoryonOracleLinux*7.
1usingOracle12.
1.
0.
2.
0(in-cludingdatabaseandgrid)with800warehouses.
DataSource:RequestNumber:2239,Benchmark:HammerDB,Score:8.
17145e+006vs.
1-Node,4xIntelXeonPlatinum8180ProcessoronLightningRidgeSKXwith768GBTotalMemoryonRedHatEnterpriseLinux*7.
3usingOracle12.
2.
0.
1(includingdatabaseandgrid)with800warehouses.
DataSource:RequestNumber:2542,Benchmark:HammerDB,Score:1.
2423e+007Higherisbetter.
10Upto1.
5xmoreVMsbasedonvirtualizationconsolidationworkload:1-Node,2xIntelXeonProcessorE5-2699v4onGrantley-EP(Wellsburg)with512GBTotalMemoryonVMwareESXi*6.
0Update1usingGuestVM'sutilizeRHEL664bitOS.
DataSource:RequestNumber:1637,Benchmark:servervirtualizationworkload,Score:1034@58vs.
1-Node,2xIntelXeonPlatinum8180ProcessoronWolfPassSKXwith768GBTotalMemoryonVMwareESXi6.
0U3GAusingGuestVM'sutilizeRHEL664bitOS.
DataSource:RequestNumber:2563,Benchmark:servervirtualizationworkload,Score:1580@90VMsHigherisbetter.
11Upto2.
7xclaimbasedonDPDKL3PacketForwarding:IntelXeonE5-2650processors2.
00GHz,8GT/sQPI,20MBL3cache,PatsburgChipset(C0stepping),GrizzlyPassPlatform(R2216G-ZBPP),DDR31333MHz,8xdualrank4GB(total32GB),4memorychannelspersocketConfiguration,1DIMMperchannel,6xIntel82599dual-portPCI-ExpressGen2x810GbEthernetNIC,1xIntel82599dual-portGen2x8I/Oexpansionmodule10GbEthernetNIC,BIOSversionS5500.
86B.
01.
00.
0048,Operatingsystem:FedoraCore15,Kernelversion:2.
6.
38.
4,IxNetwork*6.
0.
400.
22,DPDKversion:FD5_1Score:102Gbits/spacketforwardingat256Busingcoresvs.
Gold6152:EstimatesbasedonIntelinternaltestingonIntelXeon61522.
1GHz,2xIntel,FM10420(RRC)GenDualPort100GbEEthernetcontroller(100Gbit/card)2xIntelXXV710PCIExpressGenDualPort25GbEEthernetcontroller(2x25G/card),DPDK17.
02.
Score:281Gbits/spacketforwardingat256Bpacketusingcores,IOandmemoryonasinglesocket.
12Upto1.
7xclaimbasedonDPDKL3PacketForwarding:E5-2658v4:5xIntelXL710-QDA2,DPDK16.
04.
Benchmark:DPDKl3fwdsampleapplicationScore:158Gbits/spacketforwardingat256Bpacketusingcores.
Gold6152:EstimatesbasedonIntelinternaltestingonIntelXeon61522.
1GHz,2xIntel,FM10420(RRC)GenDualPort100GbEEthernetcontroller(100Gbit/card)2xIntelXXV710PCIExpressGenDualPort25GbEEthernetcontroller(2x25G/card),DPDK17.
02.
Score:281Gbits/spacketforwardingat256Bpacketusingcores,IOandmemoryonasinglesocket.
13SystemConfguration:4NodevSAN*Cluster.
PerNodeconfiguration:Supermicro*SuperServer2028U-TN24R4T+DualIntelXeonE5-2687Wv4(12Core@3.
0Ghz),Supermicro*ServerBoard,256GBDDR4RAM,BootDrive,1xIntelSSDDCS3710Series(200GB,2.
5"),vSANIntel3DNANDCluster:VirtualSANSSDs-2DiskGroupscomprisedof2xIntelSSDDCP4600Series(1.
6TB,2.
5"SFF),8xIntelSSDDCP4500Series(4TB,2.
5"SFF),vSANIntel2DNANDCluster:VirtualSANSSDs-2DiskGroupscomprisedof2xIntelSSDDCP3700Series(800GB,2.
5"SFF),8xIntelSSDDCP3500Series(2TB,2.
5"SFF),IntelEthernetServerAdapterX540-DA2.
143.
1x,1.
2x,2xISA-Lconfiguration.
IntelXeonProcessorScalableFamily:Platinum8180Processor,28C,2.
5GHz,H0,NeonCityCRB,12x16GBDDR42666MT/sECCRDIMM,BIOSPLYCRB1.
86B.
0128.
R08.
1703242666.
IntelXeonE5-2600v4SeriesProcessor,E5-2650v4,12C,2.
2GHz,AztecCityCRB,4x8GBDDR42400MT/sECCRDIMM,BIOSGRRFCRB1.
86B.
0276.
R02.
1606020546.
OperatingSystem:RedhatEnterpriseLinux7.
3,Kernel4.
2.
3,ISA-L2.
18,BIOSConfguration,P-States:Disabled,Turbo:Disabled,SpeedStep:Disabled,C-States:Disabled,ENERGY_PERF_BIAS_CFG:PERF.
15AsmeasuredbyIntelXeonProcessorScalableFamilywithIntelFPGAoptimizedworkloadandIntelXeonProcessorScalableFamilywithoutFPGAoptimizedworkload.
16Upto1.
58xclaimbasedonGhost-NodeJSworkload:1-Node,2xIntelXeonProcessorE5-2699v4onWildcatPasswith384GBTotalMemoryonUbuntu16.
04LTSusingNode.
jsversion6.
9.
2,MySQLMariaDBversion15.
1Distrib10.
0.
30.
DataSource:RequestNumber:2687,Benchmark:Ghost-NodeJS,Score:2308Higherisbettervs.
1-Node,2xIntelXeonPlatinum8180ProcessoronWolfPassSKXwith384GBTotalMemoryonUbuntu16.
10usingNode.
jsversion6.
9.
2,MySQLMariaDBversion15.
1Distrib10.
0.
30.
DataSource:RequestNumber:2687,Benchmark:Ghost-NodeJS,Score:3647Higherisbetter.
17Upto1.
77xclaimbasedonDPDKL3Forwarding:E5-2658v4:5xIntelXL710-QDA2,DPDK16.
04.
Benchmark:DPDKl3fwdsampleapplication.
Vs.
Gold6152:EstimatesbasedonIntelinternaltestingonIntelXeon61522.
1GHz,2xIntel,FM10420(RRC)GenDualPort100GbEEthernetcontroller(100Gbit/card)2xIntelXXV710PCIExpressGenDualPort25GbEEthernetcontroller(2x25G/card),DPDK17.
02.
Score:281Gbits/spacketforwardingat256Bpacketusingcores,IOandmemoryonasinglesocket.
18Upto2.
5xclaimbasedonDPDKIPSecSeg-gwbenchmark:IntelXeonE5-2658v4,IntelPCHC612,DDR4-2400Intel895XCCbasedQuickAssistAcceleratorAdapterPCIeGen3x8links,DPDK16.
11IPSec-secgw,1420Bpacket).
IntelDPDK16.
11IPsec-secgwsampleapplication.
Cores,IO,packetbuffermemory,andprocessingcoresareonasinglesocket.
6coresusedononeSocket,Cryptoalgorithm:AES-128-CBC-HMAC-SHA1vs.
IntelXeon61522.
1GHz,3xIntelCorporation,EthernetControllerX710(4x10Gbeportspercard),LewisburgB0QuickAssistAcceleratorwithPCIeGen3x24links,IntelDPDK17.
02IPsec-secgw,IntelQAT1.
7.
Upstream.
L.
1.
0.
0-15,6coresused.
Cores,IO,packetbuffermemory,andprocessingcoresareonasinglesocket.
6coresusedononeSocket,Cryptoalgorithm:AES-128-CBC-HMAC-SHA1.
2017IntelCorporation.
Intel,theIntellogo,IntelXeon,IntelOptaneandIntelXeonPhiaretrademarksofIntelCorporationintheU.
S.
and/orothercountries.
*Othernamesandbrandsmaybeclaimedasthepropertyofothers.
PrintedinUSA0717/SC/HBD/PDFPleaseRecycle336198-00119Upto2.
4xclaimbasedonTLSWebProxyusingNGINX:IntelXeonE5-2658v4,DDR4-2133,IntelPCHC612,Intel895XCCbasedQuickAssistAcceleratorAdapterPCIeGen3x8links,OpenSSL-Async(0.
4.
9-009)+NGINX-1.
6.
2(0.
1.
0-008),QAT1.
6.
L.
2.
6.
0-60.
Cores,IO,packetbuffermemory,andprocessingcoresareonasinglesocket.
6coresusedononeSocket12Coresareused,Cryptoalgorithm:AES-128-CBC-HMAC-SHA1vs.
IntelXeon61522.
10GHz,DDR4-24003xIntelCorporationEthernetControllerX710(4x10Gbeportspercard),1xIntelCorpora-tionEthernetControllerX710(2x10Gbeportspercard),PCIex16to2x8PCIebifurcationplugincard,Lewisburg-LB1QuickAssistAcceleratorwithPCIeGen3x24links,IntelOpenSSL-1.
0.
1u+NGINX-1.
9.
6,IntelQAT1.
7.
Upstream.
L.
1.
0.
0-15.
Cores,IO,packetbuffermemory,andprocessingcoresareonasinglesocket.
6coresusedononeSocket,20Coreareused.
Cryptoalgorithm:AES-128-CBC-HMAC-SHA1.
20Inference:Platform:2SIntelXeonPlatinum8180CPU@2.
50GHz(28cores),HTdisabled,turbodisabled,scalinggovernorsetto"performance"viaintel_pstatedriver,384GBDDR4-2666ECCRAM.
CentOSLinuxrelease7.
3.
1611(Core),Linuxkernel3.
10.
0-514.
10.
2.
el7.
x86_64.
SSD:IntelSSDDCS3700Series(800GB,2.
5inSATA6Gb/s,25nm,MLC).
Performancemeasuredwith:Environmentvariables:KMP_AFFINITY='granularity=fne,compact',OMP_NUM_THREADS=56,CPUFreqsetwithcpupowerfrequency-set-d2.
5G-u3.
8G–gperformance.
DeepLearningFrameworks:Caffe:(http://github.
com/intel/caffe/),revisionf96b759f71b2281835f690af267158b82b150b5c.
Inferencemeasuredwith"caffetime--forward_only"command,trainingmeasuredwith"caffetime"command.
For"ConvNet"topologies,dummydatasetwasused.
Forothertopologies,datawasstoredonlocalstorageandcachedinmemorybeforetraining.
Topologyspecsfromhttps://github.
com/intel/caffe/tree/master/models/intel_optimized_models(GoogLeNet,AlexNet,andResNet-50),https://github.
com/intel/caffe/tree/master/models/default_vgg_19(VGG-19),andhttps://github.
com/soumith/convnet-benchmarks/tree/master/caffe/imagenet_winners(ConvNetbenchmarks;fileswereupdatedtousenewerCaffeprototxtformatbutarefunction-allyequivalent).
IntelC++compilerver.
17.
0.
220170213,IntelMKLsmalllibrariesversion2018.
0.
20170425.
Cafferunwith"numactll".
21Training:Platform:2SIntelXeonCPUE5-2697v2@2.
70GHz(12cores),HTenabled,turboenabled,scalinggovernorsetto"performance"viaintel_pstatedriver,256GBDDR3-1600ECCRAM.
CentOSLinuxrelease7.
3.
1611(Core),Linuxkernel3.
10.
0-514.
21.
1.
el7.
x86_64.
SSD:IntelSSD520Series240GB,2.
5inSATA6Gb/s,25nm,MLC.
Performancemeasuredwith:Environmentvariables:KMP_AFFINITY='granularity=fne,compact,1,0',OMP_NUM_THREADS=24,CPUFreqsetwithcpupowerfrequency-set-d2.
7G-u3.
5G–gperformance.
DeepLearningFrameworks:Caffe:(http://github.
com/intel/caffe/),revisionb0ef3236528a2c7d2988f249d347d5fdae831236.
Inferencemeasuredwith"caffetime--forward_only"command,trainingmeasuredwith"caffetime"command.
For"ConvNet"topologies,dummydatasetwasused.
Forothertopologies,datawasstoredonlocalstorageandcachedinmemorybeforetraining.
Topologyspecsfromhttps://github.
com/intel/caffe/tree/master/models/intel_optimized_models(GoogLeNet,AlexNet,andResNet-50),https://github.
com/intel/caffe/tree/master/models/default_vgg_19(VGG-19),andhttps://github.
com/soumith/convnet-benchmarks/tree/master/caffe/imagenet_winners(ConvNetbenchmarks;fleswereupdatedtousenewerCaffeprototxtformatbutarefunctionallyequivalent).
GCC4.
8.
5,IntelMKLsmalllibrariesversion2017.
0.
2.
20170110.
22Upto8.
2xclaimbasedonIntelDistributionforLINPACKBenchmark:1-Node,2xIntelXeonProcessorE5-2690onIntelServerBoardS2600CP2with32GBTotalMemoryonRedHatEnterpriseLinux*6.
0(Santiago)kernelversion2.
6.
32-504.
el6.
x86_64usingIntelDistributionforLINPACKBenchmarkusing56000problemsize.
Score:366.
0GFLOPS/svs.
1-Node,2xIntelXeonPlatinum8180ProcessoronPurley-EP(Lewisburg)with192GBTotalMemoryonUbuntu17.
04usingMKL2017Update2.
DataSource:RequestNumber:2535,Benchmark:IntelDistributionforLINPACKBenchmark,Score:3007.
8GFLOPS/sHigherisbetter.
23Upto1.
63xGainsbasedonGeomeanofWeatherResearchForecasting-Conus12Km,HOMME,LSTCLS-DYNAExplicit,INTESPERMASV16,MILC,GROMACSwater1.
5M_pme,VASP-Si256,NAMDstmv,LAMMPS,AmberGBNucleosome,Binomialoptionpricing,Black-Scholes,MonteCarloEuropeanoptions.
ResultshavebeenestimatedbasedoninternalIntelanalysisandareprovidedforinformationalpurposesonly.
Anydifferenceinsystemhardwareorsoftwaredesignorconfigurationmayaffectactualperformance.
SoftwareandworkloadsusedinperformancetestsmayhavebeenoptimizedforperformanceonlyonIntelmicroprocessors.
Performancetests,suchasSYSmarkandMobileMark,aremeasuredusingspecificcomputersystems,compo-nents,software,operationsandfunctions.
Anychangetoanyofthosefactorsmaycausetheresultstovary.
Youshouldconsultotherinformationandperformanceteststoassistyouinfullyevaluatingyourcontemplatedpurchases,includingtheperformanceofthatproductwhencombinedwithotherproducts.
Formoreinformationgotohttp://www.
intel.
com/performance/datacenter.
24Upto2.
2xclaimbasedonIntelDistributionforLINPACKBenchmark:1-Node,2xIntelXeonProcessorE5-2699v4onGrantley-EP(Wellsburg)with64GBTotalMemoryonRedHatEnterpriseLinux*7.
0kernel3.
10.
0-123usingMP_LINPACK11.
3.
1(ComposerXE2016U1).
DataSource:RequestNumber:1636,Benchmark:IntelOptimizedMPLINPACK,Score:1446.
4GFLOPS/svs.
1-Node,2xIntelXeonPlatinum8180ProcessoronPurley-EP(Lewisburg)with192GBTotalMemoryonUbuntu17.
04usingMKL2017Update2.
DataSource:RequestNumber:2535,Benchmark:IntelDistributionforLINPACKBenchmark,Score:3007.
8GFLOPS/sHigherisbetter.
- Protectionintelxeon相关文档
- Terminalintelxeon
- Modeintelxeon
- Supportsintelxeon
- SMBusintelxeon
- Curveintelxeon
- rangeintelxeon
RackNerd 2022春节促销提供三款年付套餐 低至年付10.88美元
RackNerd 商家我们应该是比较熟悉的商家，速度一般，但是人家便宜且可选机房也是比较多的，较多集中在美国机房。包括前面的新年元旦促销的时候有提供年付10美元左右的方案，实际上RackNerd商家的营销策略也是如此，每逢节日都有活动，配置简单变化，价格基本差不多，所以我们网友看到没有必要囤货，有需要就选择。RackNerd 商家这次2022农历新年也是有几款年付套餐。低至RackNerd VPS...
江苏云服务器 2H2G 20M 79元/月 大宽带159元/月 高性能挂机宝6元/月 香港CN2 GIA、美国200G防御 CN2 GIA 折后18元/月 御速云
介绍：御速云成立于2021年的国人商家，深圳市御速信息技术有限公司旗下品牌，为您提供安全可靠的弹性计算服务，随着业务需求的变化，您可以实时扩展或缩减计算资源，使用弹性云计算可以极大降低您的软硬件采购成本，简化IT运维工作。主要从事VPS、虚拟主机、CDN等云计算产品业务，适合建站、新手上车的值得选择，拥有华东江苏、华东山东等国内优质云产品；香港三网直连（电信CN2GIA联通移动CN2直连）；美国高...
LOCVPS洛杉矶CN2线路KVM上线,洛杉矶/香港云地/香港邦联7折
LOCVPS发来了新的洛杉矶CN2线路主机上线通知，基于KVM架构，目前可与香港云地、香港邦联机房XEN架构主机一起适用7折优惠码，优惠后最低美国洛杉矶CN2线路KVM架构2GB内存套餐月付38.5元起。LOCPVS是一家成立较早的国人VPS服务商，目前提供洛杉矶MC、洛杉矶C3、和香港邦联、香港沙田电信、香港大埔、日本东京、日本大阪、新加坡、德国和荷兰等机房VPS主机，基于KVM或者XEN架构。...
intelxeon为你推荐
-
京沪高铁上市首秀京沪高铁怎么老是出问题？高铁的核心技术是中国自己的吗？地图应用哪个手机定位软件最好用?同ip网站查询怎么查自己的服务器挂着哪些网站曲妙玲张婉悠香艳版《白蛇传》是电影还是写真集？psbc.comwap.psbc.com网银激活冯媛甑夏如芝是康熙来了的第几期?百花百游百花净斑方效果怎么样？百度关键词分析百度关键字分析是什么意思？www.kk4kk.com猪猪影院www.mlzz.com 最新电影收费吗？51sese.comwww.51xuanh.com这是什么网站是骗人的吗？