Intel 12th-Gen Alder Lake Pricing, Benchmarks and Specs

Intel's 12th-Gen Alder Lake chips bring the company's hybrid architecture, which combines a mix of larger high-performance cores paired with ... Skiptomaincontent Tom'sHardware Tom'sHardware Search RSS UKUSAustraliaCanada Reviews BestPicks RaspberryPi CPUs GPUs Coupons More Laptops SSDs Motherboards Cooling Desktops PCBuilds Monitors RAM PCCases Keyboards Headsets Mice PowerSupplies VRHeadsets AboutTom'sHardware Forums TrendingNvidiaAdaLovelaceAMD5800X3DIntelCorei9-12900KSRTX3090TiAppleM1Ultra Tom'sHardwareissupportedbyitsaudience.Whenyoupurchasethroughlinksonoursite,wemayearnanaffiliatecommission.Learnmore (Imagecredit:Tom'sHardware) AlderLakehasfinallyarrivedtothemarketandwenowhaveourreviewsoftheleading-edgeCorei9-12900K,Corei7-12700K,andCorei5-12600K,andCorei5-12400withalloftheperformancebenchmarksandanalysis.Overall,theCorei5-12600KisnowtheBestCPUforgamingonthemarket,whiletheCorei7-12700Kslotsinasthebesthigh-endprocessorformainstreamplatforms.Intelhasalsoreleasedafulllineupof22additionallower-priceddesktopAlderLakechipsandanewlineupofLaminarcoolers.We'vealsoincludedourfullperformancebenchmarksinthearticlebelow. Intel's12th-GenAlderLakechipsbringthecompany'shybridarchitecture,whichcombinesamixoflargerhigh-performancecorespairedwithsmallerhigh-efficiencycores,todesktopx86PCsforthefirsttime.Intelhasalreadyreleasedthehigher-endmodelsforthedesktopPC,butthecompanyhasaugmentedthosechipswiththemoremainstreammodelsitrecentlylaunched.ItalsoappearsthatIntelwillalsoaddthenewdesigntoitslowest-endchips,too.WealsoexpecttheAlderLakemodelsforlaptopstoalsoarriveinJanuary.  PriceCores|ThreadsP-CoreBase/BoostE-CoreBase/BoostTDP/PBP/MTPDDR4-3200L3CacheCorei9-12900K/KF$589(K)-$564(KF)8P+8E|16Cores/24Threads3.2/5.2GHz2.4/3.9GHz125W/241WDDR4-3200/DDR5-480030MBCorei7-12700K/KF$409(K)-$384(KF)8P+4E|12Cores/20Threads3.6/5.0GHz2.7/3.8GHz125W/190WDDR4-3200/DDR5-480025MBCorei5-12600K/KF$289(K)-$264(KF)6P+4E|10Cores/16Threads3.7/4.9GHz2.8/3.6GHz125W/150WDDR4-3200/DDR5-480016MBCorei5-12400/FEstimated~$185-$1606P+0E|6Cores/12Threads4.4/~2.5GHzn/a65W/117WDDR4-3200/DDR5-480018MB Intel'sinitialAlderLakelineupstartswiththreechipsandtheirgraphics-lessvariants,withtheflagship$589Corei9-12900K,whichIntelbillsasthe'world'sfastestgamingprocessor,'leadingthechargewith16coresand24threads.Intelclaimsthischipprovidesanaverage13%generationaljumpingamingperformance,beatsAMD'sRyzen95950Xflagshipbyupto30%ingaming,andoffersuptotwicetheperformanceofitspredecessorincontentcreationworkloads.Intelevenclaimsuptoan84%generationalgaininfpsinsomegamestreamingscenarios. Wehaveourowntestinginthearticlebelow,andthere'snodoubtthattheAlderLakechipsarenowthefastestgamingchipsontheplanet. Theyalsobringamazingoverclockingheadroomforenthusiasts,offeringupmuchmoreextraperformancethanAMD'sRyzen5000chips.YoucanseethedetailedbreakdownofthatadvantageinourHowtoOverclock12thGenIntelAlderLakeCPUsfeature.YoucanalsoseehowAlderLakefaresagainstAMD'scomparablechipsinthesededicatedhead-to-headarticles:IntelCorei9-12900KvsRyzen95900Xand5950X:AlderLakeandRyzen5000FaceOffIntelCorei5-12600KvsAMDRyzen55600Xand5800XFaceOff:RyzenHasFallenIntelCorei7-12700KvsAMDRyzen95900Xand5800XFaceOff:IntelRisingAlderLakerepresentsamassivestrategicshiftasIntellookstoregaintheuncontestedperformanceleadagainstAMD'sRyzen5000seriesprocessors.AMD'sZen3architecturehastakentheleadinourBestCPUsforgamingandCPUBenchmarkshierarchy,partlyonthestrengthofRyzen'shighercorecounts.That'snottomentionthepressurefromApple'sM1processorsthatfeatureasimilarhybriddesigntoAlderLakeandcomewithexplosivesingle-threadedperformanceimprovementsoftheirown.Intel'sAlderLakebringsdisruptivenewarchitecturesandsupportsfeatureslikePCIe5.0andDDR5thatleapfrogAMDandAppleinconnectivitytechnologyandalsooutstripRyzen'scorecountsinmobiledesigns. Itallstartswithanewwayofthinkingforx86chipsbypairinghigh-performanceandhigh-efficiencycoreswithinasinglechip.Thatwell-traveleddesignphilosophypowersbillionsofArmchips,oftenreferredtoasBig.Little(IntelcallsitsimplementationBig-Bigger),butit'safirstforx86desktopPCs.TheGoldenCovearchitecturepowersAlderLake's'big'high-performancecores,whilethe'little'AtomefficiencycorescomewiththeGracemontarchitecture.Inteletchesthecoresonits'Intel7'process,markingthecompany'sfirsttrulynewnodeforthedesktopsince14nmdebutedsixlongyearsago. Intelisgoingall-in:thecompanywillreunifyitsdesktopandmobilelineswithAlderLake.IntelsaysitwilltuneAlderLakeforhigh-performance,amustfordesktopPCsandhigh-endnotebooks. Asevidencedbyitsopeningsalvoof12th-GenAlderLakemobilelineupthattopsoutat14Coresand5.0GHz,itappearsthecompanyhasdonejustthatformobile,too. (Imagecredit:Tom'sHardware)PotentadversarieschallengeIntelonbothsides.Apple'sArm-poweredM1processorshavesetahighbarforhybriddesigns,outperformingallotherprocessorsintheirclasswiththepromiseofmorepowerfuldesignstocome.Meanwhile,AMD'sRyzen5000chipshavetakentheleadoverIntel'sagingSkylakederivatives.Intel'sRocketLakechipsovertookAMDinsingle-threadedperformance,buttheystilltrailinmulti-coreworkloadsduetoRocketLake'smaximumofeightcores,whileAMDhas16-coremodelsforthemainstreamdesktop.Intelcertainlyneedsacome-from-behinddesigntothoroughlyunseatitscompetitors,swingingthetablesbackinitsfavorliketheConroechipsdidbackin2006whentheCorearchitecturedebutedwitha~40%performanceadvantagethatcementedIntel'sdominanceforadecade.Intel'sRajaKodurihasalreadylikenedthetransitiontoAlderLakewiththedebutofCore,suggestingthatAlderLakecouldindeedbeaConroe-esquemoment.Intel's12th-GenAlderLakeAtaGlanceAlderLakeisavailablenowSixflagshipmodels:$589Corei9-12900K,$409i7-12700K,$289i5-12600K.Allthreearealsoavailableasgraphics-lessKFmodels. Intelalsohasarangeof22lower-pricedmodelsfordesktopPCs.LGA1700socketrequiresnewmotherboardsTheAlderLakeSoCwillspanfromdesktopPCstoultramobiledeviceswithTDPratingsfrom9Wto125W,allbuiltontheIntel7process.ThedesktopPCcomeswithuptoeightPerformance(P)coresandeightEfficient(E)coresforatotalof16coresand24threadsandupto30MBofL3cacheforasinglechip.AlderLakesupportseitherDDR4orDDR5(LP4x/LP5,too).DesktopPCsupportsx16PCIeGen5andx4PCIeGen4,whilemobilesupportsx12PCIeGen4andx16PCIeGen3,Thunderbolt4,andWi-Fi6E.Intel'snewhyper-threadedPerformance(P)core,whichcomeswiththeGoldenCovemicroarchitecturedesignedforlow-latencysingle-threadedperformance,comeswithanaverageof19%moreIPCthantheCypressCovearchitectureinRocketLake.ItalsosupportsAVX-512andAMX(anewAI-focusedmatrix-multiplyISA)fordatacentervariants(botharedisabledonconsumerchips).Intel'snewsingle-threadedEfficiency(E)corecomeswiththeGracemontmicroarchitecturetoimprovemulti-threadedperformanceandprovideexceptionalareaefficiency(smallfootprint)andperformance-per-watt.FoursmallcoresfitinthesameareaasaSkylakecoreanddeliver80%moreperformanceinthreadedwork(atthesamepower).AsingleEcorealsodelivers40%moreperformancethanasingle-threadedSkylakecore(atthesamepower)insingle-threadedwork(caveatsapplytoboth).Intel'sThreadDirectorisahardware-basedtechnologythatassuresthreadsareassignedtoeitherthePorEcoresinanoptimizedmanner.Thisisthesleepertechthatenablesthehybridarchitecture. Fourvariants:-SfordesktopPCs,-Pformobile,-Mforlow-powerdevices,-LAtomreplacement,-Neducational(probablyChromebooks)IntelAlderLakeReleaseDateThehigh-endAlderLakeK-seriesprocessorsforthedesktopPCarenowavailable,butthatonlyencompassesthreeSKUsandtheirgraphics-lesscounterparts.IntelisexpectedtoannouncethefulldesktopPCstack,newchipsetsforlower-endAlderLakeH670,B660andH610motherboards,andthelaptopmodelsatCES2022. IntelAlderLake-SDesktopPCSpecificationsandPricing U.S.PriceCores|ThreadsP-CoreBase/BoostE-CoreBase/BoostTDP/PBP/MTPDDR4-3200L3CacheRyzen95950X$79916P|32threads3.4/4.9GHz-105WDDR4-320064MB(2x32)Corei9-12900K/KF$589(K)-$564(KF)8P+8E|16Cores/24threads3.2/5.2GHz2.4/3.9GHz125W/241WDDR4-3200/DDR5-480030MBRyzen95900X$54912P|24threads3.7/4.8GHz-105WDDR4-320032MB(1x32)Corei9-11900K$5498P|16threads3.5/5.3GHz-125WDDR4-320016MBCorei7-12700K/KF$409(K)-$384(KF)8P+4E|12Cores/20threads3.6/4.9GHz2.7/3.8GHz125W/190WDDR4-3200/DDR5-480025MBCorei7-11700K$4098P|16threads3.6/5.0GHz-125WDDR4-320016MBRyzen75800X$4498P|16threads3.8/4.7GHz-105WDDR4-320032MBCorei5-12600K/KF$289(K)-$264(KF)6P+4E|20Cores/16threads3.7/4.9GHz2.8/3.6GHz125W/150WDDR4-3200/DDR5-480016MBCorei5-11600K$2726P|12threads3.9/4.9GHz-95WDDR4-320012MBRyzen55600X$2996P|12threads3.7/4.6GHz-65WDDR4-320032MB TheAlderLakechipsusetheIntel7process,whichusedtobereferredtoas'10nmEnhancedSuperFin'beforeIntelrecentlyrenameditsprocessnodesduringitslatestprocessandpackagingroadmapupdate.TheGoldenCovecoressupportHyper-Threading,allowingtwothreadstorunonasinglecore,whilethesmallerGracemontcoresaresingle-threaded.Thatmeanssomemodelscouldcomewithseemingly-odddistributionsofcoresandthreads. AllAlderLakechipssupportDDR4-3200orDDR5-4800. Intel's$58916-coreCorei9-12900KcomeswitheightP-coresthatsupporthyper-threading,andeightsingle-threadedE-coresforatotalof24threads.That'sa33%increaseinthreadcountovertheprevious-genCorei9-11900K.TheP-coreshavea3.2GHzbase,andpeakfrequenciesreach5.2GHzwithTurboBoostMax3.0(thisfeatureisonlyonP-cores).Thischipcomeswith125WPBP(base)and241WMTP(peak)powerratings. The12900Khasa100MHzreductioninpeakclockfrequencycomparedtothe11900K,butthatisn'ttoomeaningfulgiventheentirelynewhybridarchitecture—thesechipswillrealizeperformancegainsfromusingdifferentcoretypesfordifferenttasks.Speakingofwhich,theE-coreshavea2.4GHzbaseandstretchupto3.9GHzviathestandardTurboBoost2.0algorithms.Thechipcomesarmedwith30MBofL3cacheand14MBofL2.Image1of2(Imagecredit:Intel)Image2of2(Imagecredit:Intel) At$589,theCorei9-12900Kcomesata$40premiumoveritsprior-gencounterpart,squeezinginbetweenthe$79916-coreRyzen95950Xand$549Ryzen95900X.ThatcouldbeattractiveifIntel'sperformanceclaimspanout(morelater),butitleavesasizeable$185gapbetweentheCorei9andi7familiesthatIntelinadequatelyplugswiththegraphics-less$564Corei9-12900KF.It'slogicaltoexpectafillerproductbetweenCorei7andi9inthefuture(possiblyliketheCorei9-10850K). The$409Corei7-12700Kcomeswiththesame$409traypricingastheprevious-genCorei7-11700KandhaseightP-coresandfourE-cores,foratotalof20threads.TheP-coresrunata3.6/5.0GHzbase/boost,whiletheE-coresweighinat2.7/3.8GHz,allfedby25MBofL3cacheand12MBofL2.Thegraphics-less$384Corei7-12700KFcomeswitha$25pricereduction.(Imagecredit:Tom'sHardware)The12700K's$409pricepointmeansthatIntelhaskepttheCorei7flagshipatitssamepricepoint,whereitlandsbetweenthe$449Ryzen75800Xand$399Ryzen55600X.The12700K/F'sincreasedperformancecouldmakeitamoreattractivepartthanitslacklusterprevious-gencounterpart,thehard-to-recommendCorei7-11700K. TheCorei5-12600K's$289pricepointremainsthesameastheprior-genCorei5-11600K,meaningitlandsrightsmackdabingamercountry,goingtoe-to-toewiththe$299six-coreRyzen55600XandrepresentingthelowestpointofentrytotheAlderLakefamily(atleastfornow).ThischipcomeswithsixthreadedP-coresthatoperateat3.7/4.9GHzandfourE-coresthatrunat2.8/3.6GHz,foratotalof16threads.That'spairedwith20MBofL3and9.5MBofL2cache. AMD'scompetingRyzen55600XcurrentlyleadsourBestCPUforgaminglist,butitfacesastiffchallengefromIntel's12600K. AlltheAlderLakechipssupportbothDDR4andDDR5memory,butthereareseveralcaveatstothelistedDDR5support.Asadefault,DDR5runsinGear2mode,resultinginhigherlatency,andstandardmotherboardsonlysupportDDR5-4800ifthemotherboardhasonlytwophysicalslots.Therefore,atstocksettings,thechipwillonlysupportDDR5-4400onanymotherboardwithfourslots,evenifonlytwoslotsarepopulated. Intelhasdiscardedits'TDP'(ThermalDesignPoint)nomenclature,andnowassignsaProcessorBasePower(PBP)metricinitsplace.ThecompanyalsoaddedasecondaryMaximumTurboPower(MTP)metrictoitsspecsheetstoquantifythehighestpowerlevelduringboostactivity(typicallycalledPL2). AlderLake'snewmemorycontrollerssupportfourdifferentmemorytypes:DDR5-4800andLP5-5200,alongwithDDR4-3200andLP4x-4266.Thissingledesign'sbroadmemorysupportenablesdifferenttypesofmemoryconfigurationsfordifferentuse-cases.ItappearsthatIntelwillsplititsmemorysupportintoDDR4forlower-endZ690motherboards,B-andH-seriesmodels,andmobilesystems,whileDDR5willonlyslotinforthehighest-endZ-seriesmotherboards.ThismakessensegiventheexpectedhighpricingforDDR5memoryintheearlydaysofadoption,thoughit'snotablethatIntelhasn'tconfirmeditsapproachyet.AlderLakealsosupportsuptoPCIe5.0with64GB/softhroughputacrossax16laneconnection.ThedesktopPCchipssupportax16PCIeGen5connectionwithanadditionalx4PCIeGen4connection(itisunclearifthisx4connectionisusedforthechipsetorexposedtotheuser),whilelower-powermodelssupportax12PCIeGen4configpairedwithax16PCIeGen3connection.(Imagecredit:Intel)Thefirstchipsbasedonthedesigncomeinthreedifferentpackages,eachforadifferentsegment:ThedesktopPCchipthatwilldropintonewmotherboardswithanLGA1700CPUsocket(yes,115xcoolerswithconvertersarecompatible),ahigh-performanceBGAType3packageformobileapplications(thisislikelya12-28WUP3package,thoughIntelhasn'tconfirmed),andahigh-densityBGAType4HDIpackageforUltraMobileapplications(likelya7-15WUP4equivalentforultra-thins).WehavealsolearnedthattheIntelAlderLakechipsfordesktopPCswillcomewithtwodievariants,withonebeingasix-corewithp-coresonly,whiletheotherwillcomewiththestandard(upto)eightp-coresandeighte-cores.IntelAlderLakeGamingBenchmarksandPerformanceImage1of8(Imagecredit:Tom'sHardware)Image2of8(Imagecredit:Tom'sHardware)Image3of8(Imagecredit:Tom'sHardware)Image4of8(Imagecredit:Tom'sHardware)Image5of8(Imagecredit:Tom'sHardware)Image6of8(Imagecredit:Tom'sHardware)Image7of8(Imagecredit:Tom'sHardware)Image8of8(Imagecredit:Tom'sHardware)Here'stheoveralltestingfromourreview,whichhasadditionalanalysis.AboveyoucanseegamingresultsinbothWindows11andWindows10.The$54912-coreRyzen95900XisAMD'sfastestgamingchip,butthe$589Corei9-12900Kis8.7%fasterinthecumulative1080pgamingmeasurementacrossourentiretestsuite.Andthat'swithboththeDDR4andDDR5memoryconfigurations,soyouwon'thavetodropseriouscashonaDDR5kittogetthere. Steppingdown$300,the$289Corei5-12600KwithDDR5memoryis~2.7%fasterthanthe$299Ryzen55600X,butthatgapwidensslightlywithlessexpensiveDDR4memory.TheCorei5-12600KalsoeffectivelytiestheRyzen75800X,butfor$161less. RocketLakebuyerswillbeplentydisappointed.TheCorei9-11900Klandedameresixmonthsagoat$539,butthe12900Kis~11%fasteringaming.It'salsomuchfasterinthethreadedworkloadsthatwe'llseelaterinourapplicationtesting.WeseeasimilarstoryunfoldwiththeCorei5-11600Kcomparedtothe12600K,with9%moreperformanceingamingcomingfor$27more. Naturally,movingoverto1440pbringsaGPUbottleneckintotheequation,sotheperformancedeltasbetweenthechipsshrinktremendously.HeretheCorei5-12600Keffectivelytiesthe5600Xand5800X,whilethe12900Kisamere3.6%fasterthantheRyzen95950X. TheCorei9-12900Kslotsinasthefastestgamingchipontheplanet,pricebedamned,buttheCorei7-12700Kdeliversbasicallythesamegamingperformanceforfarlesscash,makingitthego-tochoiceforperformanceaddicts.Formainstreamgamers,theCorei5-12600Koffersthebestblendofpriceandperformancethatyoucanfindonthemarket,barnone.Flippingthroughthe99thpercentilechartsshowslargerdeltas,butwehavetoviewthosewithcautionasWindows11isstillyoungandseemstosufferfrommoreframeratevariabilitythanourWindows10testplatform.Thiscouldresultfromyet-to-be-updatedgamecode,therelativelynewgraphicsdriversforWindows11,orsomeothercombinationoffactorsthatcouldbesmoothedoutinthefuture. Intel'sAlderLakecarvesoutawininWindows11,butlargeperformancedeltasinafewofthegametitlescanheavilyimpactthesetypesofcumulativemeasurements.Forinstance,IntelenjoysasizeableleadinHitman3,butthatgameisspecificallytunedtoleveragetheE-coreseffectivelybyoffloadinglow-prioritytaskslikephysicstothesmallcores.Thatcanbeseenasanadvantagebysomebecausemoregamedevscouldtakethisapproach,orasaflukebyothersthatthinkthistypeofoptimizationwillonlycometoIntel-sponsoredtitles. ThecompetitionbetweenIntelandAMDisabsolutelyclosernow,soit'sbesttomakeaninformeddecisionbasedonthetypesoftitlesyouplayfrequently.Besuretocheckouttheindividualtestsbelow.Image1of11(Imagecredit:Tom'sHardware)Image2of11(Imagecredit:Tom'sHardware)Image3of11(Imagecredit:Tom'sHardware)Image4of11(Imagecredit:Tom'sHardware)Image5of11(Imagecredit:Tom'sHardware)Image6of11(Imagecredit:Tom'sHardware)Image7of11(Imagecredit:Tom'sHardware)Image8of11(Imagecredit:Tom'sHardware)Image9of11(Imagecredit:Tom'sHardware)Image10of11(Imagecredit:Tom'sHardware)Image11of11(Imagecredit:Tom'sHardware)Therewereteethingpains,though.Aswereported,DenuvoDRMfalselyidentifiedIntel'sE-coresasaseparatesystem,andthus91Denuvo-enabledgametitleswouldn'tworkwithAlderLakechips.IntelhasworkedwithDenuvo,andthesoftwaremakerissuedaflurryofgamepatchestofixtheissue.NowthelistofbrokengamesonAlderLakeisdowntojustthree. Additionally,ProcessLassohasnowaddedadjustmentsthatwillhelpaddressanyperformanceissuesinWindows10. IntelAlderLakeApplicationBenchmarksImage1of4(Imagecredit:Tom'sHardware)Image2of4(Imagecredit:Tom'sHardware)Image3of4(Imagecredit:Tom'sHardware)Image4of4(Imagecredit:Tom'sHardware)Wecanboildownproductivityapplicationperformanceintotwobroadcategories:single-andmulti-threaded.Theseslidesshowthegeometricmean(equalweightingtoalltests)ofperformanceinseveralofourmostimportanttestsineachcategory,butbesuretolookattheexpandedresultsbelow. AlderLakedeliversafewstunningwinsinthethreadedworkloadsthatRyzenhasdominatedforsolong,highlightingtheadvantagesofthex86hybridarchitecture.Itisquitesurprisingtoseethe24-threadCorei9-12900KwithDDR5memorytiethe32-threadRyzen95950Xinthemulti-threadedranking,butevenmoresurprisingtoseeittakea3%leadwithDDR4.That'sprettyimpressiveinlightofthe5950X's$800pricetag. TheCorei5-12600Kisequallyimpressiveinitspricerangeasitis38%fasterinthreadedworkthanthecomparably-priced5600X,and7%fasterthanthe5800Xthatcosts$161more. ThedeltasinfavorofAlderLakeareevenmoreconvincinginthesingle-threadedmetric,butyoushouldn'tputundueimportanceonthismetricbecauseitiscomprisedofafewveryspecificworkloads.Youcanseeabroaderspateoflightly-threadedworkloadsbelow. Needlesstosay,AlderLakedominatesthosetypesofworkloads.Checkoutourreviewfortheblow-by-blowonallofthedifferentbenchmarks. IntelAlderLakePowerConsumption,Efficiency,andThermalsImage1of11(Imagecredit:Tom'sHardware`)Image2of11(Imagecredit:Tom'sHardware)Image3of11(Imagecredit:Tom'sHardware)Image4of11(Imagecredit:Tom'sHardware)Image5of11(Imagecredit:Tom'sHardware)Image6of11(Imagecredit:Tom'sHardware)Image7of11(Imagecredit:Tom'sHardware)Image8of11(Imagecredit:Tom'sHardware)Image9of11(Imagecredit:Tom'sHardware)Image10of11(Imagecredit:Tom'sHardware)Image11of11(Imagecredit:Tom'sHardware)Yes,AlderLakestillsucksmorepowerthanAMD'sRyzen5000serieschips,butthearrivaloftheIntel7processdoesmarkabigimprovement.Aswecansee,theAlderLakechipsconsumefarlesspowerthantheRocketLakechips—wemeasuredapeakof238Wwiththe12900K,whiletheprevious-gen11900Kdrewnearly100WmoreduringthesameBlenderworkload. Overall,Intelhasreduceditspowerconsumptionfrommeme-worthytoanacceptablelevel.Besides,AlderLakeismuchfasterthanitspredecessor,earningitsomeleeway. Forinstance,asyoucanseeinourrenders-per-daymeasurements,theCorei9-12900Kand12600Karebothtwiceasefficientastheirpredecessors,whichiscommendable.Thislowerpowerconsumptionresultsinlowercoolingrequirements,too. Image1of4(Imagecredit:Tom'sHardware)Image2of4(Imagecredit:Tom'sHardware)Image3of4(Imagecredit:Tom'sHardware)Image4of4(Imagecredit:Tom'sHardware)HerewetakeaslightlydifferentlookatpowerconsumptionbycalculatingthecumulativeamountofenergyrequiredtoperformBlenderandx264andx265HandBrakeworkloads,respectively.Weplotthis'taskenergy'valueinKilojoulesontheleftsideofthechart. Theseworkloadsarecomprisedofafixedamountofwork,sowecanplotthetaskenergyagainstthetimerequiredtofinishthejob(bottomaxis),thusgeneratingareallyusefulpowerchart. Bearinmindthatfastercomputetimes,andlowertaskenergyrequirements,areideal.Thatmeansprocessorsthatfalltheclosesttothebottomleftcornerofthechartarebest. Asyoucansee,Intel'schipshavedescendedfromtheundesirableupperrightofthechartdowntothelowerlefthand,nearlymatchingAMD'schipsinpowerconsumptionwhileactuallybeingfaster.That'sanoutstandingimprovementaftersixyearsofpower-guzzling14nmchips. IntelLaminarStockCoolers(Imagecredit:Intel)We’vealreadyseensignsofIntel’snewLaminarcoolersinthewild,butnowwehavethefulldetails.ThesecoolersaredesignedtoaddresstwomajordeficiencieswithIntel’sstockcoolers:Thermaldissipationlimitationsandaesthetics.AMD’sstockcoolershavelongbeatIntelinbothofthesedepartments,sothisisasorely-neededupgrade.Image1of3(Imagecredit:Intel)Image2of3(Imagecredit:Intel)Image3of3(Imagecredit:Intel)TheLaminarcoolersdropontotheLGA1700socketandcomeinthreevariants.TheLaminarRH1hascontrollableaRGBlightingandfeaturesalargecopperslugtoimprovethermaldissipation.Intelclaimsnear-silentoperationat2.6BA.ThiscoolershipswiththelockedCorei9model. TheLaminarRM1dropsinfortheCorei7,i5,andi3non-KmodelsandcomeswithoutRGBlighting(itdoeshaveadecorativeblueringliningthefinstack).Intelratesthiscoolerfor‘quietperformance’at3.9BA.TheLaminarRS1roundsoutthenewfamilyofcoolers,butit’sthesmallestofthebunchandwillonlyshipwiththelower-powerPentiumandCeleronchips.Thiscoolerdoesn’thavealistednoiserating,soyoushouldexpectittobefairlynoisy. Intelwillalsomakethesecoolersavailableseparatefromthechips,butyouhavetobuytheminlotsof1000.WhilethatputsitoutofreachformostofustopurchasedirectlyfromIntel,thismeansthatwewilleventuallyseethesecoolersfilterouttoretailoutletsviathird-partysellers.We’reworkingtosourcethesecoolersforin-depthtesting;staytuned.IntelAlderLake-PandAlderLake-MMobileProcessorSpecificationsImage1of4(Imagecredit:Intel)Image2of4(Imagecredit:Intel)Image3of4(Imagecredit:Intel)Image4of4(Imagecredit:Intel)TheAlderLakemobilechipssupportDDR5-4800andLPDDR5-5300,whileAppleM1-basedchipssupportLPDDR-6400andAMDhasLPDDR5-5500.You’llonlyfindtheDDR5interfacesonthehigher-endAlderlakeproducts,andsupportforstandardDDR5enablesend-userconfigurability.IntelalsohasoptionsforDDR4-3200andLRPDDR4x-4267interfaces,whichyou’llfindDDR4optionsonthemorevalue-orientedplatforms.Alltold,theAlderLakeplatformsupportsfourdifferenttypesofmemorytechnologies,whichIntelsaysisafirstforamobileplatform. IntelalsosupportsthePCIe5.0interfacefordesktopchipsbutcitedbothalackofend-userdevicesandalackoftimetoqualifythefasterinterfaceforthemobilemodels.Assuch,alloftheAlderLakemobilechipsonlysupportPCIe4.0. Intel’sH-serieslineupspanstheCorei9,i7,andi5familieswithatotalofeight45WSKUsdesignedforhigh-endlaptops.Thesechipscanconsumeupto115Wforshortperiodsduringintensework(durationvariesbasedonthelaptop’scoolingcapabilities),whichwillinvariablyimpactpowerconsumption.Intelhasn’tprovidedbatterylifeprojectionsyet(itsaysitwillinthefuture)butisfrankthattheAlderLakeseriesistunedforperformance.Assuch,weshouldexpecthighergen-on-genperformancebutsimilarbatterylife.Intel’sperformancebenchmarksbelowareimpressive,butsimilarbatterylifewon’tdomuchtoaddresstheelephantintheroom:TheAppleM1-basedlaptops. Asyou’llnotice,theAlderLakechipstopoutatsixperformancecoreswhileIntel’sprevious-genmobileprocessorshaduptotencores.IntelaugmentsthesixP-coreswithuptoeightE-cores,whichallowstheprocessortoprovidesimilarorbetterperformanceinmulti-threadedworkloadsthantheprevious-gentencoremodels.Naturally,theP-cores’higherIPCalsodeliversmoreperformanceinsingle-threadedwork. TheCorei9-12900HKisthehalochipwithsixP-coresandeightE-cores,withtheformeroperatingatupto5.0GHzwhilethelatterbooststo3.8GHz.TheP-coresfeaturehyperthreading,whichmeanseachhosttwothreads,whiletheE-coresaresingle-threaded.Asaresult,theCorei9-12900HKhas14coresand20threads.ItalsocomesarmedwiththeIntelIrisXegraphicsenginewith96EUs.(someofthei7andi5modelshavefewerEUs). IntelalsounveiledthespecificationsforitsP-andU-serieschipsbutdidn’tprovideanyadditionalcommentary.However,wecanseethattheP-serieschipshavea28WTDPandjumptoa64WTDPunderload.ThesechipscomewitheitherfourorsixP-corespairedwitheightE-cores.Thesechipsaredesignedforperformancethin-and-lightlaptops. TheU-seriesisdesignedforthin-and-lightsandsupportsdefault9or15WTDPs,butcanjumpto29Wor55W,whichisquitethedynamicrange.TheU-seriesmodelscomewithtwoP-coresandeitherfouroreightE-cores.TheU-seriesdropsdowntofive-coresix-threadPentiumandCeleronmodels,anoddarrangementthat’sreminiscentofLakefield,Intel’sfirsthybridchip.TheseU-andP-serieschipswillcometomarketlaterinQ1,andIntelsaysitwillsharemoredetailsthen.IntelAlderLake600-SeriesMotherboards,LGA1700Socket,DDR5andPCIe5.0Image1of3(Imagecredit:Intel)Image2of3(Imagecredit:Intel)Image3of3(Imagecredit:Intel)Intel'sincessantmotherboardupgrades,whichrequirenewsocketsorrestrictsupportwithinexistingsockets,haveearnedthecompanyplentyofcriticismfromtheenthusiastcommunity—especiallygivenAMD'slonglineofAM4-compatibleprocessors.ThattrendwillcontinuewithanewrequirementforLGA1700socketsandthe600-serieschipsetforAlderLake(we'vealreadyseenplentyoflistingsandpicturesofZ690motherboards).Still,ifrumorsholdtrue, Intelwillsticktothenewsocketforatleastthenextgenerationofprocessors(7nmMeteorLake)andpossiblyforanadditionalgenerationbeyondthat,rivalingAMD'sAM4longevity. Intelalsointroducedits14nmZ690chipsetfortheinitialrelease,andyoucanreadaboutthechipsetandsomeofthefirst60+motherboardsinourZ690motherboardrounduphere.There'sawideselectionofDDR5motherboardsspreadamongthevariousmotherboardmakers'high-andlower-endZ690families.However,DDR4modelsappeartobeconfinedtothelower-endZ690boards(unlikepreviousgenerations,nomotherboardsupportsbothDDR4andDDR5).WeexpectpricingforDDR5tobesubstantiallyhigherthanDDR4,currentlyprojectedtobea50to60%markup,forsometime.Z690motherboardsaremoreexpensivethanlower-endalternatives,butweexpectthelower-endAlderLakeH670,B660andH610motherboardstobeannouncedatCES2022. AlderLakesupportsupto16lanesofPCIe5.0(technicallyforstorageandgraphicsonly,nonetworkingdevices)andanadditionalfourlanesofPCIe4.0fromthechipforM.2storage.Intelhasalsoadded12lanesofPCIe4.0thathangoffthechipset,anicestepupfromtheZ590chipset'sPCIe3.0support.Thatprovidesatotalof28PCIeplatformlanesforAlderLakesystems.IntelalsodoubledthethroughputoftheDMIconnectionbetweenthechipandchipsetfromanx8DMI3.0pipe,whichclocksinat7.88GB/s,toanx8DMI4.0connectionthatdelivers15.66GB/s.IntelalsoaddedsupportfortheVolumeManagementDevicefeaturethatenablesPCIeSSDmanagementandtheabilitytocreatebootableRAIDconfigurations.BecausetheLGA1700socketisbiggerthanthecurrentsocketsusedinLGA1151/LGA1200motherboards,existingcoolersmaybeincompatible,butcoolerconversionkits,whichmostcoolermakerswillprovideforfreeforexistingcustomers,canaccommodatethelargersocket. (CoolersthatsupportbothLGA11xxandLGA2066alreadyexist,soanin-betweenoptionisn'ttoodifficult.)Thelargersocketisneededtoaccommodate500morepinsthantheLGA1200socket.Thosepinsareneededtosupportnewerinterfaces,likePCIe5.0andDDR5,amongotherpurposes,likepowerdelivery. IntelhasalsolistedAlderLake-SBGAsupportdocumentation,indicatingthatsoldered-downmodelswillalsocometomarket. PCIe5.0andDDR5supportgiveIntelaconnectivityadvantageovercompetingchips,buttherearealotofconsiderationsinvolvedwiththesebigtechnologytransitions.AswesawwiththemovefromPCIe3.0to4.0,astepuptoafasterPCIeinterfacerequiresthickermotherboards(morelayers)toaccommodatewiderlanespacing,morerobustmaterials,andretimersduetostrictertracelengthrequirements.Allofthesefactorsconspiretoincreasecost. (Imagecredit:Tom'sHardware)WecanexpectthosesamePCIe4.0requirementstobecomemorearduousformotherboardswithaPCIe5.0interface,particularlybecausetheywillrequireretimersforevenshorterlanelengthsandeventhickermotherboards.ThatmeanswecouldseeyetanotherjumpinmotherboardpricingoverwhattheindustryalreadyabsorbedwiththemovetoPCIe4.0.Additionally,PCIe5.0alsoconsumesmorepower,whichwillpresentchallengesinmobileformfactors. IntelhasannouncedthatAlderLakewillsupportDDR5memory,butthatwillcausepricingpressure.Notably,everytransitiontoanewermemoryinterfacehasresultedinhigherup-frontDIMMpricing,whichisconcerningintheprice-sensitivedesktopPCmarket. DDR4forexamplefirstcametotheHEDTsegmentonIntel'sX99platformin2014,andpricingatthetimewasmorethandoublethecostofDDR3.SkylakebroughtDDR4tothemainstreamsegmentin2015,butitstillcarrieda25-50%pricepremium.Currentsignspointtoa50%to60%premiumforDDR5memory. DDR5isintheopeningstages;somevendors,likeAdata,TeamGroup,andMicron,havealreadybegunshippingmodules.TheinauguralmodulesareexpectedtorunintheDDR5-4800toDDR5-6400range.TheJEDECspectopsoutatDDR5-8400,butaswithDDR4,itwilltakesometimebeforeweseethosepeakspeeds. Wehave,however,seensignsthatonlythehigher-endAlderLakedesktopPCplatforms,likeZ-seriesmotherboard,willsupportDDR5,whilelower-endboardswilluseDDR4forafriendlierpriceofentry. (Imagecredit:Intel)Intel’spricingmodelwiththeAlderLakechipsisbrutal—basedontheirperformance,everymodelundercutscompetingRyzen5000chipsbylargemargins.ThatmeanstheAlderLakechipsareagreatvalue,butthehighpricingassociatedwiththeexpensiveZ690motherboardshasbeenapainpointthat,insomecases,allowsAMDtoremaincompetitiveonoverallplatformcosts.ThosehighpricestendtostemprimarilyfromsupportforDDR5andthePCIe5.0interface.TheDDR5shortages,andresultingpricescalping,certainlyhaven’thelpedmatters,either.Unfortunately,therearepreciousfewhigh-endDDR4Z690optionsavailable. IntelplanstoaddressthehighmotherboardpricingissuewiththeH670,B600,andH610chipsets.We’retoldthatweshouldexpectthenewmotherboardswiththesechipsetstopredominantlysupportlower-costDDR4memory. Notably,memoryoverclockingisallowedwithallAlderLakeprocessors,includinglockedmodels,ontheB660andH470platforms.Wehaveamuchdeeperbreakdownofthenewchipsetfeaturesandaroundupofthemotherboardsweknowabouthere,soheadthereformoredetail. IntelAlderLake:TDPvsPBPandMTP(Imagecredit:Intel)IntelhaslonglistedtheTDPofaprocessorasitsguaranteedratingatbasefrequencies,alsoknownasPL1.However,thechipcanalsoopportunistically(meaningthisisn'tguaranteed)boosttohigherfrequenciesandthusconsumefarmorepower,butonlyifitissafelywithincertainpower,temperature,andcurrentlimits.ThisiscalledthePL2powerstate,andIntelhasn'tincludedthismetriconitsstandardspecsheets. NowIntelhasredefineditspowernomenclaturetohavea'ProcessorBoostPower'(PBP)valuerepresentingtheguaranteedbaseperformancelevel(PL1).ThisreplacesTDP. Intelwillalsonowlista'MaximumTurboPower'(MTP)specificationthatquantifiesthepowerconsumptionduringTurboBoost,alsoknownasPL2.Thatmeansyou'llnolongerseeaTDPratingonthespecsheet. Intel'sprocessorshavea'TauDuration'settingthatdictateshowlongtheprocessorcanstayintheboostedMTPstate(PL2)beforeitdropsbackdowntothePBPstate(PL1–basepower).Intelspecifiedthisdurationas58secondsforRocketLakechips,butthisisonlyaguideline.Motherboardvendorsarefreetoalterthisvaluetoanylengthoftimeiftheirmotherboardcanhandlethepowerdeliveryrequiredtosustaintheboost.Asshowninthegraphicabove,mostmotherboardvendorschangetheTausettingtoinfinitetostaywithinboostforaninfiniteamountoftime.GiventhatIntel'sTausettingsareonlyrecommendations,thechipremainsinsideofthewarrantyregardlessofboostduration.AsaninfiniteTaueisacommonpracticeonnearlyeveryenthusiastmotherboard,IntelwillnowsettheTautoadefaultof'infinite'forallofitsK-series(overclockable)models,butretainthesame58-seconddurationforitslockedchips.ThismeansthattheCorei9-12900K'sMTP(PL2)isnowthesameasitsPBP(PL1).Inotherwords,thechipwillalwaysoperateat241W.Additionally,thenewdesktopPCmotherboardsforAlderLakechipswillheraldthearrivalofmainstreamATX12VOmotherboardsthatleverageanewlower-powerPSUspecification.BothsystemswithsupportforstandardpowersuppliesandtheATX12VOspecareplanned,butIntelisonafull-courtpresstopushtheadoptionofthenewstandard. However,despiteIntel'sfondnessforthestandard,wehaveyettoseeasignificantnumberof600-seriesboardsthatsupportATX12VO.Almostallboardsworkwithstandardpowersupplies. Intel12th-GenAlderLakeXeLPIntegratedGraphicsThemediaengine,inthiscasethesame Gen12XeLParchitecturefoundinTigerLakebutportedtotheIntel7process,comesintwovariants:onewith32EUs(GT1)fordesktopPCs,andanotherGT2variantwith96EUsforthemobilevariants.ThedesktopPCmodelscomewith33%moreEUsthanthecurrentdesktopchipswithGen9.5UHD630Graphics,butthat'safarcryfromthe96EUsfoundin11thGenTigerLake.Butthisisonthedesktop,wheremostusersthatcareaboutgraphicsperformancewillsimplyuseadedicatedGPU.IntelsaystheXeLPenginesupports1080pgameplayandfeaturesa12-bitend-to-endvideopipeline.ThedesktopPCmodelsdon'thaveThunderbolt4connectivityoranimageprocessingunit(IPU),withthosefeaturesbeingusedonlyformobilevariants. TheUHD770engineclocksinat(upto)1550,1500,and1450MHzforthe12900K,12700K,and12600K,respectively.We'vealsoseenAlderLake-Pbenchmarks(themobilechips)withtheGT2configuration,with96EUs(768shaders).TheearlyXeLPiGPUsilicononthe-Pmodelrunsat1.15GHz,butaswithallengineeringsamples,thatcouldchangewithshippingmodels. AlderLake'sintegratedGPUssupportuptofivedisplayoutputs(eDP,dualHDMI,andDualDP++),andsupportthesameencoding/decodingfeaturesasbothRocketLakeandTigerLake,includingAV18-bitand10-bitdecode,12-bitVP9,and12-bitHEVC.  IntelAlderLakeCPUArchitectureIntelpioneeredthex86hybridarchitecturewithitsLakefieldchips,withthoseinauguralmodelscomingwithoneSunnyCovecorepairedwithfourAtomTremontcores.ComparedtoLakefield,boththehigh-andlow-performanceAlderLake-Scorestakeastepforwardtonewermicroarchitectures.AlderLake-Sactuallyjumpsforwardtwo'Cove'generationscomparedtothe'big'SunnyCovecoresfoundinLakefield.ThebigGoldenCovecorescomewithincreasedsingle-threadedperformance,AIperformance,Networkand5Gperformance,andimprovedsecurityfeaturescomparedtotheWillowCovecoresthatdebutedwithTigerLake. Image1of2(Imagecredit:Intel)Image2of2(Imagecredit:Intel)AlderLake'ssmallerGracemontcoresjumpforwardasingleAtomgenerationandofferthebenefitofbeingmorepowerandareaefficient(perf/mm^2)thanthelargerGoldenCovecores.Gracemontalsocomeswithincreasedvectorperformance,anodtoanobviousadditionofsomelevelofAVXsupport(likelyAVX2).Intelalsolistsimprovedsingle-threadedperformancefortheGracemontcores.Image1of3(Imagecredit:Intel)Image2of3(Imagecredit:Intel)Image3of3(Imagecredit:Intel)Youcanreadourdeep-divecoverageoftheGoldenCovePerformanceCorearchitecturehere.Insummary,theGoldenCovemicroarchitectureisdesignedforlow-latencysingle-threadedperformanceandcomeswithanaverageof19%moreIPCthantheCypressCovearchitectureinRocketLake.ItalsosupportsAVX-512andAMX(anewAI-focusedmatrix-multiplyISA)fordatacentervariants(botharedisabledonconsumerchips).Image1of2(Imagecredit:Intel)Image2of2(Imagecredit:Intel)YoucanalsoreadourdeepdivecoverageoftheGracemontEfficiencyCorearchitecturehere.Insummary,thesingle-threadedEfficiency(E)core,whichcomeswiththeGracemontmicroarchitecture,isdesignedtoimprovemulti-threadedperformanceandprovideexceptionalareaefficiency(smallfootprint)andperformance-per-watt.FourofthesesmallcoresfitinthesameareaasaSkylakecoreanddeliver80%moreperformanceinthreadedwork(atthesamepower).AsingleEcorealsodelivers40%moreperformancethanasingle-threadedSkylakecore(atthesamepower)insingle-threadedwork(caveatsapplytoboth).Image1of3(Imagecredit:Intel)Image2of3(Imagecredit:Intel)Image3of3(Imagecredit:Intel)LakefieldservedasaprovinggroundnotonlyforIntel's3DFoverospackagingtechbutalsoforthesoftwareandoperatingsystemecosystem.AtitsArchitectureDay2020,InteloutlinedtheperformancegainsabovefortheLakefieldchipstohighlightthepromiseofhybriddesigns.Still,theresultscomewithanimportantcaveat: Thesetypesofperformanceimprovementsareonlyavailablethroughbothhardwareandoperatingsystemoptimizations. Let'slookatIntel'ssolutiontothatproblem. IntelThreadDirectorIntelunveiledtheanswertothesoftwarechallengeatitsArchitectureDay2021—thenewThreadDirector.DuetoAlder'suseofbothfasterandslowercoresthatareoptimizedfordifferentvoltage/frequencyprofiles,unlockingthemaximumperformanceandefficiencyrequirestheoperatingsystemandapplicationstohaveanawarenessofthechiptopologytoensureworkloads(threads)landinthecorrectcorebasedonthetypeofapplication. (Imagecredit:Intel)Thecurrentthreadschedulingsystemsarebasedentirelyonstaticrules(priority,foreground,background)andtendtobeinefficientandcreatesoftwareprogrammingoverhead.That'swhereIntel'sThreadDirectortechnologycomesin.Thishardware-basedtechnologyprovidesenhancedtelemetrydatatoWindows11toassurethatthreadsarescheduledtoeitherthePorEcoresinanoptimizedandintelligentmanner,butinawaythat'stransparenttosoftware.ThistechnologyworksbyfeedingtheWindows11operatingsystemwithlow-leveltelemetrydatacollectedfromwithintheprocessoritself,thusinformingthescheduleraboutthestateofthecore,beitpower,thermalorotherwise.(Aswecoveredhere,IntelhasintegratedanewpowermicrocontrollerineachGracemontcore,afirst,thatcollectssimilardataontheorderofmicrosecondsinsteadofmilliseconds,soitmightbepartofthenewtelemetrysystem.)(Imagecredit:Intel)Additionally,ThreadDirectorcanalsodetecttheinstructionmix(scalar/vector)usedinanygiventhreadatananosecondgranularity,andthencommunicatewiththeWindows11schedulertosteerthethreadtothecorrectexecutioncore,bethatahigh-performanceP-CoreoranefficientE-Core.Typically,vector/AIworkloadswillbeprioritizedtoperformancecoreswhilescalarinstructionsandbackgroundtasksaremovedtoefficiencycores.However,thesystemisdynamic,sothreadplacementdecisionscanvarybasedonthedynamicmixofconditionsandworkloadspresentontheprocessoratanygiventime.Additionally,threadscangothroughvariousphasesandinstructionmixesovertheirlifetime,sotheschedulerconstantlyre-adjustsbasedonthereal-timetelemetrydata.Thisishelpfulwhenthenumberofthreadsdesignatedfor'performance'outnumbertheavailablecores,forinstance.Inthatcase,lessdemanding'performance'threads,suchasaprograminaspinloop,canbemovedofftotheefficiencycoreswhilemoredeservingworkloadsareassignedtotheperformancecore.Previously,theoperatingsystemdidn'thaveaccesstothistypeoftelemetrydatatoinformschedulingdecisions,insteadusingsimpledatalikewhethertheprocesswasaforegroundorbackgroundtask.Thisenhancedsystemallowstheoperatingsystemandprocessortoworkintandemtoassurecorrectschedulinginreal-time,thusavoidingcostlysoftwarere-coding.ThisisapromisingsignthatexistingcodewillrunwellontheAlderLakeprocessors.Programmerscanaccessmoregranularcontrol,too,byspecifyingthatcertainthreadsareusedinacertainmannerthroughanexpansionofthePowerThrottlingAPIthatallowsdeveloperstoassignaQoSattributetotheirthreads.Additionally,anewEcoQosclassificationletssoftwaretagthreadsthatrespondbestontheefficiencycorestoassuretheyareprioritizedtoexecuteontheE-Cores.MicrosoftsaysthattheEdgebrowserand'various'Windows11componentsnowtakeadvantageoftheEcoQosclassificationsystem,andwecanexpectsupporttobroadenquickly. (Imagecredit:Intel)Thislookstobeapromisingandless-intrusive(atleastfromacodingstandpoint)methodofensuringthatthecorrectthreadslandonthecorrectcores,thusdeliveringoptimalperformance.Thatsaid,we'llhavetoseeitinactionbeforewecanpassjudgementonitsefficacy–muchofitspotencywillboildowntothelatencyinvolvedwiththeprocessofcommunicatingtelemetrydataandmovingthethread,andintelisn'tsharingthosedetailsyet.Additionally,it'spossiblethatanexcessofcommunicationbetweentheThreadDirectorandtheWindows11schedulercouldcreateachallengingworkloadofitsown,sofindingtherightamountofgranularitywillbekeytoassuringbothtimelythreadplacementandaminimumofsystemoverhead.Thesystemisalreadyfarindevelopment,andMicrosoftsaysthatfurtherenhancementstotheenginearealreadyunderwayandinplanningforWindows11,withmoredetailstobesharedatalaterdate.AlderLakechipswillalsoworkfinewithabog-standardWindows10operatingsystem–existingthread-schedulingtechniquescontinuetoworkwiththeprocessors,justnotaswell.Whilethechipswork,you'llmissoutontheenhancedcapabilitiesofThreadDirector(that'sWindows11only),whichwillhaveavaryingimpactonperformanceandpowerconsumptionbasedoninstructiontypeandapplicationusagemodels.Inotherwords,yourmileagewillvary.Thehybridarchitecturecouldstillresultinsometeethingpains,asIntelitselfrecentlydivulgedthatsomeoldergameswithDRMmightnotworkwiththenewchipsunlessdevelopersaddspecificsoftwareoptimizations.Finally,ithaslongbeenknownthattheGracemontcoresdonotsupporttheAVX-512instructionset,andspeculationhasbeenrifeabouthowthecodewouldworkonAlderLakeprocessors,ifatall.Intel'sanswerissimple:AVX-512willnotworkoneithertypeofcorepresentinAlderLake.Thehigh-performancecoresdofeaturetheGoldenCovearchitecturethatsupportsAVX-512natively,butIntelhasfusedthatfeatureoff(yes,the512-bitFMAisstillpresentandconsumesdiearea)fortheconsumerchips.Incontrast,serverchipswithGoldenCovehavetwo512-bitFMAsandfullysupportAVX-512.Meanwhile,theGracemontcoresaresimplynotAVX-512capable,anddisablingsupportallowstheAlderLakechiptohaveuniformISAsupport. Itiscompletelyclearthatthe AMDvsIntel battlehasbeenreignited.HeadtoourIntelCorei9-12900KandCorei5-12600Kreviewforadditionalin-depthtestingandbuyingadvice.  PaulAlcorn PaulAlcornistheDeputyManagingEditorforTom'sHardwareUS.HewritesnewsandreviewsonCPUs,storageandenterprisehardware. Topics CPUs Seeallcomments(93) 93Comments Commentfromtheforums Anarticlefromtomshardwarelastyear,"DDR5SpecificationReleased:FastRAMWithBuilt-InVoltageRegulators",statesthataDDR5LRDIMMsupportsupto4TBmemory,providedbystackedchipsupto16levels. IwonderifAlderLakewillsupportthisamount. Reply Ican'tsayI'mtoofondoftheBigLittleapproachonDesktopCPUs.. PerhapsonMobile,butIjustdon'tgetthebenefitondesktop. Maybeothersfeeldifferently. Reply mspencerl87said:Ican'tsayI'mtoofondoftheBigLittleapproachonDesktopCPUs.. PerhapsonMobile,butIjustdon'tgetthebenefitondesktop. Maybeothersfeeldifferently.EverybodythoughtthatintelintroducingBGAondesktopwouldmeanthatitwouldendupmeaningallCPUsbutit'sonlyasmallpercentageforwhatevernicheintelthoughtthatitwouldmakesensefor. Myguessisthatthisbig.biggerisgoingtobethesamething,it'sgoingtobeinafewthingswhereitmakessomesenseandeverythingelsewillbenormalCPUsasweknowthem. Reply TheGracemontcoresarealsogoingintheGrandRidgechipswhichwillbethesuccessortotheP5900familybasestationchips. DoestheiruseinAlderLakealsosuggestsomeplanbyInteltointegratemorenetworkingcapabilitiesintotheirdesktopchips? Forexample,acceleratednetworkingmightfitintoIntel'shybridcloudplans. ThisisanarticleshowingaGrandRidgeslideleak...4x100Gbe. https://www.overclock3d.net/news/cpu_mainboard/intel_grand_ridge_architecture_leaked_with_ddr5_and_pcie_4_0/1 Reply mspencerl87said:Ican'tsayI'mtoofondoftheBigLittleapproachonDesktopCPUs.. PerhapsonMobile,butIjustdon'tgetthebenefitondesktop. Maybeothersfeeldifferently.Whiletheaveragehomeusermaynotseeapoint,thiscanbeaprettybigdealforcompaniesthatemploythousandsofcomputersinsomeformorfashion.Likesayforexampleaserverthat'snotreallybeingusedbecauseoflowdemand.RatherthanshutitdownorputitindeepsleepandhopeWoLgetsitup,whateverappsandservicesthatmakesuptheskeletoncrewforsaidservercanremainrunningonthelowerpowercoresuntilpowerneedstorampupagain.There'sthepossibilityofbothasignificantpowersavingandreducingdowntime. Thoughforme,Iwouldreallyliketoreducemycomputer'sidle/lowpowerusageevenfurther.I'mnothurtingonmyelectricbill,butnonetheless,chippingoffouncescanleadtopounds. Reply TerryLazesaid:EverybodythoughtthatintelintroducingBGAondesktopwouldmeanthatitwouldendupmeaningallCPUsbutit'sonlyasmallpercentageforwhatevernicheintelthoughtthatitwouldmakesensefor. Myguessisthatthisbig.biggerisgoingtobethesamething,it'sgoingtobeinafewthingswhereitmakessomesenseandeverythingelsewillbenormalCPUsasweknowthem.ExceptIntelhasalreadyannounceditasthearchitecturalsequeltoRocketLakeandTigerLake.SounlessIntelplansonreleasingAlderLakeCPU'swithoutthehybridconfiguration,whichisoneofthemajorfeaturestheyarepromoting,orthereisasurpriseunannounced/leakedarchitectureforhighperformancedesktops,it'snotgoingtobeanichefeature. CES2021:IntelAnnouncesFourNewProcessorFamilies|IntelNewsroom "Dueinsecondhalfof2021,AlderLakewillcombinehigh-performancecoresandhigh-efficiencycoresintoasingleproduct.AlderLakewillalsobeIntel’sfirstprocessorbuiltonanew,enhancedversionof10nmSuperFinandwillserveasthefoundationforleadershipdesktopandmobileprocessorsthatdeliversmarter,fasterandmoreefficientreal-worldcomputing." Reply Intel'srecentpriceadjustmentshavegivenCometLakeasolidvaluepropositioncomparedtoAMD'sRyzen5000chips. I'vebeenrockin'ani7-4790Kforover5yearsandtheZen3processorswereabigenticementtofinallyupgrade.TroublewasIcouldn'tfindoneatMSRP,orevenclose.So,withCometLakepricingbeingreduced,Ioptedforani9-10850onaZ490MB.Iamreallyhappywiththeresults! Wonderwhatwillbeavailablein5moreyears?BythenI'llbe80yearsoldandmightnotcare.🆒 John Reply spongiemastersaid:ExceptIntelhasalreadyannounceditasthearchitecturalsequeltoRocketLakeandTigerLake.SounlessIntelplansonreleasingAlderLakeCPU'swithoutthehybridconfiguration,whichisoneofthemajorfeaturestheyarepromoting,orthereisasurpriseunannounced/leakedarchitectureforhighperformancedesktops,it'snotgoingtobeanichefeature. CES2021:IntelAnnouncesFourNewProcessorFamilies|IntelNewsroom "Dueinsecondhalfof2021,AlderLakewillcombinehigh-performancecoresandhigh-efficiencycoresintoasingleproduct.AlderLakewillalsobeIntel’sfirstprocessorbuiltonanew,enhancedversionof10nmSuperFinandwillserveasthefoundationforleadershipdesktopandmobileprocessorsthatdeliversmarter,fasterandmoreefficientreal-worldcomputing."Fromthisarticle. Whilethisisnoconfirmationbutitonlymakessense. Therearealsosignsthatsomemodelswillcomewithonlythebigcoresactive,whichshouldperformexceedinglywellingaming. Reply mspencerl87said:Ican'tsayI'mtoofondoftheBigLittleapproachonDesktopCPUs.. PerhapsonMobile,butIjustdon'tgetthebenefitondesktop. Maybeothersfeeldifferently.Itssimple.AMDscoresareweakerthanIntelthatswhytheycanbeatintelinmulticorebenchmarkscoreswith2Xthecorecountinagivenpowerbudget.WhileIntel'sbigcoresaresubstantiallybetteratsinglecorescore.WithAlderlake,IntelhopestocheckmateAMDinbothsinglecorescore(withBigcore)aswellasinmulticorescorewithlargertotalcomputepoweratroughlyisoprocesstechnology. Reply TerryLazesaid:Fromthisarticle. Whilethisisnoconfirmationbutitonlymakessense.LeaksfromawhilebackshowedtheCPU'swithnolittlecoreswereclusteredatlowercorecounts.Intelwillfindawaytoleveragethosesmallercoresforbetteroverallperformance,sowhileyoumaybeabletobuyan8coreALwithnosmallcores,you'relikelynotgoingtowantto.ItwouldbelikebuyingahighendCPUtodaywithouthyperthreading,exceptyou'llbeforgoingevenmoreperformance. https://www.tomshardware.com/news/intel-alder-lake-already-looks-confusing-12-configurations-possible Reply ViewAll93Comments Showmorecomments BeIntheKnow Getinstantaccesstobreakingnews,in-depthreviewsandhelpfultips. ContactmewithnewsandoffersfromotherFuturebrands Receiveemailfromusonbehalfofourtrustedpartnersorsponsors ThankyouforsigninguptoTom'sHardware.Youwillreceiveaverificationemailshortly. Therewasaproblem.Pleaserefreshthepageandtryagain. MOSTPOPULARMOSTSHARED1GPUBenchmarksandHierarchy2022:GraphicsCardsRanked2IntelInvests$700MillioninImmersionLiquidCoolingSolutions3TheFutureGamesShowPoweredbyManaAirsSaturday,June114SKHynix'sNewSSDBoasts1.4MillionIOPS5TSMCtoInitiate1.4nmProcessTechnologyR&D1GPUBenchmarksandHierarchy2022:GraphicsCardsRanked2IntelInvests$700MillioninImmersionLiquidCoolingSolutions3TheFutureGamesShowPoweredbyManaAirsSaturday,June114SKHynix'sNewSSDBoasts1.4MillionIOPS5TSMCtoInitiate1.4nmProcessTechnologyR&D