Imbali yanamhlanje yokujongana ne-Intel kunye ne-AMD kwimarike yeprosesa ibuyele kwisiqingatha sesibini se-90s. Ixesha leenguqu ezinkulu kunye nokungena kwizinto eziqhelekileyo, xa i-Intel Pentium ibekwe njengesisombululo sendalo yonke, kwaye i-Intel Inside yaba phantse isilogeni esaziwayo kwihlabathi, saphawulwa ngamaphepha aqaqambileyo kwimbali engekho nje eluhlaza okwesibhakabhaka, kodwa kwakhona. obomvu - ukuqala kwisizukulwana se-K6, i-AMD ikhuphisane ngokungadinwa ne-Intel kumacandelo amaninzi emarike. Nangona kunjalo, yayizizehlo zenqanaba lamva kancinci-isiqingatha sokuqala seminyaka yoo-XNUMX-esidlale indima ebalulekileyo ekuveleni kolwakhiwo lweCore oludumileyo, olusephantsi komgca we-Intel processor.
Imbali encinci, imvelaphi kunye nenguquko
Ukuqala kweminyaka yoo-2000 kunxulunyaniswa kakhulu nezigaba ezininzi kuphuhliso lweeprosesa - ugqatso lwe-1 GHz olunqwenelekayo rhoqo, ukubonakala kweprosesa yokuqala ye-double-core, kunye nomzabalazo oqatha wokubaluleka kwinqanaba le-desktop. Emva kokuba iPentium iphelelwe lithemba kwaye i-Athlon 64 X2 ingene kwimarike, i-Intel yazisa i-Core generation processors, eyathi ekugqibeleni yaba yinto yokutshintsha kuphuhliso lweshishini.
Iiprosesa zokuqala ze-Core 2 Duo zabhengezwa ekupheleni kukaJulayi 2006 - ngaphezu konyaka emva kokukhululwa kwe-Athlon 64 X2. Kumsebenzi wayo kwisizukulwana esitsha, i-Intel yayikhokelwa ikakhulu yimiba yokulungiswa kwezakhiwo, ukufezekisa ezona zibonakaliso ziphezulu zokusebenza kwamandla esele kwizizukulwana zokuqala zeemodeli ezisekwe kuyilo lweCore, enekhowudi ebizwa ngokuba yiConroe-babengaphezulu kwesihlandlo esinye nesiqingatha. I-Pentium 4, kunye ne-package ye-thermal ebhengeziweyo ye-65 W, insimbi, mhlawumbi, awona maprosesa anamandla kakhulu kwiimarike ngelo xesha. Isebenza njenge-catch-up (eyenzeka ngokungaqhelekanga), i-Intel iphunyezwe kwinkxaso yesizukulwana esitsha kwimisebenzi ye-64-bit kunye noyilo lwe-EM64T, isethi entsha yemiyalelo ye-SSSE3, kunye nephakheji ebanzi ye-x86-based virtualization technologies.
Core 2 Duo microprocessor die
Ukongeza, enye yeempawu eziphambili ze-Conroe processors yayiyi-cache enkulu ye-L2, impembelelo yayo ekusebenzeni ngokubanzi kweeprosesa kwabonakala kakhulu nangoko. Emva kokuthatha isigqibo sokwahlula amacandelo eprosesa, i-Intel ivaliwe isiqingatha se-4 MB L2 cache kubameli abancinci bomgca (E6300 kunye ne-E6400), ngaloo ndlela iphawula icandelo lokuqala. Nangona kunjalo, iimpawu zetekhnoloji ye-Core (isizukulwana esisezantsi sobushushu kunye nokusebenza okuphezulu kwamandla okuhambelana nokusetyenziswa kwe-lead solder) kuvumele abasebenzisi abaphambili ukuba bafezekise iifrikhwensi eziphezulu kakhulu kwizisombululo zenkqubo ephucukileyo - iibhodi zebhodi ezikumgangatho ophezulu zenze ukuba kube lula ukugqithisa ibhasi yeFSB. , ukwandisa ukuphindaphinda kweprosesa encinci ukuya kwi-3 GHz nangaphezulu (ukubonelela ngokunyuka kwe-60%), ngenxa yokuba iikopi eziphumeleleyo ze-E6400 zinokukhuphisana nabazalwana babo abadala i-E6600 kunye ne-E6700, nangona ixabiso leengozi ezinkulu zeqondo lokushisa. . Nangona kunjalo, kunye nokugqithisa okuthobekileyo kwenza ukuba kube lula ukufumana iziphumo ezinzulu - kwiibhentshi, iiprosesa ezindala zithathe indawo ye-Athlon 64 X2 ephambili, ephawula isikhundla seenkokeli ezintsha kunye nezinto ezithandwa ngabantu.
Ukongeza, i-Intel iqalise uguqulo lokwenyani - iiprosesa ze-quad-core zosapho lwase-Kentsfield kunye nesimaphambili se-Q, esakhelwe kwi-nanometers efanayo ye-65, kodwa sisebenzisa ulwakhiwo lweetshiphusi ezimbini ze-Core 2 Duo kwi-substrate enye. Emva kokuba iphumelele amandla aphakamileyo asebenzayo (iqonga lisebenzise inani elifanayo njengeekristale ezimbini ezisetyenzisiweyo ngokwahlukeneyo), i-Intel okokuqala ibonise indlela enamandla ngayo inkqubo enemicu emine - kwizicelo zemultimedia, ukugcinwa kunye nemidlalo enzima esebenzisa umthwalo ngokusebenzayo. ukuhambelana kwimisonto emininzi (ngo-2007, ezi yayiziCrysis ezichukumisayo kunye neeGears zeMfazwe ezingekho ngaphantsi kwe iconic), umahluko ekusebenzeni ngoqwalaselo lweprosesa enye ungafikelela kwi-100%, eyayiluncedo olukhulu kuye nawuphi na umthengi. inkqubo Core 2 Quad-based.
Ukuncamathelisa ii-C2D ezimbini kwi-substrate enye - i-Core 2 Quad
Njengomgca wePentium, iiprosesa ezikhawulezayo zonyulwa ngokugqithisileyo kunye nesiqalo se-QX, kwaye zafumaneka kubantu abanomdla kunye nabakhi benkqubo ye-OEM ngexabiso eliphezulu kakhulu. Isithsaba sesizukulwana se-65-nm sasiyi-QX6850 kunye ne-frequency ye-3 GHz kunye nebhasi ye-FSB ekhawulezayo esebenza kwi-frequency ye-1333 MHz. Le prosesa iye yathengiswa nge $999.
Ewe kunjalo, impumelelo enjalo yayingenako ukuhlangabezana nokhuphiswano oluvela kwi-AMD, kodwa isigebenga esibomvu ngelo xesha sasingekagqibi kwimveliso yeeprosesa ze-quad-core, ukuchasana neemveliso ezintsha ezivela kwi-Intel, iqonga lovavanyo le-Quad FX. , ephuhliswe ngentsebenziswano ne-NVidia, yaboniswa kwaye yafumana imodeli enye ye-serial ye-ASUS L1N64 motherboard, eyenzelwe ukusebenzisa ii-Athlon FX X2 ezimbini kunye ne-Opteron processors.
I-ASUS L1N64
Iqonga liye laba yinto entsha yobugcisa obunomdla kwizinto eziqhelekileyo, kodwa ezininzi iindibano zobugcisa, ukusetyenziswa kwamandla amakhulu kunye nokusebenza okuphakathi (ngokuthelekisa imodeli ye-QX6700) ayizange ivumele iqonga ukuba likhuphisane ngempumelelo kwicandelo eliphezulu lemarike. - I-Intel yafumana isandla esiphezulu, kwaye iiprosesa ze-Phenom FX ezine-cores ezine zabonakala zibomvu kuphela ngoNovemba 2007, xa umqhubi ekulungele ukuthatha isinyathelo esilandelayo.
Umgca wePenryn, owawubizwa ngokuba yinto ebizwa ngokuba yi-die-shrink (ukunciphisa ubungakanani bokufa) ye-65 nm chips ukusuka kwi-2007, yaqala kwimarike ngoJanuwari 20, i-2008 kunye neeprosesa zeWolfdale - kwiinyanga nje ezi-2 emva kokukhululwa kwe-AMD's Phenom FX. . Ukutshintshela kwi-teknoloji yenkqubo ye-45-nm usebenzisa i-dielectrics yakutshanje kunye nezixhobo zokuvelisa zasivumela ukuba sandise i-horizons ye-architecture ye-Core ngakumbi. Iiprosesa zafumana inkxaso ye-SSE4.1, inkxaso yezinto ezintsha zokugcina amandla (ezifana ne-Deep Power Down, ephantse ikhuphe amandla okusetyenziswa kwamandla kwi-hibernation state kwi-mobile versions of processors), kwaye yaba yinto epholileyo kakhulu - kwezinye iimvavanyo umahluko inokufikelela kwi-10 degrees xa kuthelekiswa nochungechunge lwangaphambili lweConroe. Emva kokwandisa i-frequency kunye nokusebenza, kunye nokufumana i-cache ye-L2 eyongezelelweyo (ye-Core 2 Duo umthamo wayo unyuke ukuya kwi-6 MB), iiprosesa ezintsha ze-Core zikhusele izikhundla zabo eziphambili kwi-benchmarks, kwaye zavula indlela yokuqhubela phambili ukhuphiswano oluqatha kwaye ukuqala kwexesha elitsha. Ii-Epochs zempumelelo engazange ibonwe ngaphambili, amaxesha okudodobala nokuzola. Ixesha leCore i processors.
Inyathelo elinye phambili kwaye zero emva. Isizukulwana sokuqala Core i7
Sele ngoNovemba ka-2008, i-Intel yazisa i-architecture entsha ye-Nehalem, ephawula ukukhululwa kweeprosesa zokuqala kwi-Core i series, eyaziwa kakhulu kuwo wonke umsebenzisi namhlanje. Ngokungafaniyo ne-Core 2 Duo eyaziwayo, i-architecture ye-Nehalem ekuqaleni ibonelele ngee-cores ezine zomzimba kwi-chip enye, kunye nenani leempawu zokwakha ezaziwa kuthi kwizinto ezintsha zobugcisa ezivela kwi-AMD - umlawuli wememori odibeneyo, i-cache yenqanaba lesithathu ekwabelwana ngalo. , kunye ne-QPI- ujongano oluthatha indawo ye-HyperTransport.
Intel Core i7-970 microprocessor die
Xa umlawuli wememori eshukunyiswe phantsi kwesigqubuthelo somprosesa, i-Intel yanyanzelwa ukuba iphinde yakhe yonke i-cache isakhiwo, ukunciphisa ubukhulu be-L2 cache ngokuthanda i-L3 cache edibeneyo ye-8 MB. Nangona kunjalo, eli nyathelo lenze ukuba kube lula ukunciphisa inani lezicelo, kunye nokunciphisa i-L2 cache ukuya kwi-256 KB ngondoqo ngamnye kwaba sisisombululo esisebenzayo malunga nesantya somsebenzi kunye nokubalwa kweentambo ezininzi, apho ubuninzi bomthwalo. yabhekiswa kwi-cache ye-L3 eqhelekileyo.
Ukongeza kuhlengahlengiso lwe-cache, i-Intel yathatha inyathelo eliya phambili ne-Nehalem, ibonelela ngeeprosesa ngenkxaso ye-DDR3 kwi-frequencies ye-800 kunye ne-1066 MHz (nangona kunjalo, imigangatho yokuqala yayikude nokunciphisa ezi processors), kunye nokususa inkxaso ye-DDR2, ngokungafaniyo ne-AMD, esebenzise umgaqo wokubuyela ngasemva kwi-Phenom II processors, ekhoyo kuzo zombini i-AM2 + kunye ne-AM3 sockets entsha. Umlawuli wememori ngokwayo e-Nehalem unokusebenza kwenye yeendlela ezintathu kunye nenye, ezimbini okanye ezintathu zememori itshaneli kwi-64, 128 okanye i-192-bit ibhasi, ngokulandelelana, enkosi apho abavelisi be-motherboard babeke ukuya kuthi ga kwi-6 DIMM DDR3 i-memory connectors kwi-PCB. . Ngokubhekiselele kwi-interface ye-QPI, ithathe indawo yebhasi ye-FSB esele iphelelwe lixesha, inyusa i-bandwidth yeqonga ubuncinci kabini-ebesisisombululo esihle kakhulu ukusuka kwindawo yokujonga ukunyusa iimfuno zokuphindaphindwa kwememori.
I-Hyper-Threading elityelweyo ibuyele eNehalem, inika ii-cores ezine ezinamandla ezinemisonto esibhozo, kwaye ibangela "loo SMT." Ngapha koko, i-HT yaphunyezwa emva kwiPentium, kodwa ukusukela ngoko i-Intel khange icinge ngayo kude kube ngoku.
Itekhnoloji ye-Hyper-Treading
Olunye uphawu lobuchwephesha lwesizukulwana sokuqala i-Core i yayisisiqhelo sokusebenza kwe-cache kunye nabalawuli beememori, ulungelelwaniso olubandakanya ukutshintsha iiparamitha eziyimfuneko kwi-BIOS - i-Intel icetyiswa ukuba iphindaphinde imemori yememori yokusebenza ngokufanelekileyo, kodwa nokuba yinto encinci. inokuba yingxaki kubasebenzisi abathile, ngakumbi xa begqithisa iibhasi ze-QPI (ibhasi ye-BCLK), kuba kuphela i-flagship ebiza kakhulu yomgca we-i7-965 kunye nethegi ye-Extreme Edition ifumene umphindaphindi ongavulwanga, ngelixa i-940 kunye ne-920 ine-frequency esisigxina. kunye nesiphindaphindi sama-22 kunye nama-20, ngokulandelelanayo.
I-Nehalem iye yaba nkulu ngokwasemzimbeni (ubungakanani beprosesa bonyuke kancinci xa kuthelekiswa ne-Core 2 Duo ngenxa yesilawuli sememori esishukunyiswa phantsi kwekhava) kwaye phantse.
Ukuthelekiswa kobungakanani beprosesa
Ndiyabulela "ekubekweni esweni" kwenkqubo yamandla, umlawuli wePCU (iYunithi yoLawulo lwaMandla), kunye nemowudi yeTurbo, yenze ukuba kube lula ukufumana ifrikhwensi encinci (kwaye, ke, ukusebenza) nangaphandle kohlengahlengiso lwezandla, olulinganiselweyo kuphela. kwi-nameplate amaxabiso e-130 W. Kuyinyani, kwiimeko ezininzi lo mda unokutyhalelwa ngasemva ngokuguqula useto lwe-BIOS, ufumane i-100-200 MHz eyongezelelweyo.
Lilonke, i-architecture ye-Nehalem yayinokubonelela kakhulu - ukunyuka okukhulu kwamandla xa kuthelekiswa ne-Core 2 Duo, ukusebenza kweentambo ezininzi, ii-cores ezinamandla kunye nenkxaso yemigangatho yamva nje.
Kukho ukungaqondi kakuhle okunxulumene nesizukulwana sokuqala se-i7, oko kukuthi ubukho beesokethi ezimbini ze-LGA1366 kunye ne-LGA1156 efanayo (ekuqaleni) i-Core i7. Nangona kunjalo, iiseti ezimbini zengqiqo zazingekho ngenxa ye-whim yequmrhu elibawayo, kodwa kwinguqu kwi-architecture yaseLynnfield, inyathelo elilandelayo kuphuhliso lomgca we-Core i processor.
Ngokuphathelele ukhuphiswano oluvela kwi-AMD, isigebenga esibomvu sasingangxamanga ukutshintshela kuyilo olutsha lwenguquko, singxamele ukuhambisana nesantya se-Intel. Ukusebenzisa i-K10 endala, inkampani yakhulula i-Phenom II, eyaba yinguqu kwi-teknoloji yenkqubo ye-45-nm yesizukulwana sokuqala iPhenom ngaphandle kotshintsho olubalulekileyo lwezakhiwo.
Ngombulelo ekunciphiseni kwendawo yokufa, i-AMD yakwazi ukusebenzisa indawo eyongezelelweyo ukulungiselela i-cache ye-L3 ekhangayo, ekubunjweni kwayo (kunye nokuhlelwa ngokubanzi kwezinto kwi-chip) ngokuhambelana nophuhliso lwe-Intel kunye ne-Nehalem, kodwa inani lezinto ezingeloncedo ngenxa yomnqweno woqoqosho kunye nokuhambelana ngasemva kunye neqonga le-AM2 eliguga ngokukhawuleza.
Emva kokulungisa iziphene kumsebenzi we-Cool'n'Quiet, engazange isebenze kwisizukulwana sokuqala se-Phenom, i-AMD yakhupha iinguqulelo ezimbini zePhenom II, eyokuqala eyabhekiswa kubasebenzisi kwii-chipsets ezindala ukusuka kwisizukulwana se-AM2, kunye neyesibini yeqonga le-AM3 elihlaziyiweyo kunye nenkxaso yememori ye-DDR3. Yayingumnqweno wokugcina inkxaso yeeprosesa ezintsha kwiibhodi zomama ezindala ezazidlala ihlaya elikhohlakeleyo kwi-AMD (leyo, nangona kunjalo, iya kuphinda iphindwe kwixesha elizayo) - ngenxa yeempawu zeqonga ngendlela yebhulorho ecothayo yasentla, iPhenom entsha. I-II X4 ayikwazanga ukusebenza kwi-frequency elindelekileyo yebhasi ye-uncore (umlawuli wememori kunye ne-L3 cache), ilahlekelwa yintsebenzo engaphezulu kuhlaziyo lokuqala.
Nangona kunjalo, i-Phenom II yayifikeleleka kwaye inamandla ngokwaneleyo ukubonisa iziphumo kwinqanaba lesizukulwana sangaphambili se-Intel - oko kukuthi i-Core 2 Quad. Ngokuqinisekileyo, oku kuthetha kuphela ukuba i-AMD yayingakulungele ukukhuphisana neNehalem. Kukonke.
Kwaye kwafika iWestmere...
eWestmere. Ixabiso eliphantsi kune-AMD, ngokukhawuleza kuneNehalem
Iingenelo zePhenom II, ezivezwa yi-giant ebomvu njengenye indlela yohlahlo lwabiwo-mali kwi-Q9400, ibeka kwizinto ezimbini. Eyokuqala ihambelana ngokucacileyo kunye neqonga le-AM2, elifumene abalandeli abaninzi beekhompyuter ezingabizi kakhulu ngexesha lokukhutshwa kwePhenom yokuqala. Eyesibini lixabiso elimnandi, elingeyiyo i-i7 9xx ebizayo okanye efikelelekayo (kodwa ayisenangeniso) Iiprosesa ze-Code 2 Quad series zinokukhuphisana nazo. I-AMD yayibheja ngokufikeleleka kolona luhlu lubanzi lwabasebenzisi, abadlali abaqhelekileyo kunye neengcali eziqaphela uhlahlo lwabiwo-mali, kodwa i-Intel yayisele inesicwangciso sokubetha onke amakhadi e-chipmaker abomvu ngasekhohlo.
Embindini wayo yayiyiWestmere, uphuhliso olulandelayo lolwakhiwo lweNehalem (undoqo weBloomfield), oluye lwazibonakalisa phakathi kwabanomdla kunye nabo bakhetha ukuthatha okona kulungileyo. Ngeli xesha, i-Intel yazishiya izisombululo ezibiza kakhulu - iseti entsha yengqiqo esekwe kwi-LGA1156 socket iphulukene nesilawuli se-QPI, yafumana i-DMI eyenziwe lula ngolwakhiwo, yafumana isilawuli sememori ye-DDR3, kwaye yaphinda yaqondisa kwakhona eminye yemisebenzi phantsi kwesixhobo. iprosesa coverage - ngeli xesha yaba umlawuli PCI.
Ngaphandle kwento yokuba ngokubonakalayo i-Core i7-8xx entsha kunye ne-Core i5-750 ziyafana ngobukhulu kwi-Core 2 Quad, enkosi kutshintsho ukuya kwi-32 nm, ikristale yajika yaba nkulu ngakumbi kunaleyo ye-Nehalem - idini. iziphumo ezongezelelweyo ze-QPI kunye nokudibanisa ibhloko yezibuko ze-I / O eziqhelekileyo, iinjineli ze-Intel zidibanise umlawuli we-PCI, othatha i-25% yendawo ye-chip kwaye yenzelwe ukunciphisa ukulibaziseka ekusebenzeni kunye ne-GPU, ngenxa yokuba iindlela ezongezelelweyo ze-PCI ze-16 azizange zigqithise.
E-Westmere, imodi ye-Turbo nayo yaphuculwa, eyakhelwe kumgaqo "we-cores ezininzi - ngaphantsi kwe-frequency", esetyenziswe yi-Intel ukuza kuthi ga ngoku. Ngokwengqiqo yeenjineli, umda we-95 W (leyo kanye ingakanani i-flagship ehlaziyiweyo bekufanele idle) ayizange ihlale ifunyenwe kwixesha elidlulileyo ngenxa yogxininiso lokugqithisa zonke ii-cores kuyo nayiphi na imeko. Imowudi ehlaziyiweyo yenza ukuba kube lula ukusebenzisa i-overclocking "smart", i-dosing frequencies ngendlela yokuba xa i-core enye isetyenzisiwe, ezinye zivaliwe, zikhulula amandla ongezelelweyo ukugqithisa i-core echaphazelekayo. Ngendlela elula ngolo hlobo, kwavela ukuba xa i-overclocking enye ingundoqo, umsebenzisi ufikelele kwi-frequency ye-clock ephezulu, xa i-overclocking ezimbini, yayingaphantsi, kwaye xa i-overclocking yonke emine, yayingabalulekanga. Yile ndlela i-Intel iqinisekise ngayo ukusebenza okuphezulu kwimidlalo emininzi kunye nezicelo zisebenzisa intambo enye okanye emibini, ngelixa igcina amandla okusebenza ukuba i-AMD yayinokuphupha ngayo emva koko.
Iyunithi yokulawula amandla, ejongene nokusabalalisa amandla phakathi kwee-cores kunye nezinye iimodyuli kwi-chip, nayo iphuculwe kakhulu. Ngombulelo kuphuculo lwenkqubo yobugcisa kunye nophuculo lobunjineli kwizinto eziphathekayo, i-Intel yakwazi ukwenza inkqubo ephantse ibe yinyani apho iprosesa, ngelixa ikwimo engasebenziyo, iyakwazi ukusebenzisa amandla angekho ONKE. Kuyaphawuleka ukuba ukufezekisa isiphumo esinjalo akuhambelani notshintsho lwezakhiwo - iyunithi yokulawula ye-PSU ihanjiswe phantsi kwe-Westmere cover ngaphandle kotshintsho, kwaye kuphela iimfuno ezongeziweyo zemathiriyeli kunye nomgangatho jikelele owenze ukuba kube nokwenzeka ukunciphisa imisinga yokuvuza ukusuka kwii-cores ezivaliweyo ukuya kwi-zero. okanye phantse ukuya kwiqanda) iprosesa kunye neemodyuli ezikhaphayo zikwimo engasebenziyo.
Ngokutshintshiselana ngesilawuli sememori esineendlela ezintathu kwitshaneli ezimbini, iWestmere ibinokulahlekelwa ngumsebenzi othile, kodwa ngenxa yokwanda kokuphindaphinda kwenkumbulo (1066 yeNehalem eqhelekileyo, kunye ne-1333 yeqhawe lale nxalenye yenqaku), entsha. I-i7 ayizange ilahlekelwe yintsebenzo kuphela, kodwa kwezinye iimeko yavela ngokukhawuleza kune-Nehalem processors. Nakwizicelo ezingasebenzisi zonke ii-cores ezine, i-i7 870 iye yaphantse yafana nomzalwana wayo omdala ngenxa yenzuzo kwi-DDR3 frequency.
Ukusebenza kwemidlalo ye-i7 ehlaziyiweyo yayiphantse yafana nesona sisombululo silungileyo sesizukulwana sangaphambili - i-i7 975, ebiza ngokuphindwe kabini. Kwangaxeshanye, isisombululo esincinci silungelelanise elunxwemeni kunye nePhenom II X4 965 BE, ngamanye amaxesha ngokuzithemba phambi kwayo, kwaye ngamanye amaxesha kancinci.
Kodwa ixabiso lalingumba kanye owabhidanisa bonke abalandeli be-Intel - kunye nesisombululo ngendlela engakholelekiyo yeedola ezili-199 zeCore i5 750 ifanele wonke umntu ngokugqibeleleyo. Ewe, bekungekho mowudi ye-SMT apha, kodwa iicores ezinamandla kunye nokusebenza okugqwesileyo kwenze ukuba kwenzeke ukuba kungapheleli nje ekugqweseni iprosesa ye-AMD, kodwa nokuyenza ngexabiso eliphantsi.
La yayingamaxesha amnyama kwiiReds, kodwa babene-ace phezulu-iprosesa entsha ye-AMD FX yayiza kukhutshwa. Enyanisweni, i-Intel ayizange ifike ingaxhobanga.
Ukuzalwa kwentsomi kunye nemfazwe enkulu. Sandy Bridge vs AMD FX
Ukujonga emva kwimbali yobudlelwane phakathi kwezi ngxilimbela zimbini, kuyacaca ukuba yayilixesha lika-2010-2011 elalinxulunyaniswa nolindelo olumangalisayo lwe-AMD, kunye nezisombululo ezingalindelekanga ze-Intel. Nangona zombini iinkampani zithathe umngcipheko ngokubonisa uyilo olutsha ngokupheleleyo, kuba iiReds ukubhengezwa kwesizukulwana esilandelayo kunokuba yintlekele, ngelixa i-Intel, ngokubanzi, yayingenamathandabuzo.
Ngelixa iLynnfield yayisisilungiso esikhulu sebug, iSandy Bridge yabuyisela iinjineli kwibhodi yokuzoba. Ukutshintshela kwi-32 nm kuphawulwe ukudala isiseko se-monolithic, akusekho konke okufanayo kwi-layout ehlukeneyo esetyenziswe e-Nehalem, apho iibhloko ezimbini ze-cores ezimbini zahlula i-crystal zibe ziinxalenye ezimbini, kwaye iimodyuli zesibini zibekwe emacaleni. Kwimeko yeSandy Bridge, i-Intel yenza i-monolithic layout, apho i-cores ibekwe kwibhloko enye, isebenzisa i-cache ye-L3 eqhelekileyo. Umbhobho olawulayo owenza umbhobho womsebenzi uhlaziywe ngokupheleleyo, kwaye ibhasi ye-ring-speed ring inikeze ukulibaziseka okuncinci xa isebenza ngememori kwaye, ngenxa yoko, ukusebenza okuphezulu kuyo nayiphi na imisebenzi.
Intel Core i7-2600k microprocessor chip
Iigraphics ezidityanisiweyo nazo zavela phantsi kwe-hood, ehlala kwi-20% ye-chip efanayo kwindawo - okokuqala ngqa kwiminyaka emininzi, i-Intel yagqiba ekubeni ijongane ne-GPU eyakhelwe-ngaphakathi. Kwaye nangona ibhonasi enjalo ingabalulekanga ngokwemigangatho yamakhadi adityanisiweyo, awona makhadi emizobo eSandy Bridge anokungafuneki. Kodwa ngaphandle kwe-112 yezigidi ze-transistors ezabelwe i-chip yegraphic, kwiSandy Bridge iinjineli ze-Intel zithembele ekwandiseni ukusebenza okungundoqo ngaphandle kokunyusa indawo yokufa, ekuqaleni ukuyijonga ayingomsebenzi olula-isizukulwana sesithathu sokufa yi-2 mm2 kuphela enkulu kuno I-Q9000 yayikhe yaba . Ngaba iinjineli ze-Intel zikwazile ukufezekisa okumangalisayo? Ngoku impendulo ibonakala icacile, kodwa masiyigcine inika umdla. Siza kubuyela kule nto kungekudala.
Ukongeza kwi-architecture entsha ngokupheleleyo, iSandy Bridge iye yaba ngowona mgca mkhulu weprosesa kwimbali ye-Intel. Ukuba ngexesha le-Lynnfield i-blues ibonise imodeli ye-18 (i-11 yee-PC eziphathwayo kunye ne-7 ye-desktops), ngoku uluhlu lwazo luye lwanda ukuya kwi-29 (!) Ii-SKU zazo zonke iiprofayili ezinokwenzeka. IiPC zeDesktop zifumene i-8 yazo ekukhululweni - ukusuka kwi-i3-2100 ukuya kwi-i7-2600k. Ngamanye amazwi, onke amacandelo emarike agutyungelwe. Eyona nto ifikelelekayo i-i3 yanikezelwa nge-117 yeedola, kwaye i-flagship ixabisa i-317 yeedola, eyayiphantsi kakhulu ngemigangatho yezizukulwana ezidlulileyo.
Kwiintetho zentengiso, i-Intel yabiza iSandy Bridge "isizukulwana sesibini se-Core processors," nangona ngokobugcisa kwakukho izizukulwana ezithathu ngaphambi kwayo. I-blues ichaze ingqiqo yabo ngokubalwa kweeprosesa, apho inani emva kwe-i * ukutyunjwa lilinganiswe nesizukulwana - kungenxa yesi sizathu abaninzi basakholelwa ukuba i-Nehalem yayikuphela koyilo lwe-i7 yesizukulwana sokuqala.
Eyokuqala kwimbali ye-Intel, iSandy Bridge ifumene igama leeprosesa ezivuliwe - unobumba u-K kwigama lomzekelo, elithetha ukuphindaphinda kwamahhala (njengoko i-AMD yayithanda ukwenza, okokuqala kuthotho lweeprosesa eziMnyama, emva koko yonke indawo). Kodwa, njengakwimeko ye-SMT, ubunewunewu obunjalo babufumaneka kuphela ngentlawulo eyongezelelweyo kwaye ngokukodwa kwiimodeli ezimbalwa.
Ukongeza kumgca weklasikhi, iSandy Bridge nayo yayineeprosesa ezibhalwe T kunye no-S, ezijoliswe kubakhi beekhompyutha kunye neenkqubo eziphathekayo. Ngaphambili, i-Intel ayizange icinge nzulu ngeli candelo.
Ngotshintsho ekusebenzeni kwe-multiplier kunye nebhasi ye-BCLK, i-Intel ivalwe amandla okugqithisa iimodeli zeSandy Bridge ngaphandle kwesalathisi se-K, ngaloo ndlela ivala i-lophole esebenza ngokugqibeleleyo eNehalem. Ubunzima obuhlukeneyo kubasebenzisi yayiyinkqubo "yokugqithiswa okulinganiselweyo", eyenza ukuba kube lula ukuseta ixabiso le-Turbo frequency kwiprosesa eyayivinjwe ulonwabo lwemodeli evuliwe. Umgaqo wokusebenza we-overclocking ngaphandle kwebhokisi uhlala ungatshintshi kunye noLynnfield - xa usebenzisa ingundoqo enye, inkqubo ivelisa ubuninzi obukhoyo (kuquka ukupholisa) ukuphindaphinda, kwaye ukuba iprosesa ilayishwe ngokupheleleyo, i-overclocking iya kuba ngaphantsi kakhulu, kodwa kuzo zonke ii-cores. .
Ukugqithisa ngesandla kweemodeli ezivuliweyo, ngokuchaseneyo, kuye kwehla kwimbali ngenxa yamanani iSandy Bridge evumeleke ukuba iwafezekise nangona idityaniswe neyona nto ilula ipholile. 4.5 GHz ngaphandle kokuchitha ekupholiseni? Akukho mntu wakha watsibela phezulu kangaka ngaphambili. Ungasathethi ke ukuba i-5 GHz yayisele ifikeleleke ukusuka kwindawo yokujonga ngaphezulu kunye nokupholisa okwaneleyo.
Kunye nezinto ezintsha zokwakha, iSandy Bridge yayikhatshwa zizinto ezintsha zobugcisa - iqonga elitsha le-LGA1155 elixhotyiswe ngenkxaso ye-SATA 6 Gb/s, ukubonakala kwe-UEFI interface ye-BIOS, kunye nezinye izinto ezincinci ezithandekayo. Iqonga elihlaziyiweyo lifumene inkxaso yemveli ye-HDMI 1.4a, i-Blu-Ray 3D kunye ne-DTS HD-MA, ngenxa yoko, ngokungafaniyo nezisombululo zedeskithophu ezisekelwe kwi-Westmere (i-Clarkdale core), iSandy Bridge ayizange ifumane ubunzima obungathandekiyo xa ikhupha ividiyo kwiiTV zanamhlanje kunye ukudlala iimuvi kwiifreyimu ezingama-24, nto leyo ngokungathandabuzekiyo yabakholisa abalandeli bethiyetha yasekhaya.
Nangona kunjalo, izinto zazingcono ngakumbi kumbono wesoftware, kuba yayikukukhutshwa kweSandy Bridge apho i-Intel yazisa itekhnoloji yabo eyaziwa ngokuba yi-decoding yevidiyo isebenzisa izixhobo ze-CPU - Ulungelelwaniso oluKhawulezayo, olungqineke sesona sisombululo xa usebenza ngevidiyo. . Ukusebenza kwemidlalo ye-Intel HD Graphics, ewe, akusivumelanga ukuba sibhengeze ukuba imfuno yamakhadi evidiyo ngoku yinto yakudala, nangona kunjalo, i-Intel ngokwayo yaphawula ukuba i-GPU ixabisa i-50 yeedola okanye ngaphantsi, i-chip yabo yegraphics inokuba nayo. ube ngumkhuphisani onzima, owawungekho kude nenyaniso - ngexesha lokukhululwa, i-Intel yabonisa ukusebenza kwe-2500k ye-graphics core kwinqanaba le-HD5450 - elona khadi lemizobo ye-AMD Radeon efikelelekayo.
I-Intel Core i5 2500k ithathwa njengeyona iprosesa idumileyo. Oku akumangalisi, kuba ngenxa ye-multiplier evuliwe, i-solder phantsi kwekhava kunye nokuchithwa kobushushu obuphantsi, iye yaba yintsomi yokwenyani phakathi kwe-overclockers.
Ukusebenza kwemidlalo yeSandy Bridge kwakhona kugxininise indlela ebekwe yi-Intel kwisizukulwana sangaphambili-ukubonelela ngomsebenzi womsebenzisi ngokuhambelana nezona zisombululo zeNehalem ezixabisa i-999 yeedola. Kwaye isigebenga esiluhlaza siphumelele- ngesixa esincinci esingaphezulu kwe-300 yeedola, umsebenzisi ufumene ukusebenza ngokuthelekiswa ne-i7 980X, ebonakala ingenakucingelwa kwiinyanga nje ezintandathu ezidlulileyo. Ewe, i-horizons entsha yokusebenza ayizange yoyiswe sisizukulwana sesithathu (okanye sesibini?) se-Core processors, njengoko kwakunjalo nge-Nehalem, kodwa ukuncitshiswa okubalulekileyo kweendleko zezisombululo eziphezulu ezixabisekileyo kwenza ukuba kube nokwenzeka ukuba ube "ngabantu" ngokwenene. ukhetho.
I-Intel Core i5-2500k
Kubonakala ngathi lifikile ixesha lokuba i-AMD iqale ngoyilo lwayo olutsha, kodwa kuye kwafuneka silinde ixesha elide ukubonakala komntu okhuphisana naye ngokwenene-ngokukhululwa kweSandy Bridge, i-arsenal ye-giant ebomvu iquka kuphela iPhenom eyandisiweyo kancinane. Umgca we-II, uncediswa ngezisombululo ezisekelwe kwi-Thuban cores - i-6-core eyaziwayo ye-X1055 1090 processors kunye ne-800T. Ezi processors, ngaphandle kotshintsho oluncinci lwezakhiwo, zinokuqhayisa kuphela ngokubuya kwetekhnoloji yeTurbo Core, apho umgaqo wohlengahlengiso olugqithisiweyo lwee-cores ubuyele kuhlengahlengiso lomntu ngamnye kubo, njengoko kwakunjalo kwiPhenom yasekuqaleni. Enkosi kolu bhetyebhetye, zombini ezona ndlela zisebenzayo zinoqoqosho (ngokwehla kwimowudi yokungenzi nto ukuya kwi-500 MHz) kunye neprofayili yokusebenza enobunkunkqele (ii-cores ezingaphezulu kwe-XNUMX MHz ngaphezulu kwe-frequency yefektri) yenzeka. Ngaphandle koko, i-Thuban yayingahlukanga kubazalwana bayo abancinci kolu chungechunge, kwaye ii-cores zayo ezimbini ezongezelelweyo zasebenza ngakumbi njengeqhinga lokuthengisa i-AMD, enikezela ngee-cores ezininzi ngemali encinci.
Yeha, inani elikhulu lama-cores alizange lithethe konke ukusebenza okukhulu - kwiimvavanyo zokudlala, i-X6 1090T inqwenela ukuya kwinqanaba le-Clarkdale ephantsi, kuphela kwezinye iimeko ezicela umngeni ekusebenzeni kwe-i5 750. Ukusebenza okuphantsi ngondoqo ngamnye, I-125 W yokusetyenziswa kwamandla kunye nezinye iintsilelo zakudala ze-architecture ye-Phenom II, esekho kwi-45 nm, ayizange ivumele i-Reds ukuba ibeke ukhuphiswano olunzima kwisizukulwana sokuqala i-Core kunye nabazalwana bayo abahlaziyiweyo. Kwaye ngokukhululwa kweSandy Bridge, ukufaneleka kwe-X6 kwanyamalala, kushiyeke kunomdla kuphela kwisangqa esimxinwa sabasebenzisi babalandeli abaziingcali.
Impendulo evakalayo ye-AMD kwiimveliso ezintsha ezivela kwi-Intel ilandelwe kuphela ngo-2011, xa umgca omtsha we-AMD FX processors esekwe kwi-architecture ye-Bulldozer yaziswa. Ukukhumbula olona chungechunge luphumeleleyo lwabaqhubekekisi bayo, i-AMD ayizange ithobeke, kwaye yaphinda yagxininisa amabhongo ayo amangalisayo kunye nezicwangciso zexesha elizayo-isizukulwana esitsha sathembisa, njengangaphambili, ii-cores ezingaphezulu kwimarike yedesktop, ulwakhiwo olutsha, kwaye, kunjalo. , ukusebenza okumangalisayo kwiindidi zexabiso-kuya-ukusebenza.
Ukusuka kumbono woyilo, iBulldozer ijongeka inesibindi-ulungiselelo lwemodyuli yee-cores kwiibhloko ezine kwi-cache ye-L3 eqhelekileyo phantsi kweemeko ezifanelekileyo zenzelwe ukuqinisekisa ukusebenza kakuhle kwimisebenzi enemisonto emininzi kunye nezicelo, nangona kunjalo, ngenxa yomnqweno wokugcina ukuhambelana. kunye neqonga le-AM2 eliguga ngokukhawuleza, i-AMD yagqiba ekubeni igcine isigqubuthelo somprosesa wesilawuli sebhulorho engasentla, yenze enye yezona ngxaki zibalulekileyo kuyo kwiminyaka elandelayo.
Crystal Bulldozer
Nangona ii-cores ezi-4 zomzimba, iiprosesa ze-Bulldozer zanikezelwa kubasebenzisi njengezona zisibhozo-oku kwakungenxa yobukho bee-cores ezimbini ezinengqondo kwiyunithi nganye yekhompyutha. Nganye kuzo yazingca nge-2 MB L2 cache yayo enkulu, idikhowuda, i-256 KB buffer yemiyalelo kunye neyunithi yendawo edadayo. Olu lwahlulo lwamacandelo asebenzayo lwenza ukuba kube lula ukubonelela ngokucwangciswa kwedatha kwimicu esibhozo, kugxininise ukugxininiswa kolwakhiwo olutsha kwikamva elibonakalayo. I-Bulldozer ifumene inkxaso ye-SSE4.2 kunye ne-AESNI, kunye neyunithi enye ye-FPU ngombindi womzimba ngamnye waba nako ukwenza imiyalelo ye-256-bit ye-AVX.
Ngelishwa kwi-AMD, i-Intel sele yazisa iSandy Bridge, ngoko ke iimfuno zecandelo leprosesa ziye zanda kakhulu. Ngexabiso elingaphantsi kwe-X6 1090T, umsebenzisi oqhelekileyo unokuthenga i-i5 2500k enkulu kwaye afumane intsebenzo ngokuhambelana neyona minikelo igqwesileyo yokugqibela, kunye neeReds kufuneka zenze okufanayo. Ewe, iinyani zexesha lokukhululwa zazinombono wazo kulo mbandela.
Sele ii-cores ezi-6 zePhenom II zikhululekile kwiimeko ezininzi, singasathethi ke ngeentambo ezisibhozo ze-AMD FX - ngenxa yokucaciswa kobuninzi bemidlalo kunye nezicelo ezisebenzisa iintambo ze-1-2, ngamanye amaxesha ukuya kuthi ga kwimisonto emi-4, imveliso entsha. ukusuka kwinkampu ebomvu yajika yakhawuleza kancinci iPhenom II yangaphambili, ngokungenathemba ilahlekelwe yi-2500k. Nangona ezinye iingenelo kwimisebenzi yobuchwephesha (umzekelo, kugcino lwedatha), iflegi yeFX-8150 yajika ingenamdla kubathengi abasele bemfamekiswe ngamandla e-i5 2500k. Uvukelo aluzange lwenzeke, kwaye nembali ayizange iphinde. Kufanelekile ukukhankanya uvavanyo lwe-WinRAR olwakhelwe ngaphakathi, olwalunemisonto emininzi, ngelixa umsebenzi wokwenyani umgcini wogcino wasebenzisa ngokupheleleyo imisonto emibini kuphela.
Enye ibhulorho. Ivy Bridge okanye ngelixa ulindile
Umzekelo we-AMD wawubonisa izinto ezininzi, kodwa okokuqala kugxininise isidingo sokudala uhlobo oluthile lwesiseko sokwakha impumelelo (kuzo zonke iinkalo) ukwakhiwa kweprosesa. Yile ndlela i-AMD yaba yeyona nto ingcono kakhulu kwixesha le-K7 / K8, kwaye kwaba ngumbulelo kwi-postulates efanayo ukuba i-Intel yathatha indawo yabo ngokukhululwa kweSandy Bridge.
Ukulungiswa kwezakhiwo kwabonakala kungenamsebenzi xa i-win-win indibaniselwano ibonakala kwizandla zeBlues - ii-cores ezinamandla, i-TDP ephakathi kunye nefomathi yeqonga eqinisekisiweyo kwibhasi yeringi, ngokukhawuleza ngokukhawuleza kwaye isebenza kakuhle kuwo nawuphi na umsebenzi. Ngoku konke okwakusele kukuhlanganisa impumelelo, kusetyenziswa yonke into eyayifikile ngaphambili - kwaye le yimpumelelo kanye yokuba i-Ivy Bridge yenguqu, eyesithathu (njengoba ibango le-Intel) isizukulwana seCore processors, yaba.
Mhlawumbi olona tshintsho lubalulekileyo kumbono woyilo yayiyintshukumo ye-Intel ukuya kwi-22 nm - hayi ukutsiba, kodwa inyathelo lokuzithemba ekunciphiseni ubungakanani bokufa, okuthe kwakhona kwajika kwancinci kunokwandulelayo. Ngendlela, ubungakanani beprosesa ye-AMD FX-8150 kunye neteknoloji ye-32 nm endala yayiyi-315 mm2, ngelixa i-Intel Core i5-3570 iprosesa yayinobukhulu obungaphezu kwesiqingatha esikhulu: 133 mm2.
Ngeli xesha, i-Intel iphinde yathembela kwimizobo ekwibhodi, kwaye yabelwa indawo engaphezulu kwi-chip yayo - nangona incinci ngakumbi. Eminye i-topology ye-chip ayizange iqhube naluphi na utshintsho - iibhloko ezine ezifanayo ze-cores kunye ne-block L3 ye-cache eqhelekileyo, umlawuli wememori kunye nomlawuli we-I / O wenkqubo. Omnye unokuthi uyilo lujongeka ngokufanayo, kodwa yayiyeyona nto iphambili yeqonga le-Ivy Bridge-ukugcina okona kulungileyo kweSandy, ngelixa usongeza i-pluses kwi-treasury iyonke.
Ibhulorho yeCrystal Ivy
Ngombulelo kwinguqu kwi-teknoloji yenkqubo encinci, i-Intel yakwazi ukunciphisa ukusetyenziswa kwamandla apheleleyo kwi-processors ukuya kwi-77 W - ukusuka kwi-95 kwisizukulwana sangaphambili. Nangona kunjalo, amathemba eziphumo ezigqwesileyo ezigqwesileyo awazange athetheleleke - ngenxa yobume be-Ivy Bridge, ukufikelela kwii-frequencies eziphakamileyo kufuna i-voltages enkulu kunakwimeko yeSandy, ngoko kwakungekho ngxamiseko ethile yokuseta iirekhodi kunye nosapho lwabaqhubekisi. Kwakhona, ukutshintshwa kwe-interface ye-thermal phakathi kwesigubungelo sokusasazwa kwe-thermal yeprosesa kunye ne-chip yayo ukusuka kwi-solder ukuya kwi-paste ye-thermal yayingeyona nto ingcono kakhulu yokugqithisa.
Ngethamsanqa kubanini besizukulwana sangaphambili i-Core, i-socket ayizange itshintshe, kwaye iprosesa entsha inokufakwa ngokulula kwi-motherboard yangaphambili. Nangona kunjalo, ii-chipsets ezintsha zibonelela ngolonwabo olunje njengenkxaso ye-USB 3.0, ke abasebenzisi abalandela iinguqulelo zetekhnoloji mhlawumbi bangxamele ukuthenga ibhodi entsha kwi-Z-chipset.
Ukusebenza ngokubanzi kwe-Ivy Bridge akunyukanga kakhulu ngokwaneleyo ukuba kubizwe ngolunye uguquko, kodwa ngokuqhubekayo. Kwimisebenzi yobuchwephesha, i-3770k ibonise iziphumo ezinokuthelekiswa neeprosesa ze-X-series yobuchwephesha, kwaye kwimidlalo yayiphambi kwezinto ezazithandwa ngaphambili 2600k kunye ne-2700k kunye nomahluko malunga ne-10%. Abanye banokuthatha oku akwanelanga ukuphucula, kodwa iSandy Bridge ithathwa njengenye yezona ntsapho ziqhuba ixesha elide kwimbali ngesizathu.
Ekugqibeleni, kwanabasebenzisi bemidlalo ye-PC abanoqoqosho bakwazi ukuziva bephambili - i-Intel HD Graphics 4000 yajika yaba ngokukhawuleza kakhulu kunesizukulwana sangaphambili, ebonisa ukunyuka okuphakathi kwe-30-40%, kwaye yafumana inkxaso ye-DirectX 11. Ngoku kwakunokwenzeka ukuba udlale imidlalo edumileyo kwiindawo ezisezantsi-eziphantsi, ukufumana ukusebenza kakuhle.
Ukuyishwankathela, i-Ivy Bridge yayilulongezelelo olwamkelekileyo kusapho lwe-Intel, luthintela zonke iintlobo zemingcipheko evela kulwakhiwo olugqithileyo, kwaye ilandela umgaqo-siseko wokuba i-Blues ingaphambuki kuwo. I-Reds yenza umzamo wokuqhuba umsebenzi omkhulu kwiimpazamo ngendlela yePiledriver - isizukulwana esitsha kwisimo esidala.
I-32 nm ephelelwe lixesha ayizange ivumele i-AMD ukuba yenze olunye uguquko, ngoko ke iPiledriver yabizwa ukuba ilungise iziphene zeBulldozer, ithathele ingqalelo eyona miba ibuthathaka yoyilo lwe-AMD FX. I-Zambezi cores yatshintshwa yi-Vishera, ebandakanya ukuphuculwa kwezinye izisombululo ezisekelwe kwi-Triniti - iiprosesa zeselula ze-giant ebomvu, kodwa i-TDP yahlala ingatshintshi - i-125 W yemodeli yeflegi kunye nesalathisi se-8350. Ngokwesakhiwo, yayifana nomzalwana wayo omdala. , kodwa ukuphuculwa kwezakhiwo kunye nokunyuka kwamaxesha amaninzi nge-400 MHz kusivumele ukuba sibambe.
Izilayidi zentengiso ye-AMD ngobusuku bokukhutshwa kweBulldozer yathembisa abalandeli bolu phawu i-10-15% yokonyuka kwentsebenzo ukusuka kwisizukulwana ukuya kwesinye, kodwa ukukhutshwa kweSandy Bridge kunye nokutsibela phambili okukhulu akuzange kuvumele ukuba ezi zithembiso zibizwe ngokuba ngamabhongo kakhulu. - ngoku i-Ivy Bridge yayisele ikwishelufa, ibuyisela umva umda ongaphezulu wemveliso yomda nangakumbi. Ukuze ugweme ukwenza impazamo kwakhona, i-AMD yazisa i-Vishera njengenye indlela yohlahlo lwabiwo-mali lomgca we-Ivy Bridge - i-8350 yayichasene ne-i5-3570K, eyayingabangelwa nje kukulumkisa kwe-Reds, kodwa nakwinkampani. umgaqo-nkqubo wamaxabiso. I-Piledriver ye-flagship ifumaneke kuluntu nge-$ 199, eyenza ukuba ingabizi kakhulu kunomntu okhuphisana naye - nangona kunjalo, okufanayo akunakuthethwa ngokuqinisekileyo malunga nokusebenza.
Imisebenzi yobuchwephesha yayiyeyona ndawo iqaqambileyo yeFX-8350 ukutyhila amandla ayo - iicores zasebenza ngokukhawuleza, kwaye kwezinye iimeko imveliso entsha evela kwi-AMD yayiphambi kwe-3770k, kodwa apho uninzi lwabasebenzisi lujonge (ukusebenza kwemidlalo), iprosesa ibonise iziphumo ezifana ne-i7-920, kwaye ngokungcono ayikho kude kakhulu emva kwe-2500k. Nangona kunjalo, le meko ayizange imangalise nabani na - i-8350 yayiyi-20% enemveliso ngaphezu kwe-8150 kwimisebenzi efanayo, ngelixa i-TDP yahlala ingatshintshi. Umsebenzi wokulungisa iimpazamo ube yimpumelelo, nangona ubungekho ngokuqaqambileyo njengoko abaninzi bebenokuthanda.
Irekhodi yehlabathi yokugqithisa iprosesa ye-AMD FX 8370 yaphunyezwa yi-Finnish overclocker I-Stilt ngo-Agasti ka-2014. Wakwazi ukugqithisa ikristale ukuya kwi-8722,78 MHz.
Haswell: Kuhle kakhulu ukuba ube yinyani kwakhona
Indlela yoyilo ye-Intel, njengoko sele ibonwa, ifumene intsingiselo yayo yegolide-ukunamathela kwisikimu esisekwe kakuhle ekwakheni ulwakhiwo oluyimpumelelo, ukwenza uphuculo kuzo zonke iinkalo. ISandy Bridge yaba ngumseki woyilo olusebenzayo olusekwe kwibhasi yeringi kunye neyunithi edibeneyo engundoqo, i-Ivy Bridge yayicoca ngokwemigaqo ye-hardware kunye nonikezelo lwamandla, kwaye i-Haswell yaba luhlobo lokuqhubekeka kwangaphambi kwayo, ithembisa imigangatho emitsha yomgangatho kunye nokusebenza. .
Izilayidi zoyilo ezivela kwinkcazo ye-Intel ngobunono zabonisa ukuba uyilo luya kuhlala lungatshintshi. Uphuculo luchaphazele kuphela iinkcukacha ezithile kwifomathi yokuphucula - amazibuko amatsha ongezwa kumphathi wemisebenzi, i-L1 kunye ne-L2 cache yalungiswa, kunye ne-TLB buffer ekugqibeleni. Akunakwenzeka ukuba ungaqapheli ukuphuculwa komlawuli we-PCB, ojongene nokusebenza kwenkqubo kwiindlela ezahlukeneyo kunye neendleko zombane ezinxulumene nazo. Ukubeka nje, ekuphumleni i-Haswell iye yanoqoqosho ngakumbi kune-Ivy Bridge, kodwa kwakungekho ntetho yokuncipha ngokubanzi kwi-TDP.
Iibhodi zomama eziphucukileyo ezinenkxaso yeemodyuli ze-DDR3 ezinesantya esiphezulu zinika abantu abathandanayo ngovuyo oluthile, kodwa ukusuka kwindawo yokujonga ngaphezulu kwe-overclocking yonke into yajika yaba lusizi-iziphumo zikaHaswell zazimbi ngakumbi kunesizukulwana sangaphambili, kwaye oku kwakubangelwa ubukhulu becala kutshintsho kwi-overclocking. ezinye iindawo zokusebenzela ezishushu, ezithi kuphela ngamavila angahleki ngazo ngoku. Iigraphics ezidityanisiweyo nazo zafumana izibonelelo zokusebenza (ngenxa yokunyuka kogxininiso kwihlabathi leelaptops eziphathekayo), kodwa ngokuchasene nemvelaphi yokunqongophala kokukhula okubonakalayo kwi-IPC, uHaswell wabizwa ngokuba "Hasfail" ngenxa yokunyuka okubuhlungu kwe-5-10% ekusebenzeni xa kuthelekiswa. kwisizukulwana sangaphambili. Oku, kunye neengxaki zemveliso, kukhokelele kwinto yokuba i-Broadwell - isizukulwana esilandelayo se-Intel - yajika yaba yintsomi engekhoyo, kuba ukukhululwa kwayo kumaqonga eefowuni kunye nokunqumamisa unyaka wonke kuchaphazele kakubi imbono yomsebenzisi jikelele. Ubuncinci ukulungisa imeko, i-Intel yakhupha i-Haswell Refresh, ekwabizwa ngokuba yiDevil Canyon - nangona kunjalo, inqaku layo lonke yayikukunyusa isiseko se-Haswell processors (4770k kunye ne-4670k), ke asizukunikela icandelo elahlukileyo kuyo.
Broadwell-H: Nokuba ukonga ngakumbi, ngokukhawuleza
Ikhefu elide ekukhutshweni kwe-Broadwell-H ngenxa yobunzima obunxulumene nokutshintshwa kwenkqubo entsha yeteknoloji, nangona kunjalo, ukuba sihlolisisa uhlalutyo lwezakhiwo, kuyacaca ukuba ukusebenza kwe-Intel processors kufikelele kwinqanaba elingenakufikeleleka ngabakhuphisana nabo. ukusuka kwi-AMD. Kodwa oku akuthethi ukuba iiReds zazichitha ixesha labo-enkosi kutyalo-mali kwii-APU, izisombululo ezisekwe kwi-Kaveri zazifunwa kakhulu, kwaye iimodeli ezindala zoluhlu lwe-A8 zinokunika isiqalo sentloko kuyo nayiphi na imizobo edibeneyo evela kwiBlues. Kuyabonakala ukuba, i-Intel yayingonwabanga kwaphela yile meko-kwaye ke ngoko undoqo wemizobo ye-Iris Pro yahlala kwindawo ekhethekileyo kuyilo lwe-Broadwell-H.
Idityaniswe nenguqu ukuya kwi-14 nm, ubungakanani bokufa be-Broadwell-H buhlala bufana - kodwa ukucwangciswa okuthe kratya kusivumele ukuba sigxile ngakumbi ekwandiseni amandla emizobo. Emva koko, kwakukho kwii-laptops kunye namaziko e-multimedia apho i-Broadwell ifumene ikhaya layo lokuqala, ngoko ke izinto ezintsha ezifana nenkxaso ye-hardware decoding ye-HEVC (H.265) kunye ne-VP9 ibonakala ingaphezulu kunengqiqo.
Intel Core i7-5775C microprocessor chip
Ikristale ye-eDRAM ifanelwe ukukhankanywa ngokukodwa, ithathe indawo eyahlukileyo kwi-crystal substrate kwaye yaba luhlobo lwedatha ye-high-speed buffer - i-L4 cache - ye-processor cores. Ukusebenza okusivumeleyo ukuba sithembele kwinyathelo elibi ukuya phambili kwimisebenzi yobuchwephesha enovakalelo ngokukodwa kwisantya sokucubungula idatha egciniweyo. Umlawuli we-eDRAM uthathe indawo kwiprosesa engundoqo; iinjineli ziyisebenzise ukubuyisela indawo ethe yakhululeka emva kotshintsho lwenkqubo entsha yetekhnoloji.
I-eDRAM yadityaniswa nokukhawulezisa ukusebenza kwemizobo ekwibhodi, isebenza njengecache yefreyimu ekhawulezayo - enomthamo we-128 MB, amandla ayo anokwenza lula kakhulu umsebenzi we-GPU ekwibhodi. Ngapha koko, yayiluzuko lwekristale ye-eDRAM ukuba unobumba C wongezwa kwigama leprosesa-Intel yabiza itekhnoloji yokugcinwa kwedatha ekhawulezayo kwi-chip Crystal Wall.
Iimpawu ze-frequency zemveliso entsha, ngokungaqhelekanga ngokwaneleyo, ziye zathozama ngakumbi kune-Haswell-i-5775C endala yayine-frequency ye-3.3 GHz, kodwa kwangaxeshanye inokuqhayisa ngokuphindaphinda okuvuliwe. Ngokunciphisa ama-frequency, i-TDP nayo iyancipha - ngoku yayiyi-65 W kuphela, leyo iprosesa yeli nqanaba mhlawumbi iyona mpumelelo engcono kakhulu, kuba ukusebenza kwahlala kungatshintshi.
Ngaphandle kokuthozama kwayo (ngemigangatho yeSandy Bridge) enokubakho ngaphezulu, uBroadwell-H wothuswa kukusebenza kakuhle kwamandla, wajika waba ngowona qoqosho kwaye ubanda kakhulu phakathi kwabantu okhuphisana nabo, kwaye imizobo ekwibhodi yayiphambi kwezisombululo zosapho lwe-AMD A10, ebonisa ukuba ukubheja kungundoqo wemizobo phantsi kwe-hood kwakufanelekile.
Kubalulekile ukukhumbula ukuba i-Broadwell-H yabonakala iphakathi kangangokuba kwiinyanga ezintandathu iiprosesa ezisekelwe kuyilo lwe-Skylake zaziswa, eyaba sisizukulwana sesithandathu kwintsapho yaseCore.
I-Skylake - Ixesha le-revolutions liphelile kudala
Okumangalisayo kukuba, izizukulwana ezininzi zidlulile ukusukela kwiSandy Bridge, kodwa akukho namnye kubo okwazileyo ukothusa uluntu ngento engakholelekiyo kunye nentsha, ngaphandle, mhlawumbi, yeBroadwell-H - kodwa apho bekungaphezulu malunga nokutsiba okungazange kubonwe kwimizobo. kunye nokusebenza kwayo (xa kuthelekiswa nee-APU ze-AMD), kunokuba malunga nenkqubela phambili enkulu ekusebenzeni. Iintsuku zeNehalem ngokuqinisekileyo zihambile kwaye aziyi kubuya, kodwa i-Intel yaqhubeka iqhubela phambili ngamanyathelo amancinci.
Uyilo lwezakhiwo, i-Skylake yahlengahlengiswa, kwaye ulungelelwaniso oluthe tyaba lweeyunithi zekhompuyutha zathatyathelwa indawo yi-classics square layout, apho ii-cores zahlulwe yi-cache ye-LLC ekwabelwana ngayo, kwaye undoqo wegraphics onamandla ubekwe ngasekhohlo.
Intel Core i7-6700k microprocessor chip
Ngenxa yeempawu zobugcisa, umlawuli we-eDRAM ngoku ubekwe kwindawo yeyunithi yokulawula ye-I / O njengokongezwa kwimodyuli yokulawula ukukhutshwa komfanekiso ukwenzela ukubonelela ngeyona ndlela ingcono yokuhanjiswa komfanekiso ukusuka kwi-core graphics core. Isilawuli sombane esakhelwe ngaphakathi esisetyenziswe eHaswell sanyamalala phantsi kwekhava, ibhasi ye-DMI yahlaziywa, kwaye ngenxa yomgaqo wokubuyela ngasemva, iiprosesa zeSkylake zixhase zombini i-DDR4 kunye ne-DDR3 inkumbulo- umgangatho omtsha we-SO-DIMM DDR3L wenzelwe bona. , isebenza kumbane ophantsi .
Kwangaxeshanye, umntu akanakunceda kodwa aqaphele ukuba i-Intel ihlawula kangakanani na ekupapasheni isizukulwana esilandelayo semizobo ekwibhodi-kwimeko yeSkylake, yayisele ingowesithandathu kumgca oluhlaza. I-Intel inebhongo ngokukodwa ngokunyuka kwentsebenzo, eyayibaluleke kakhulu kwimeko ye-Broadwell, kodwa ngeli xesha ithembisa ngokukodwa abadlali be-budget-conscious games inqanaba eliphezulu lokusebenza kunye nenkxaso yazo zonke ii-API zanamhlanje, kuquka i-DirectX 12. I-subsystem yegraphics iyingxenye. yento ebizwa ngokuba yi-System on Chip (SOC) , leyo i-Intel nayo yakhuthaza ngokunyanisekileyo njengomzekelo wesisombululo esiyimpumelelo sokwakha. Kodwa ukuba uyakhumbula ukuba isilawuli sombane esidityanisiweyo sinyamalele, kwaye inkqubo engaphantsi kwamandla ixhomekeke ngokupheleleyo kwiVRM yebhodi yomama, ewe, iSkylake ayikafiki kwiSOC epheleleyo. Akukho ntetho konke konke malunga nokudibanisa ibhulorho esezantsi phantsi kwekhava.
Nangona kunjalo, i-SOC apha idlala indima yomlamli, uhlobo "lwebhulorho" phakathi kwe-chip ye-graphics ye-Gen9, i-processor cores kunye nomlawuli we-I / O wenkqubo, ojongene nokusebenzisana kwamacandelo kunye neprosesa kunye nokucubungula idatha. Kwangelo xesha, i-Intel yabeka ugxininiso olubalulekileyo ekusebenzeni kwamandla kunye namanyathelo amaninzi athathwe yi-Intel kumlo wokusebenzisa iiwatts ezimbalwa-i-Skylake ibonelela "ngamasango ombane" ahlukeneyo (masiwabize ngokuba ngamandla) kwicandelo ngalinye le-SOC, kuquka ibhasi yeringi enesantya esiphezulu, inkqubo engaphantsi kwemizobo kunye nomlawuli wemidiya. Inkqubo yangaphambili yolawulo lwesigaba seprosesa esekwe kwiprosesa iguqukele kwitekhnoloji yeSpeed Shift, ebonelela ngokutshintsha okuguquguqukayo phakathi kwezigaba ezahlukeneyo (umzekelo, xa uvuka kwimowudi yokulala ngexesha lomsebenzi osebenzayo okanye uqalisa umdlalo onzima emva kokusefa okulula. ) kunye nokulinganisa iindleko zamandla phakathi kweeyunithi ze-CPU ezisebenzayo ukufezekisa ukusebenza kakuhle kakhulu ngaphakathi kwe-TDP.
Ngenxa yohlengahlengiso olunxulumene nokunyamalala kwesilawuli samandla, i-Intel yanyanzelwa ukuba isuse i-Skylake kwi-socket entsha ye-LGA1151, apho kwakhululwa iibhodi ze-motherboard ezisekelwe kwi-chipset ye-Z170, eyafumana inkxaso ye-20 PCI-E 3.0 imizila, enye i-USB 3.1 Uhlobo lwe-A port, inani elongezelelweyo le-USB 3.0 port, inkxaso ye-eSATA kunye ne-M2 drives. Imemori yaxelwa ukuba ixhase iimodyuli zeDDR4 ezinamaza ukuya kuthi ga kwi-3400 MHz.
Ngokuphathelele ukusebenza, ukukhutshwa kwe-Skylake akuzange kuphawule nayiphi na into eyothusayo. Ukonyuka komsebenzi okulindelekileyo ngeepesenti ezintlanu xa kuthelekiswa noDevil Canyon kushiye abalandeli abaninzi bedidekile, kodwa kwacaca kwi-Intel presentation slides ukuba ugxininiso oluphambili lwalusekusebenzeni kwamandla kunye nokuguquguquka kweqonga elitsha, elikwaziyo ukufaneleka kuzo zombini iincinci ezinexabiso eliphantsi. -Iinkqubo ze-ITX kunye kunye namaqonga okudlala aphambili. Abasebenzisi abalindele ukutsibela phambili ukusuka eSandy Bridge Skylake baye baphoxeka; imeko yayisikhumbuza ukukhululwa kweHaswell; ukukhutshwa kwesokethi entsha nako kuyadanisa.
Ngoku lixesha lokuthemba iKaby Lake, kuba umntu, kwaye wayefanele ukuba nguye ...
Ichibi laseKaby. Ichibi elitsha kunye nobubomvu obungalindelekanga
Nangona ingqiqo yokuqala yesicwangciso se-"tick-tock", i-Intel, iqonda ukungabikho naluphi na ukhuphiswano oluvela kwi-AMD, yagqiba ekubeni yandise umjikelo ngamnye ukuya kwizigaba ezintathu, apho, emva kokuqaliswa koyilo olutsha, isisombululo esikhoyo sihlanjululwa phantsi. igama elitsha leminyaka emibini ezayo. Isinyathelo se-14 nm sasiyi-Broadwell, ilandelwa yi-Skylake, kunye ne-Kaby Lake, ngokufanelekileyo, yenzelwe ukubonisa inqanaba eliphezulu lezobuchwepheshe xa kuthelekiswa ne-Nebesnozersk yangaphambili.
Umahluko omkhulu phakathi kweKaby Lake kunye ne-Skylake kukunyuka kwee-frequencies nge-200-300 MHz - zombini ngokwemiqathango yesiseko kunye nokunyusa. Ngokwezakhiwo, isizukulwana esitsha asizange sifumane naluphi na utshintsho - kunye nemizobo edibeneyo, nangona uhlaziyo lweempawu, zahlala zifana, kodwa i-Intel yakhupha i-chipset esekelwe kwi-Z270 entsha, eyongeze i-4 PCI-E 3.0 imizila ekusebenzeni kwangaphambili. I-Sunrise Point, kunye nenkxaso ye-Intel iteknoloji ye-Optane Memory yezixhobo eziphambili zesigebenga. Ukuphindaphinda okuzimeleyo kumacandelo ebhodi kunye nezinye iimpawu zeplatifomu yangaphambili zigcinwe, kwaye izicelo ze-multimedia zifumene umsebenzi we-AVX Offset, ovumela ukunciphisa i-processor frequencies xa kusetyenzwa imiyalelo ye-AVX ukunyusa ukuzinza kwi-frequencies ephezulu.
Intel Core i7-7700k microprocessor chip
Ngokubhekiselele ekusebenzeni, iimveliso ze-Core zesizukulwana sesixhenxe okokuqala ziphantse zafana nabangaphambi kwazo - ziphinde zanikela ingqalelo ekwandiseni ukusetyenziswa kwamandla, i-Intel yalibala ngokupheleleyo ngezinto ezintsha malunga ne-IPC. Nangona kunjalo, ngokungafaniyo ne-Skylake, imveliso entsha yayisombulula ingxaki yokufudumeza okugqithisileyo kumanqanaba okugqithisa kakhulu, kwaye yenza ukuba ivakale ngathi ngeentsuku zeSandy Bridge, iprosesa engaphezulu kwe-4.8-4.9 GHz ngokusetyenziswa kwamandla okuphakathi kunye namaqondo okushisa aphantsi. Ngamanye amazwi, i-overclocking iye yaba lula, kwaye iprosesa iye yaba yi-10-15 degrees epholileyo, enokuthiwa yimiphumo yolo kulungelelaniswa kakhulu, umjikelo wayo wokugqibela.
Akukho mntu wayenokuqikelela ukuba i-AMD yayisele ilungiselela impendulo yokwenyani kwiminyaka emininzi yophuhliso lwe-Intel. Igama layo yi-AMD Ryzen.
I-AMD Ryzen-Xa wonke umntu ehleka kwaye akukho mntu wayekholelwa
Emva kohlaziyo lweBulldozer, uyilo lwePiledriver lwaziswa ngo-2012, i-AMD yafudukela ngokupheleleyo kwezinye iindawo zentengiso yeprosesa, ikhupha imigca eliqela ye-APU eyimpumelelo, kunye nezinye izisombululo ezinoqoqosho neziphathwayo. Nangona kunjalo, inkampani ayizange ilibale malunga nomlo ohlaziyiweyo wendawo elangeni kwiikhomputha zedeskithophu, ezenza ubuthathaka, kodwa kwangaxeshanye isebenza kuyilo lweZen - isisombululo esitsha sokwenene esenzelwe ukuvuselela umoya olahlekileyo wokhuphiswano kwi-CPU. imarike.
Ukuphuhlisa imveliso entsha, i-AMD yaphendukela kuncedo lukaJim Keller, "uyise wee-cores ezimbini" onamava akhe omsebenzi akhokelela ekubeni i-giant ebomvu ibe nodumo kunye nokuqatshelwa kwi-2000 yokuqala. Yayinguye lowo, kunye nezinye iinjineli, baphuhlisa ulwakhiwo olutsha oluyilelwe ukuba lukhawuleze, lube namandla kwaye lube nentsha. Ngelishwa, wonke umntu wakhumbula ukuba iBulldozer yayisekelwe kwimigaqo efanayo - indlela eyahlukileyo yayifuneka.
UJim Keller
Kwaye i-AMD yathatha ithuba lokuthengisa, ibhengeza ukunyuka kwe-52% kwi-IPC xa kuthelekiswa nesizukulwana se-Excavator - eyona core yamva nje ikhule kwiBulldozer efanayo. Oku kuthetha ukuba xa kuthelekiswa ne-8150, abaqhubekisi beZen bathembisa ukuba baya kuba ngaphezulu kwe-60% ngokukhawuleza, kwaye oku kwamchukumisa wonke umntu. Ekuqaleni, kwiinkcazo-ntetho ze-AMD bachitha ixesha kuphela kwimisebenzi yobuchwephesha, bethelekisa iprosesa yabo entsha kunye ne-5930K, kwaye kamva kunye ne-6800K, kodwa ekuhambeni kwexesha baqala ukuthetha malunga necala lokudlala lengxaki - eyona nto icinezela kakhulu kwindawo yokuthengisa. yembono. Kodwa nalapha i-AMD yayikulungele ukulwa.
Uyilo lweZen lusekwe kubuchwephesha benkqubo entsha ye-14 nm, kunye noyilo, iimveliso ezitsha azifani kwaphela nolwakhiwo lweemodyuli ukusuka ngo-2011. ube ne cores ezine ezisebenzayo . Njengakwimeko ye-Skylake, abalawuli benkqubo abahlukeneyo babekwe kwi-chip substrate, kubandakanywa neendlela ezingama-24 ze-PCI-E 3.0, inkxaso ukuya kuthi ga kwi-4 USB 3.1 Udidi lwezibuko lwe-A, kunye nesilawuli sememori ye-DDR4. Kubaluleke kakhulu ukuba uqaphele ubukhulu be-cache ye-L3 - kwiisombululo ze-flagship umthamo wayo ufikelela kwi-16 MB. Undoqo ngamnye wafumana iyunithi yayo edadayo (FPU), eyasombulula enye yeengxaki eziphambili kuyilo lwangaphambili. Ukusetyenziswa kweprosesa kuye kwehle kakhulu - kwi-flagship ye-Ryzen 7 1800X yonyulwa kwi-95 W xa kuthelekiswa ne-220 W "yeyona ishushu" (ngendlela yonke) imodeli ye-AMD FX.
I-AMD Ryzen 1800X microprocessor iyafa
Ukuzaliswa kwetekhnoloji kuye kwabonakala kungabi sisityebi kancinci kwizinto ezintsha - ke abaqhubekekisi abatsha be-AMD bafumana iseti epheleleyo yetekhnoloji entsha phantsi kwesihloko esithi SenseMI, ebandakanya i-Smart Prefetch (ukulayisha idatha kwi-cache buffer ukukhawulezisa ukusebenza kweenkqubo), Amandla acocekileyo (ngokubalulekileyo i-analogue yonikezelo lwamandla "obukrelekrele" wolawulo lweprosesa kunye namacandelo ayo, aphunyezwe kwi-Skylake), i-Neural Net Prediction (i-algorithm esebenza kwimigaqo ye-self-learning neural network), kunye ne-Frequency Extendency. Uluhlu (okanye i-XFR), eyenzelwe ukubonelela abasebenzisi ngeenkqubo zokupholisa eziphucukileyo kunye ne-100 MHz eyongezelelweyo. Ngexesha lokuqala ukusukela kwiPiledriver, i-overclocking ayiqhutywanga yiTurbo Core, kodwa nge-Precision Boost - itekhnoloji ehlaziyiweyo yokunyusa ubuninzi ngokuxhomekeke kumthwalo kwii-cores. Sibone itekhnoloji efanayo evela kwi-Intel ukusukela kwiSandy Bridge.
Uyilo olutsha lweRyzen lusekwe kwibhasi ye-Infinity Fabric, eyenzelwe ukudibanisa zombini ii-cores kunye neebhloko ezimbini zeCCX kwi-chip substrate. I-interface enesantya esiphezulu yenzelwe ukuqinisekisa intsebenziswano ekhawulezayo enokwenzeka phakathi kwee-cores kunye neebhloko, kwaye ikwazi ukuphunyezwa kwamanye amaqonga - umzekelo, kwii-APUs zoqoqosho kunye nakumakhadi emizobo ye-AMD VEGA, apho ibhasi idityaniswe nememori ye-HBM2. kufuneka isebenze nge-bandwidth ubuncinane ye-512 Gb / s
Infinity Fabric
Konke oku kudityaniswe nezicwangciso zamabhongo zokwandisa umgca weZen ukuya kumaqonga asebenza kakhulu, iiseva kunye nee-APU - ukumanyana kwenkqubo yemveliso, njengesiqhelo, kukhokelela kwimveliso engabizi kakhulu, kwaye amaxabiso aphantsi ahendayo ebesoloko elilungelo le-AMD.
Ekuqaleni, i-AMD ibonise kuphela i-Ryzen 7 - iimodeli ezindala zomgca, ezijoliswe kubasebenzisi abakhethiweyo kunye nabenzi beendaba, kwaye kwiinyanga ezimbalwa kamva balandelwa nguRyzen 5 kunye neRyzen 3. KwakuyiRyzen 5 eyathi yavela yaba ezona zisombululo zikhangayo ngokubhekiselele kwixabiso kunye nokusebenza kwemidlalo, apho i-Intel, ngokunyanisekileyo, yayingakulungele konke. Kwaye ukuba kwinqanaba lokuqala kwakubonakala ngathi i-Ryzen imiselwe ukuphinda isiphelo seBulldozer (nangona inqanaba elincinane ledrama), emva kwexesha kwacaca ukuba i-AMD yakwazi ukunyanzelisa ukhuphiswano kwakhona.
Ezona ngxaki ziphambili ngeRyzen yayiziimpawu zobuchwephesha ezihamba nabanini bohlaziyo lwangaphambili kwiinyanga ezimbalwa zokuqala - ngenxa yeengxaki zememori, uRyzen akazange akhawuleze ukucetyiswa ukuba athenge, kunye nokuxhomekeka kweeprosesa kubuninzi be-RAM. ibonise imfuneko yeendleko ezongezelelweyo. Nangona kunjalo, abasebenzisi abanamava kuseto lwexesha bafumanise ukuba ngeemodyuli zememori enesantya esiphezulu ezilungiselelwe ukuba zibe namaxesha amancinci, iRyzen iyakwazi ukutyhala nokuba yi-7700k, ebangele ulonwabo lokwenyani kwinkampu yabalandeli be-AMD. Kodwa nangaphandle kolonwabo olunjalo, usapho lwe-Ryzen 5 lweeprosesa luye lwaphumelela kangangokuba i-wave yentengiso yabo yanyanzela i-Intel ukuba yenze uguquko olungxamisekileyo kuyilo lwayo. Impendulo kwintshukumo eyimpumelelo ye-AMD yaba kukukhutshwa kwangoku (ngexesha lokubhala) i-architecture yeCoffee Lake, eyafumana ii-cores ezi-6 endaweni yezine.
IChibi leKofu. Umkhenkce wophukile
Ngaphandle kwenyani yokuba i-7700k ibambe iwonga leyona projekthi ibalaseleyo yokudlala ixesha elide, i-AMD ikwazile ukufumana impumelelo engakholelekiyo kumbindi womgca, iphumeza owona mgaqo mdala “weecores ezingaphezulu, kodwa ezinexabiso eliphantsi.” I-Ryzen 1600 yayinee-cores ezi-6 kunye neentambo ze-12, kwaye i-7600k yayisabambekile kwii-cores ezi-4, inika i-AMD impumelelo elula yokuthengisa, ngakumbi ngenkxaso yabahlaziyi abaninzi kunye neeblogi. Emva koko i-Intel yatshintsha ishedyuli yokukhupha kwaye yazisa iLake yeKofi kwimarike- hayi nje esinye isibini seepesenti kunye neewatts ezimbalwa, kodwa inyathelo lokwenyani eliya phambili.
Enyanisweni, nalapha kwenziwa ngokubhukisha. Ii-cores ezintandathu ezilindelwe ixesha elide, ngaphandle kolonwabo lwe-SMT, zivele ngokwenene kwisiseko se-Skylake efanayo, eyakhiwe kwi-14 nm. E-Kaby Lake, isiseko sayo salungiswa, ukuxazulula iingxaki ngokugqithiswa kwe-overclocking kunye nokushisa, kunye ne-Coffee Lake yaphuculwa ukwandisa inani leebhloko eziphambili nge-2, kwaye ilungiselelwe ukusebenza okupholileyo kunye nokuzinza ngakumbi. Ukuba sivavanya i-architecture ukusuka kwindawo yokujonga izinto ezintsha, ngoko akukho tshintsho (ngaphandle kokunyuka kwenani lama-cores) luvele kwiLake Lake.
Intel Core i7-8700k microprocessor chip
Kodwa bekukho imida yobuchwephesha eyayanyaniswa nesidingo seebhodi zomama ezintsha ezisekwe kwiZ370. Ezi zithintelo zinxulunyaniswa nokunyuka kweemfuno zamandla, okoko kongezwe ii-cores ezintandathu kunye nohlengahlengiso lwenkqubo kuthathelwa ingqalelo ukukhula kobudla-kudla bekristale obufunekayo ukuphakamisa amanqanaba ombane obonelelo oluncinci. Njengoko sikhumbula kwimbali ye-Broadwell, i-Intel ibizama kwiminyaka yakutshanje ukwenza okwahlukileyo - ukunciphisa uxinzelelo kuyo yonke imida, kodwa ngoku esi sicwangciso sifikelele esiphelweni. Ngokobuchwephesha, i-LGA1151 yahlala ifana, nangona kunjalo, ngenxa yomngcipheko wokonakalisa isilawuli se-VRM, i-Intel yanciphisa ukuhambelana kweprosesa kunye neebhodi zangaphambili zangaphambili, ngaloo ndlela izikhusela kwi-scandals enokwenzeka (njengoko kwakunjalo nge-RX480 kunye ne-PCI etshisiweyo ye-AMD. -E izihlanganisi). I-Z370 ehlaziyiweyo ayisayixhasi imemori ye-DDR3L yangaphambili, kodwa akukho mntu wayelindele ukuhambelana okunjalo.
I-Intel ngokwabo babelungiselela inguqu ehlaziyiweyo yeqonga ngenkxaso ye-USB 3.1 yesizukulwana sesibini, iimemori khadi ze-SDXC kunye nomlawuli we-Wi-Fi 802.11 eyakhelwe-ngaphakathi, ngoko ke ukukhutshwa ngokukhawuleza kunye ne-Z370 kwajika kwaba yenye yezo ziganeko. yenza kwaba nokwenzeka ukufikelela kwizigqibo ngenkangeleko yeqonga. Nangona kunjalo, bekukho izinto ezothusayo ezininzi kwiLake Lake-kwaye inxalenye yazo yayigxile ekugqityweni.
I-Intel inike ingqwalasela eninzi kuyo, igxininisa umsebenzi owenziweyo wokwandisa inkqubo yokugqithiswa-umzekelo, kwiLake yeKofi kuye kwenzeka ukuba kuqwalaselwe iiseti ezininzi zamanyathelo-nge-nyathelo ngaphezulu kweemeko ezahlukeneyo zokulayisha ezingundoqo, ukukwazi ukutshintsha imemori ngokuguquguqukayo. amaxesha ngaphandle kokushiya inkqubo yokusebenza, inkxaso yayo nayiphi na, nkqu nezona ziphindaphindayo zeDDR4 ezingenakwenzeka (inkxaso echaziweyo yeefrikhwensi ukuya kuthi ga kwi-8400 MHz), kunye nenkqubo yamandla eyongeziweyo eyenzelwe imithwalo ephezulu. Nangona kunjalo, enyanisweni, ukugqithisa i-8700k kwakukude kakhulu kuneyona nto ingakholelekiyo - ngenxa yokungafezeki kwe-interface ye-thermal esetyenzisiweyo ngaphandle kokuphazamiseka, iprosesa yayihlala inqunyelwe kwi-4.7-4.8 GHz, ifikelela kumaqondo obushushu agqithisileyo, kodwa ngotshintsho kwi-interface inoko. bonisa iirekhodi ezintsha kwisitayile se-5.2 okanye nakwi-5.3 GHz. Nangona kunjalo, uninzi lwabasebenzisi aluzange lube namdla koku, ngoko ke amandla okugqithisa kwiLake yeKofi engundoqo emithandathu inokubizwa ngokuba ithintelwe. Ewe, ewe, uSandy akakalityalwa.
Umdlalo wokudlala weCoffee Lake awuzange ubonise imimangaliso ekhethekileyo - nangona ukubonakala kwee-cores ezimbini zomzimba kunye nemicu emine, i-8700k ngexesha lokukhululwa yayinenyathelo elifanayo lentsebenzo ye-5-10% ngaphezu kwe-flagship yangaphambili. Ewe, iRyzen ayikwazanga ukukhuphisana nayo kwi-niche yokudlala, kodwa ukusuka kwindawo yokujonga ukuphuculwa kwezakhiwo, kuyavela ukuba iLake Lake lelinye nje "langoku", kodwa hayi "ithiki", eyayikho iSandy Bridge ngo-2011. .
Ngethamsanqa kubalandeli be-AMD, emva kokukhululwa kwe-Ryzen, inkampani yabhengeza izicwangciso zexesha elide ze-AM4 socket kunye nophuhliso loyilo lweZen de kube ngu-2020- kwaye emva kokuba iLake Lake libuyisele ingqalelo kwicandelo le-Intel eliphakathi, lalifikile ixesha. kwi-Ryzen 2 - emva kwayo yonke loo nto, i-AMD kufuneka ibe neyakhe "yangoku".
Inyaniso ekhohlakeleyoAsinakubona i-Intel njengoko injalo namhlanje ukuba ayizange isebenzise ukhuphiswano olungalunganga ukukhuthaza iimveliso zayo. Ke ngoMeyi 2009, inkampani yahlawuliswa yiKomishini yaseYurophu isixa esikhulu se-1,5 yebhiliyoni yeedola zaseMelika ngokunyoba abavelisi bekhompyuter kunye nenkampani enye yokurhweba ngokukhetha iiprosesa kwi-Intel. Ulawulo lwe-Intel ke lwathi akukho basebenzisi abanokuthenga iikhompyuter ngexabiso eliphantsi okanye ubulungisa abazukuzuza kwisigqibo sokufaka isimangalo.
I-Intel nayo indala kunye nendlela esebenzayo yokhuphiswano. Ngokuquka umyalelo we-CPUID okokuqala, ngokuqala ngeeprosesa ze-i486, kunye nokudala kunye nokusabalalisa i-compiler yayo yamahhala, i-Intel yaqinisekisa impumelelo yayo kwiminyaka emininzi ezayo. Lo mqokeleli uvelisa ikhowudi efanelekileyo yeeprosesa ze-Intel kunye nekhowudi ephakathi kuzo zonke ezinye iiprosesa. Ke, kwaneprosesa enamandla yobuchwephesha evela kubakhuphisana “yadlula” kumasebe enkqubo engowona ulungileyo. Oku kunciphise ukusebenza kokugqibela kwisicelo kwaye akuzange kuvumele ukuba ibonise malunga nenqanaba elifanayo lokusebenza njengeprosesa ye-Intel eneempawu ezifanayo.
Kwiimeko ezinjalo zokukhuphisana, i-VIA ayikwazanga ukumelana nokhuphiswano, ukunciphisa ngokukhawuleza ukuthengiswa kweeprosesa. Iprosesa yayo yeNano eyonga amandla yayingaphantsi kwiprosesa entsha ye-Intel Atom. Yonke into yayiya kuba ilungile ukuba omnye umphandi onobuchule, uAgner Fog, wayesilele ukutshintsha i-CPUID kwiprosesa yeNano. Njengoko bekulindelekile, imveliso yanda kwaye yadlula leyo yomntu okhuphisana naye. Kodwa iindaba azizange zivelise umphumo webhombu yolwazi.
Ukhuphiswano kunye ne-AMD (umvelisi wesibini omkhulu we-x86/x64 microprocessors kwihlabathi) nalo aluzange luhambe kakuhle kolu lwamva; ngo-2008, ngenxa yeengxaki zemali, i-AMD kwafuneka ihlukane nomenzi wayo weesekethe ezidibeneyo ze-semiconductor, i-GlobalFoundries. I-AMD, ekulweni kwayo ne-Intel, ixhomekeke kwi-multi-cores, inikezela ngeeprosesa ezifikelelekayo kunye neentsimbi ezininzi, ngelixa i-Intel inokuphendula kule mveliso yemveliso kunye neeprosesa ezinama-cores ambalwa, kodwa nge-teknoloji ye-Hyper-Threading.
Kwiminyaka emininzi, i-Intel ibisonyusa isabelo sayo sentengiso kwiiprosesa zeselfowuni kunye nedesktop, isusa indawo ekhuphisana nayo. Imarike yeprosesa yeseva sele iphantse yabanjwa ngokupheleleyo. Kwaye kutshanje imeko yaqalisa ukutshintsha. Ukukhutshwa kweeprosesa ze-AMD Ryzen kunyanzelise i-Intel ukuba itshintshe amaqhinga ayo asisiseko okunyusa kancinane ukusebenza kweeprosesa. Nangona iipakethe zovavanyo zanceda i-Intel ukuba ingakhathazeki kwakhona. Umzekelo, kwiimvavanyo zokwenziwa zeSYSMark, umahluko phakathi kwesizukulwana sesithandathu kunye nesesixhenxe seCore i7 desktop processors ubungahambelani nokonyuka kwamaza kunye neempawu ezingundoqo ezifanayo.
Kodwa ngoku i-Intel sele iqalisile ukonyusa inani lee-cores ze-desktop processors, kwaye iphinde yavelisa ngokuyinxenye iimodeli zeprosesa ezikhoyo. Eli linyathelo elilungileyo elibhekisa kubathengi bayo ukuba babe nolwazi ngobugcisa.
Umbhali wenqaku nguPavel Chudinov.
2019-iBlue Point yokungabuyi okanye iChiplet Revolution
Emva kwezizukulwana ezibini eziphumelele kakhulu zeeprosesa zeRyzen, i-AMD yayikulungele ukuthatha inyathelo elingazange libonwe ngaphambili kungekuphela nje ekusebenzeni, kodwa nakwitekhnoloji yamva nje yokuvelisa-ukufudukela kwitekhnoloji yenkqubo ye-7nm, ibonelela ngokunyuka kwe-25% ekusebenzeni ngelixa igcina iphakheji eshushu. , kunye nophuhliso oluninzi lwezakhiwo kunye nokulungiswa kwenza kube lula ukuthatha iqonga le-AM4 kwinqanaba elitsha, ukubonelela bonke abanikazi beenkqubo "ezidumileyo" zangaphambili kunye nokuphuculwa okungenabuhlungu kunye nohlaziyo lokuqala lwe-BIOS.
Kwaye uphawu olubalulekileyo lwengqondo lwe-4 GHz, oluye ngeendlela ezininzi lube sisikhubekiso kwindlela eya kukhuphiswano oluqatha kunye ne-Intel, abanomdla abaxhalabileyo ngendlela eyahlukileyo - ekubeni kwavela amahemuhemu okuqala, abaninzi baqaphela ngokufanelekileyo ukuba ukwanda kwe-frequency kwi-Ryzen 3000. usapho alunakwenzeka ukuba lube ngaphezu kwe-20%, kodwa akukho mntu unokuyeka ukuphupha malunga ne-5 GHz eyaboniswa yi-Intel. "Ukuvuza" okuninzi kuye kwavuselela umdla, kunye nemigca yeprosesa epheleleyo kunye neenkcukacha ezimangalisayo, uninzi lwazo lwajika lwakude nenyani. Kodwa ngobulungisa, kubalulekile ukuqaphela ukuba okunye ukuvuza bekuhambelana neziphumo eziboniweyo - ewe, kunye nogcino oluthile.
Ngokobuchwephesha, i-architecture ye-Zen 2 ifumene inani leeyantlukwano ezinzulu ukusuka kwangaphambi kwayo, ephantsi kwezizukulwana ezibini zokuqala zeRyzen. Umahluko obalulekileyo wawuyi-layout of the processor, ngoku equlethwe ngamakristali amathathu ahlukeneyo, ezimbini zazo ziqulethe iibhloko ze-cores, kwaye okwesithathu, okumangalisayo ngobukhulu, kubandakanya ibhloko yabalawuli kunye neziteshi zonxibelelwano (I / O). Ngaphandle kwazo zonke izibonelelo ezininzi ze-7nm eyonga amandla kunye nenkqubo ephucukileyo, i-AMD ayikwazanga ukukunceda kodwa ijongane neendleko zokuvelisa ngokubonakalayo, kuba inkqubo ye-7nm yayingekavavanywa kwaye iziswe kumlinganiselo ofanelekileyo weetshiphusi ezineziphene kwezo zicocekileyo. Nangona kunjalo, bekukho esinye isizathu - ukuhlanganiswa ngokubanzi kwemveliso, okwenza kube lula ukudibanisa imigca yemveliso eyahlukeneyo ibe nye, kwaye ukhethe iikristale kuzo zombini iRyzen 5 efikelelekayo kunye ne-EPYC engakholelekiyo. Esi sisombululo singabizi kakhulu sivumele i-AMD ukuba igcine amaxabiso kwinqanaba elifanayo, kwaye kwakumnandi ukukholisa abalandeli ngokukhululwa kwe-Ryzen 3000.
Uyilo lwesakhiwo se-chiplets
Ukwahlula i-chip processor ibe ngamacandelwana amathathu amancinci avumele inkqubela phambili ebalulekileyo ekusombululeni eyona misebenzi ibalulekileyo ejongene neenjineli ze-AMD - ukunciphisa i-Infinity Fabric latency, ukulibaziseka ekufikeleleni kwi-cache kunye nokutshintshiselana kwedatha kwiibhloko ezahlukeneyo zeCCX. Ngoku ubungakanani be-cache buphindwe kabini ubuncinane (32 MB L3 kwi-3600 xa kuthelekiswa ne-16 MB kunyaka odlulileyo we-2600), iindlela zokusebenza kunye nazo ziye zaphuculwa, kwaye i-Infinity Fabric frequency ine-multiplier yayo ye-FCLK, evumela ukusetyenziswa I-RAM ukuya kwi-3733 MHz eneziphumo ezifanelekileyo (ukulibaziseka kule meko akuzange kudlule i-65-70 nanoseconds). Nangona kunjalo, i-Ryzen 3000 isenovelwano kumaxesha enkumbulo, kwaye izinti ezibizayo eziphantsi-latency zinokuzisa abanini bezixhobo ezitsha ukuya kuthi ga kwi-30% okanye ngaphezulu kokunyusa ukusebenza- ngakumbi kwiimeko ezithile kunye nemidlalo.
Iphakheji ye-thermal ye-processors yahlala ifana, kodwa i-frequencies yanda njengoko bekulindelekile - ukusuka kwi-4,2 ekunyuseni kwi-3600 ukuya kwi-4,7 kwi-3950X. Emva kokungena kwimarike, abaninzi abasebenzisi badibana nengxaki ye "malaise", xa iprosesa ingabonakali i-frequencies echazwe ngumenzi naphantsi kweemeko ezifanelekileyo - "obomvu" kufuneka kuphunyezwe ukuhlaziywa okukhethekileyo kwe-BIOS (1.0.0.3ABBA), apho ingxaki yalungiswa ngempumelelo, kwaye kwinyanga edlulileyo i-Global 1.0.0.4 yakhululwa, iqulethe izilungiso ezingaphezulu kwekhulu elinesiqingatha kunye nokulungiswa - kubasebenzisi abathile, emva kohlaziyo, i-frequency processor yanda ukuya kwi-75 MHz, kunye nomgangatho. amandla ombane ehle kakhulu. Nangona kunjalo, oku akuzange kuchaphazele amandla okugqithisa ngayo nayiphi na indlela-i-Ryzen 3000, njengabanduleli bayo, isebenza kakuhle ngaphandle kwebhokisi, kwaye ayikwazi ukubonelela nge-overclocking enokubakho ngaphaya kokunyuka okungokomfuziselo - oku kuyenza idikwe kubantu abayithandayo, kodwa kuninzi. yovuyo kwabo Kutheni engafuni ukubamba useto kwi-BIOS?
I-Zen 2 ifumene ukonyuka okubalulekileyo kwintsebenzo engundoqo (ukuya kuthi ga kwi-15% kwizicelo ezahlukeneyo), ivumele i-AMD ukuba ikhulise ngokumandla umthamo kuwo onke amacandelo emarike, kwaye okokuqala ngqa kumashumi eminyaka iguqule umsinga ngokuthanda kwayo. Yintoni eyabangela ukuba oku kwenzeke? Makhe sihlolisise.
I-Ryzen 3 - Ingcamango yeTekhnoloji
Abaninzi abalandela ukuvuza malunga nesizukulwana se-Zen 2 babenomdla ngakumbi kwi-Ryzen 3 entsha. Iiprosesa ezifumanekayo zathenjiswa ii-cores ezi-6, iigraphics ezinamandla ezidibeneyo kunye nexabiso elihlekisayo. Ngelishwa, abaphumeleleyo abalindelekileyo kwi-Ryzen 3, apho i-AMD ixhobise inxalenye ephantsi yeqonga layo ngo-2017, ayizange ibone ukukhanya kwemini. Endaweni yoko, i-Reds iqhubekile nokusebenzisa i-brand ye-Ryzen 3 njengophawu olusezantsi, kubandakanywa nezisombululo ezimbini ze-APU ezingabizi kakhulu kunye nezilula-i-overclocked kancinane (xa kuthelekiswa neyandulelayo) i-3200G kunye nemizobo edibeneyo ye-Vega 8 ekwazi ukuphatha imithwalo yenkqubo esisiseko. kunye nemidlalo enesisombululo se-720p, kunye nomntakwabo omdala i-3400G, eyafumana i-core yevidiyo ekhawulezayo kunye nemizobo ye-Vega 11, kunye ne-SMT esebenzayo + yokwandisa amaza kuzo zonke iindawo. Esi sisombululo sinokwanela kwimidlalo elula kwi-1080p, kodwa ezi zisombululo zenqanaba lokungena zikhankanyiwe apha kungekhona ngenxa yeso sizathu, kodwa ngenxa yokungahambelani nokuvuza okuxelwe kwangaphambili i-Ryzen 3 kungekhona nje i-6 cores, kodwa nokugcina ixabiso elihlekisayo (malunga ne-$ 120). -150 ). Nangona kunjalo, asifanele silibale malunga nobume bokwenyani be-APU - basasebenzisa i-Zen + cores, kwaye enyanisweni ngabameli bechungechunge lwe-3000 kuphela ngokusemthethweni.
Nangona kunjalo, ukuba sithetha ngexabiso lesizukulwana esitsha xa sisonke, i-AMD iqinisekise ukuba iqinisekisa ubunkokeli obungenakuphikiswa kumacandelo amaninzi - iphumelele impumelelo ethile kudidi lwabaqhubekisi boluhlu oluphakathi.
I-Ryzen 5 3600-Igorha lesizwe ngaphandle kogcino
Enye yezinto eziphambili kuyilo lweZen 2 yeprosesa yayiyinguqu ukusuka kuyilo lwetshiphu eyodwa ukuya ekudalweni koyilo “lwemodyuli”- i-AMD iphumeze eyayo ipatent ye “chiplets”, iikristale ezincinci ezinecore processor ezidityaniswe yiInfinity. Ibhasi yelaphu. Ke, "obomvu" abungenanga kuphela kwimarike kunye nebhetshi entsha yezinto ezintsha, kodwa buye benza umsebenzi onzima kwenye yeengxaki ezicinezelayo zezizukulwana zangaphambili - ukubambezeleka okuphezulu kokubini xa usebenza ngememori naxa utshintshiselana ngedatha phakathi kweecores ezahlukeneyo. Iibhloko zeCCX.
Kwaye le ntshayelelo yayilapha ngesizathu-i-Ryzen 3600, ukumkani ongenakuphikiswa wecandelo eliphakathi, uzuze uloyiso olungenamiqathango ngokuchanekileyo ngenxa yezinto ezintsha eziphunyezwe yi-AMD kwisizukulwana esitsha. Ukonyuka okubalulekileyo ekusebenzeni okungundoqo kunye nokukwazi ukusebenza ngememori ngokukhawuleza kune-3200 MHz (eyona nto ininzi yayiyisilingi esebenzayo yesizukulwana sangaphambili) yenza ukuba kube lula ukuphakamisa ibha ukuya kwiindawo eziphakamileyo ezingazange zibonwe ngaphambili. i-i5-9600K ekhawulezayo, kodwa nakwi-flagship i7-9700.
Xa kuthelekiswa nowandulelayo, i-Ryzen 2600, umntu omtsha akafumananga nje uphuculo oluninzi kwicandelo loyilo, kodwa kunye nesimo sengqondo esincinci (i-3600 ifudumeza kancinci, yiyo loo nto i-AMD ikwazile ukugcina kwindawo epholileyo. ngokususa undoqo wobhedu), intloko epholileyo kunye nekhono lokungabi nazintsilelo. Ngoba? Kulula - i-3600 ayinayo, nangona oku kubonakala kungenangqondo. Zigwebe ngokwakho - i-frequency ephezulu inyuke nge-200 MHz, i-nameplate 65 W ayisafuni, kwaye ii-cores ezi-6 zilingana (okanye zigqithise!) I-Intel cores yangoku kwiLake Lake. Kwaye konke oku kwanikwa abalandeli nge-199 yakudala yeedola, ezinongiwe ngokuhambelana ngasemva kunye neebhodi zoomama ezininzi ze-AM4. I-Ryzen 3600 yayilungiselelwe impumelelo - kwaye ukuthengisa emhlabeni jikelele kubonisa ngokucacileyo oku kwinyanga yesithathu ngokulandelelana. Kweminye imimandla ehlala ithembekile kwi-Intel, imeko yemarike yatshintsha ngobusuku, kwaye amazwe aseYurophu (kunye neRashiya!) Azisa iqhawe elitsha lokuthengisa lelizwe kwincopho yempumelelo. Kubukhulu belizwe lethu, iprosesa ithathe i-10% yemarike kuzo zonke iintengiso ze-CPU elizweni, ngaphambi kwe-i7-9700K kunye ne-i9-9900K idibene. Kwaye ukuba umntu ucinga ukuba konke malunga nexabiso elimnandi, ngoko yonke into ayilula kakhulu: i-Ryzen 2600, ngokuthelekisa, ngexesha elifanayo emva kokungena kwimarike ayithathi ngaphezu kwe-3%. Imfihlo yempumelelo ilele kwenye indawo-i-AMD yabetha i-Intel kwelona candelo linabantu abaninzi kwimakethi yeprosesa, kwaye yathetha oku ngokuphandle kumboniso ngexesha lokuqala kweeprosesa kwi-CES2019. Kwaye ixabiso elimnandi, ukuhambelana okubanzi kunye nokupholisa kubandakanywa kuphela komeleza ubunkokeli obusele bungenakuphikiswa.
Ke kwakutheni ukuze umkhuluwa, 3600X, afuneke? Ngokufanayo kuzo zonke iimpawu, le prosesa yayikhawuleza ngenye i-200 MHz (kwaye yayinomfutho ophindaphindayo we-4.4 GHz), kwaye yasivumela ukuba sizuze inzuzo yokomfuziselo ngaphezulu kweprosesa encinci, engakhange ijonge ukukholisa ngokupheleleyo ngokuchasene nomva wendlela enkulu. ixabiso elonyuka ($229). Nangona kunjalo, imodeli yakudala yayisenazo iingenelo ezithile-oku ibikukungabikho kwesidingo sokuguqula izilayidi kwi-BIOS ukusukela amaza omoya ngaphezulu kwesiseko, kunye ne-Precision Boost 2.0, enokutshintsha ngamandla iprosesa kwiimeko zoxinzelelo, kunye nobunzima. cooler (Wraith Spire endaweni Wraith Stealth). Ukuba konke oku kuvakala ngathi sisindululo esihendayo, i-3600X lilitye elinqabileyo elivela kuluhlu olutsha lwe-AMD. Ukuba ukuhlawula ngokugqithiseleyo akusiyo inketho yakho, kwaye umehluko ekusebenzeni kwe-2-3% awubonakali ubalulekile, zikhululeke ukukhetha i-3600 - awuyi kuzisola.
I-Ryzen 7 3700X-Iflegi eNtsha eNdala
I-AMD ilungiselele ukutshintshwa kwenkokeli yangaphambili ngaphandle kwe-pathos eninzi - wonke umntu waqonda ukuba xa kuthelekiswa nabakhuphisana bangoku, i-2700X ibonakala incinci, kwaye inyathelo elikhulu eliya phambili (njengakwimeko ye-3600) lalicacile kwaye lilindelwe. Ngaphandle kokutshintsha ibhalansi yamandla ngokubhekiselele kwi-cores kunye neentambo, "obomvu" bazisa isibini seprosesa kwimarike, engenanto yantlukwano ekhethekileyo, kodwa ihluke kakhulu kwixabiso.
I-3700X yaboniswa njengokutshintshwa ngokuthe ngqo kwi-flagship yangaphambili - ngexabiso elicetyiswayo le-329 yeedola, i-AMD ibonise umntu okhuphisana ngokupheleleyo kwi-i7-9700K, igxininisa nganye yeenzuzo zayo, ezifana nezisombululo eziphezulu zeteknoloji kunye nobukho bezinto ezininzi. -umsonto, apho i-Intel yagqiba ekubeni igcine kuphela iiprosesa zayo "zasebukhosini" zodidi oluphezulu. Kwangelo xesha, i-AMD yazisa i-3800X, eyathi, eneneni, yayikhawuleza kancinci (300 MHz kwisiseko kunye ne-100 ekunyuseni) inguqulelo, kwaye ayizange ikwazi ukuzahlula ngayo nayiphi na indlela kwisihlobo sayo esincinci. Nangona kunjalo, kubantu abasaziva besoyikeka ngegama elithi "i-overclocking manual", olu khetho lukhangeleka lulungile, kodwa kuya kufuneka uhlawule kakhulu kwizinto ezincinci ezinjalo - ukuya kuthi ga kwi-70 yeedola ngaphezulu.
I-Ryzen 9 3900X kunye ne-3950X - Bonisa amandla
Nangona kunjalo, eyona nto ibalulekileyo (kunye nokuthetha ngokunyanisekileyo, iyimfuneko!) Isalathisi sempumelelo yeZen 2 yayizizisombululo ezindala ezivela kwintsapho yaseRyzen 9 - i-12-core 3900X kunye ne-16-core champion kwifom ye-3950X. Aba baprosesa, banonyawo olunye kwindawo yezisombululo ze-HEDT, bahlala benyanisekile kwingqikelelo yeqonga le-AM4, linovimba omkhulu wezixhobo ezinokumangalisa nabalandeli bonyaka ophelileyo beThreadripper.
I-3900X, ewe, yayijongwe ngokuyintloko ukuba incedise umgca we-Ryzen 3000 ngokuchasene nentsomi yemidlalo yangoku-i-9900K, kwaye kulo mba iprosesa yabonakala ilungile ngendlela emangalisayo. Ngokunyuswa kwe-4.5 GHz ngondoqo ngamnye kunye ne-4.3 kuzo zonke ezikhoyo, i-3900X ithathe inyathelo elibalulekileyo elibhekiselele kwi-parity ekudala ilindelwe kunye ne-Intel ekusebenzeni kwemidlalo, kwaye kwangaxeshanye amandla othusayo kuyo nayiphi na eminye imisebenzi - ukunikezela, i-computing, ukusebenza ngoovimba, njl. Imisonto engama-24 ivumele i-3900X ukuba ibambe i-Threadripper encinci ekusebenzeni okucocekileyo, kwaye kwangaxeshanye ingakhathazwa kukunqongophala kwamandla kumbindi ngamnye (njengoko kwakunjalo nge-2700X) okanye isiphene seendlela ezininzi zokusebenza ezingundoqo (kwaye. i-Game Mode eyaziwayo, ekhubaze isiqingatha se-cores kwi-AMD HEDT processors ). I-AMD idlalwe ngaphandle kokuthomalalisa, kwaye ngelixa isithsaba seprosesa yemidlalo ekhawulezayo sisesezandleni ze-Intel (esandul' ukutyhila i-9900KS, iprosesa ephikisanayo yohlelo lwabaqokeleli), iiReds zikwazile ukuhambisa eyona nto iguquguqukayo ephezulu. gem okwangoku kwimarike. Kodwa ayisiyiyo eyona inamandla- kwaye wonke umbulelo kwi3950X.
I-3950X yaba yintsimi yovavanyo lwe-AMD - ukudibanisa amandla omthombo we-HEDT kunye nesihloko "sokuqala kwe-16-core game processor" inokubizwa ngokuba yi-gamble ecocekileyo, kodwa eneneni "ibomvu" yayiphantse ingaxoki. Elona xesha liphezulu lokunyusa izinga le-4.7 GHz (nomthwalo kwi-1 core), ukukwazi ukusebenzisa zonke ii-cores ezili-16 kwi-frequency ye-4.4 GHz ngaphandle kokupholisa okungaqhelekanga, kunye nee-chiplets ezikhethiweyo zeklasi ephezulu, ekuvumela ukuba wenze I-monster entsha inoqoqosho ngakumbi kunomntakwabo we-12-core ngenxa yokunciphisa amandla ombane osebenzayo. Enyanisweni, ukhetho lokupholisa ngeli xesha luhlala luhlala kwisazela somthengi - i-AMD ayizange ithengise iprosesa nge-cooler, inciphisa kuphela ekukhuthazeni ukuthengwa kwe-240 okanye i-360 mm epholileyo.
Kwiimeko ezininzi, i-3950X ibonisa ukusebenza kwemidlalo kwinqanaba lesisombululo se-12-core, epholile kakhulu, ukhumbula ibali elibuhlungu lendlela yokuziphatha kwe-Threadripper. Nangona kunjalo, kwimidlalo apho ukusetyenziswa kweentambo kuncitshiswe kakhulu (umzekelo, kwi-GTA V), i-flagship ayikholisi iliso - kodwa oku kunokuba kuhluke kumgaqo.
Iprosesa entsha ye-16-core izibonakalisa ngendlela eyahluke ngokupheleleyo kwimisebenzi yobuchwephesha-ayisiyonto ukuba ukuvuza okuninzi kuthe i-AMD ilutshintshile ugxininiso lwayo kwicandelo labathengi kangangokuba i-3950X entsha iziva inentembelo nakwi-analogues ezibizayo njenge-i9. -9960X, ebonisa ukwanda okukhulu kwentsebenzo kwi-Blender, i-POV Mark, i-Premiere kunye nezinye izicelo ezinobuncwane. Kusuku olungaphambili, i-Threadripper yayisele ithembise umboniso omkhulu wamandla ekhompyuter, kodwa ne-3950X ibonise ukuba icandelo labathengi linokwahluka ngokupheleleyo - kwaye nokuba li-semi-professional. Ukukhumbula impumelelo ye-16-core flagship yeqonga le-AM4, umntu akanakunceda kodwa ukhumbule indlela i-Intel ephendule ngayo kuhlaselo olubhekiselele kwi-HEDT.
I-Intel 10xxxX - Ukuthobela kwi-Compromise
Kwangobusuku bokukhutshwa kwesizukulwana esitsha se-Threadripper, idatha ephikisanayo yavela apha naphaya malunga nomgca we-HEDT ozayo ovela kwi-Intel. Uninzi lwesiphithiphithi sasinxulumene namagama eemveliso ezintsha - emva kokukhutshwa kwezinto eziphikisanayo, kodwa zisentsha iiprosesa zeselfowuni ezivela kumgca we-Ice Lake kwitekhnoloji yenkqubo ye-10 nm, abaninzi abanomdla bakholelwa ukuba i-Intel yagqiba ekubeni ikhuthaze iimveliso kwizinto ezinqwenelekayo. I-10 nm kumanyathelo amancinci, engahlali kwiindawo ezininzi. Ukusuka kwindawo yokujonga imakethi yelaptop, ukukhutshwa kwe-Ice Lake akuzange kubangele ukothuka okukhethekileyo - isigebenga esiluhlaza kudala silawula intengiso yesixhobo esiphathwayo, kwaye i-AMD ayikabi nako ukukhuphisana nomatshini omkhulu we-OEM kunye namafutha. izivumelwano zeenkampani ezisebenze ngokusondeleyo ne-Intel ukususela ekuqaleni kwe-XNUMXs. Nangona kunjalo, kwimeko yecandelo leenkqubo eziphezulu zokusebenza, yonke into yahluke ngokupheleleyo.
Sazi yonke into malunga nomgca we-i9-99xxX - emva kwezizukulwana ezibini ze-Threadripper, i-AMD sele izibhengeze ngesibindi njengombaleki kwimakethi ye-HEDT, kodwa ukongamela kweentengiso eziluhlaza kwahlala kungagungqi. Ngelishwa kwi-Intel, iiReds azizange ziyeke kwimpumelelo yazo yangaphambili-kwaye emva kokuqala kweZen 2, kwacaca ukuba kungekudala iinkqubo eziphezulu ze-AMD ziya kuphakamisa kakhulu ibha yokusebenza, apho i-Intel yayingenamandla okuphendula, kuba isigebenga esiluhlaza sasinezisombululo ezitsha ngokusisiseko yayingabalulekanga.
Okokuqala, i-Intel kwafuneka ithathe inyathelo elingazange libonwe ngaphambili - ukunciphisa amaxabiso ngamaxesha e-2, engazange yenzeke ngaphambili kwiminyaka emininzi yokhuphiswano kunye ne-AMD. Ngoku i-flagship i9-10980XE ene-18 cores ebhodini ixabisa i-$ 979 kuphela endaweni ye-$ 1999 ye-predecessor yayo, kunye nezinye izisombululo ziye zehla kwixabiso ngexabiso elilinganayo. Nangona kunjalo, abaninzi sele beqondile ukuba balindele ntoni kwezi zimbini zikhutshiweyo kwaye ngubani oza kuvela ephumelele, ngoko i-Intel yathatha amanyathelo aqatha ngokuphakamisa i-embargo ekupapasheni uphononongo lweemveliso ezintsha kwiiyure ezi-6 ngaphambi komhla ocwangcisiweyo.
Kwaye uphononongo lwaqala ukubonakala. Nokuba awona matshaneli makhulu kunye nezixhobo zihleli zidanile kakhulu ngomgca omtsha - ngaphandle kotshintsho olukhulu kumgaqo-nkqubo wamaxabiso, umgca omtsha we-109xx uye waba yinto elula "umsebenzi kwiibugs" zesizukulwana sangaphambili - iifrikhwensi zatshintsha kancinci, iPCI eyongezelelweyo. Iindlela ze-E zavela, kwaye iphakheji ye-thermal yayinamandla okugqithisa kakhulu ayizange ishiye ithuba nakwi-hardcore abalandeli abane-SVOs ezinkulu-encotsheni i-10980X inokutya ngaphezu kwe-500 W, iqhayisa kungekuphela nje ukusebenza kakuhle kwiibenchmarks, kodwa ibonisa ngokucacileyo ukuba kukho. ayisiyonto ingaphaya kokucudisa ngaphandle kwe-14 nm kayisemkhulu.
Ayizange incede i-Intel ukuba iiprosesa ziyahambelana neqonga elikhoyo le-HEDT lesizukulwana sangaphambili - iimodeli ezincinci zomgca omtsha zilahlekile kwi-3950X ngokudilika komhlaba, zishiya abalandeli abaninzi be-Intel bedidekile. Kodwa okona kubi kwakuseza kufika.
I-Threadripper 3000 - 3960X, 3970X. Izilo zehlabathi lekhompyuter.
Ngaphandle kokuthandabuza okokuqala malunga nenani elincinci lee-cores (i-24 kunye ne-32 cores ayizange idale imvakalelo efana nokuphindaphinda kabini ii-cores eziye zenziwa kwi-Threadrippers yangaphambili), kwacaca ukuba i-AMD ayizukuzisa zisombululo kwimarike "yomboniso" -ukwanda okukhulu kwentsebenzo ngenxa yokulunga okuninzi kweZen 2 kunye nophuculo olumandla lwe-Infinity Fabric, ithembise ukusebenza okungazange kubonwe ngaphambili kwiqonga le-semi-pro - kwaye besingathethi nge-10-20%, kodwa into eyoyikisayo ngokwenene. . Kwaye xa i-embargo iphakanyisiwe, wonke umntu wabona ukuba amaxabiso amakhulu e-Threadripper entsha awazange athathwe emoyeni omncinci, kwaye kungekhona kumnqweno we-AMD wokukrazula abalandeli.
Ukusuka kwimbono yokonga iindleko, i-Threadripper 3000 yi-apocalypse yesipaji sakho. Iiprosesa ezinexabiso eliphezulu zifudukele kwiqonga elitsha ngokupheleleyo, eliphucuke ngakumbi kubuchwepheshe kunye nobunzima be-TRx40, ukubonelela ukuya kuthi ga kwi-88 PCI-e 4.0 imizila, kwaye ngaloo ndlela ibonelela ngenkxaso yoluhlu oluntsonkothileyo lwe-RAID ukusuka kwii-SSD zamva nje okanye iqela lamakhadi evidiyo aqeqeshiweyo. Umlawuli wememori ye-channel ezine kunye ne-subsystem enamandla ngokumangalisayo ayilungiselelwe kuphela kwiimodeli zangoku, kodwa kunye ne-flagship yexesha elizayo yomgca - i-64-core 3990X, ethembisa ukukhululwa emva koNyaka omtsha.
Kodwa nangona iindleko zinokubonakala ngathi yingxaki enkulu, ngokwemigaqo yokusebenza kwe-AMD ishiye kungekho litye lingaguqukiyo kwiimveliso ezintsha ze-Intel - kwinani lezicelo i-Threadripper ebonisiweyo yayikhawuleza kabini njenge-flagship 10980XE, kwaye ukonyuka kokusebenza komndilili bekumalunga ne-70% . Kwaye oku nangona i-appetites ye-3960X kunye ne-3970X iphakathi kakhulu - zombini iiprosesa azidli ngaphezu kwe-280 W, kunye ne-overclock ephezulu ye-4.3 GHz kuzo zonke ii-cores zihlala ziyi-20% ngaphezulu kwezoqoqosho kune-red- iphupha elibi elishushu elivela kwi-Intel.
Ngaloo ndlela, i-AMD yakwazi okokuqala kwimbali ukubonelela kwimarike imveliso yeprimiyamu engaguqukiyo ebonelela nje ngokunyuka okukhulu kwentsebenzo, kodwa nayo ayinayo nayiphi na imiqobo ebalulekileyo - ngaphandle kwexabiso, kodwa, njengoko besithi, kufuneka uhlawule ngaphezulu kokungcono. Kwaye i-Intel, njengokuba kungenangqondo njengoko kubonakala ngathi, iguquke yaba yindlela yoqoqosho, leyo, nangona kunjalo, ayikhangeleki ithembele kakhulu kwimvelaphi ye-3950 750X yeedola kwiqonga elifikelelekayo kakhulu.
I-Athlon 3000G - Nhlangula ngepeni entle
I-AMD ayizange ilibale malunga necandelo lohlahlo lwabiwo-mali lwe-processors zamandla aphantsi kunye nemizobo esemthethweni ebhodini - apha entsha (kodwa nayo indala) i-Athlon 5400G ingxamele ukuhlangula abo bajonga iPentium G3000 ngokudelela okukhulu. I-2 cores kunye ne-4 imicu, i-3.5 GHz i-frequency frequency kunye ne-Vega 3 yevidiyo engundoqo eyaziwayo (ejijelweyo ukuya kwi-100 MHz) kunye ne-TDP ye-35 W - kwaye konke oku kwi-$ 49 ehlekisayo. I-Reds iphinde yanikela ingqalelo ekhethekileyo kwithuba lokugqithisa iprosesa, ukubonelela ubuncinane enye i-30% yokusebenza kwi-frequency ye-3.9 GHz. Kwangaxeshanye, awuzukufuneka uchithe imali kwi-cooler ebiza kakhulu kuhlahlo lwabiwo-mali - i-3000G iza nokupholisa okugqwesileyo okwenzelwe i-65 W yobushushu - oku kwanele nokuba kugqithise kakhulu.
Kwimiboniso, i-AMD ithelekisa i-Athlon 3000G kunye nomntu okhuphisana naye ngoku kwi-Intel - i-Pentium G5400, eyathi yabiza kakhulu (ixabiso elicetyiswayo - i-$ 73), ithengiswa ngaphandle kokupholisa, kwaye iphantsi kakhulu ekusebenzeni kwimveliso entsha. . Kwakhona kuyahlekisa ukuba i-3000G ayakhiwe kwi-architecture ye-Zen 2 - isekelwe kwi-Zen endala enhle + kwi-12 nm, esivumela ukuba sibize umkhiqizo omtsha uhlaziyo oluncinci lwe-Athlon 2xx GE yonyaka odlulileyo.
Iziphumo zenguqu "ebomvu".
Ukukhutshwa kweZen 2 kube nefuthe elikhulu kwimakethi yeprosesa-mhlawumbi olo tshintsho lukhulu aluzange lubonwe kwimbali yangoku yee-CPU. Sinokukhumbula imashi ephumelelayo ye-AMD 64 FX, sinokukhankanya uloyiso lwe-Athlon phakathi kwiminyaka elishumi edlulileyo, kodwa asikwazi ukunika isifaniso esivela kwixesha elidlulileyo le-giant "ebomvu", apho yonke into yatshintsha ngokukhawuleza. kwaye impumelelo yayimangalisa nje. Kwiminyaka nje emi-2, i-AMD ikwazile ukwazisa izisombululo zeseva ye-EPYC ezinamandla kakhulu, yafumana iikhontrakthi ezininzi ezinengeniso kwiinkampani ze-IT zehlabathi, yabuyela kumdlalo kwicandelo labathengi benkqubo yokudlala kunye ne-Ryzen, kwaye yakhupha i-Intel kwimarike ye-HEDT ngoncedo lwe iThreadripper engenakuthelekiswa nanto. Kwaye ukuba ngaphambili kwakubonakala ngathi ngumbono oqaqambileyo kaJim Keller kuphela owaba semva kwayo yonke impumelelo, emva koko ngokukhululwa koyilo lweZen 2 kwimarike, kwacaca ukuba uphuhliso lwalo mbono luye lwahamba phambili. iskimu sokuqala-sifumene izisombululo zebhajethi ezigqwesileyo (i-Ryzen 3600 yaba yeyona iprosesa idumileyo kwihlabathi- kwaye isahlala injalo), izisombululo ezinamandla zendalo iphela (3900X inokukhuphisana ne-9900K, kwaye imangaliswe ngempumelelo yayo kwimisebenzi yobuchwephesha), imifuniselo enesibindi (3950X !), kunye nezisombululo ze-ultra-economical zeyona misebenzi elula yemihla ngemihla (Athlon 3000G). Kwaye i-AMD iqhubela phambili - kulo nyaka uzayo siya kuba nesizukulwana esitsha, impumelelo entsha kunye neziganeko ezitsha eziya kuthi zoyiswe ngokuqinisekileyo!
Indlu yekholamu ye-NHTi "Iimfazwe zeProcessor" kwiziqendu ezi-7 kuYouTube -
Umbhali wenqaku: Alexander Lis.
Ngabasebenzisi ababhalisiweyo kuphela abanokuthatha inxaxheba kuphando. , ndiyacela.
Ngoko ke yeyiphi engcono?
-
68,6%I-AMD327
-
31,4%Intel 150
Bali-477 abasebenzisi abavotileyo. Abasebenzisi abasi-158 abakhange.
umthombo: www.habr.com