Qaybinta Tusaalooyinka badan ee GPU
Multi-Instance GPU (MIG) waa tignoolajiyada NVIDIA oo u qaybisa hal GPU jireed qaybo badan oo go'doonsan.
Dulmar
It matters because it lets one expensive accelerator serve many small workloads at once without them interfering with each other.
quusid qoto dheer
Waxaa lagu soo bandhigay NVIDIA A100 (Ampere) oo lagu sii waday H100 iyo GPU-yada cusub ee xarunta xogta, MIG wuxuu u sawiraa GPU ilaa todoba xaaladood oo madax banaan. Si ka duwan goynta software-ka, MIG waxay bixisaa go'doomin qalabeed run ah: tusaale kastaa wuxuu helayaa qalabyo badan oo qulqulaya oo u go'an (SMs), jeexjeexyada kaydinta L2, kontaroolayaasha xusuusta, iyo jeex go'an oo xusuusta bandwidth-sare ah. A100 oo leh 40GB waxa loo kala qaybin karaa todobo 5GB, ama kuwo ka yar. Qayb kastaa waxay u dhaqantaa sida GPU yar oo gooni ah, markaa shaqo buuq badan ama burbursan hal tusaale ma gaajoon karto ama ma musuqmaasuqi karto mid kale. Adeeggan tayada la dammaanad qaaday ayaa ka dhigaya MIG mid ku habboon u adeegidda soo jeedinta, kooxo-kiraysteyaal badan, iyo deegaan horumarineed oo ay isticmaaleyaal badani wadaagaan hal kaar.
Aragtida Farsamada
MIG waxay u shaqeysaa iyadoo jir ahaan u xireysa isgoysyada gudaha ee GPU-da si tusaale kasta uu u yeesho waddo go'an oo loo maro qaybtiisa xusuusta iyo SM-yada. NVIDIA waxay ku qeexdaa profiles sida jajabyo sida 1g.5gb (hal xisaabiye, 5GB) ilaa 7g.40gb. Tusaalaha GPU wuxuu kaydiyaa xusuusta iyo SMS-yada; dhexdeeda Tusmada Xisaabinta ayaa sii kala qaybisa SMS-yada. Sababtoo ah qaybuhu waa qalab-xoog-gudbiyeedka, cilladaha, khaladaadka ECC, iyo baaxadda xajinta xusuusta waxay ku ekaanayaan hal tusaale.
Saamaynta Istiraatijiyadeed
Qiimaha iyo miisaaniyada
Go'aamada qaab-dhismeedku waxay horseedaan waxqabadka iyo kharashka hawlgalka sannadaha.
Go'aamo cad
Waxbarashada farsamada waxay ka caawisaa kooxaha inay doortaan xidhmo sax ah, ma aha oo kaliya kan ugu cusub.
Xakamaynta tayada
Doorashooyinka injineernimada ee wanaagsan waxay yareeyaan shilalka la isku halleyn karo ee wax soo saarka.
Mustaqbalka Qaybinta Tusaalooyinka Badan ee GPU
Marka ay GPU-yadu u koraan ilaa 80GB, 141GB, iyo wixii ka dambeeya, qaybintu waxay noqotaa mid soo jiidasho leh sababtoo ah moodooyinka shakhsi ahaaneed marar dhif ah ayay u baahan yihiin kaar dhan fikradda. Filo Kubernetes adag iyo is-dhexgalka daruuraha, kala qaybsanaan firfircoon iyada oo aan la daadin noodhka, iyo profiles-ka-fiican. Iibiyeyaasha tartamaya ayaa daba jooga qaab-dhismeedka GPU-da ee la midka ah ee SR-IOV, iyo aaladaha kala-soocida server-la'aanta ah waxay si isa soo taraysa ugu tiirsan yihiin qaybinta si ay u xirxiraan moodooyin badan oo cufan oo ay gooyaan qashinka aan shaqayn.
Dhaqangelinta Adduunka-dhabta ah
Bixiyaha daruuraha ayaa hal A100 u kala qaybiya todobo xaaladood si todobo macaamiil midkiiba u helo dammaanad, jeex GPU go'doonsan.
Koox cilmi baaris jaamacadeed ah ayaa siisa arday kasta oo PhD ah tusaale 10GB MIG ah oo loogu talagalay prototying halkii ay ka koobnaan lahaayeen kaararka oo dhan.
Adeegga ka-fiirsashadu waxa uu ku xidhaa dhawr nooc oo luuqado yaryar ah iyo noocyo aragga oo dul saaran hal H100, mid walbana qaybtiisa ayuu ku leeyahay daahitaan la saadaalin karo.
Kooxda Kubernetes ayaa u xayeysiisa tusaalooyinka MIG kheyraadka la jadwalsan karo si ay boosasku u codsadaan 'nvidia.com/mig-1g.5gb' sida kheyraadka kale.
Khatarta & Dariiqyada Ilaalada
Hagaajinta hal bartilmaameed waxay qarin kartaa daciifnimada nidaamka ballaaran.
Kaabayaasha dhaqaalaha iyo dayactirka inta badan waa la dhayalsadaa.
Nabadgelyada iyo daldaloolada u fiirsashada ayaa kori kara marka nidaamyadu noqdaan kuwo aad u adag.
Qorshe Hawleedka Dhaqangelinta
Qeex daahida, tayada, iyo bartilmaameedyada qiimaha ka hor inta aan la hirgelin.
Benchmark marka la eego culeyska dhabta ah iyo xaaladaha xogta.
La socodka qalabka khaladaadka, leexashada, iyo saamaynta isticmaalaha.
U diyaari dib-u-noqoshada iyo dariiqyada jawaab-celinta dhacdada ka hor inta aanad miisaan.
Sii wad Sahaminta
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the Multi-Instance GPU Partitioning quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Hagaha xiga
Barashada Hawlo Badan
Su'aalaha soo noqnoqda
What is Multi-Instance GPU Partitioning?
Multi-Instance GPU (MIG) waa tignoolajiyada NVIDIA oo u qaybisa hal GPU jireed qaybo badan oo go'doonsan. Waa arrin sababtoo ah waxay u ogolaataa hal dardariye qaali ah inuu u adeego culaysyo shaqo oo badan oo yar yar hal mar iyaga oo aan midba midka kale faragelin.
Go'doon noocee ah ayay MIG bixisaa inta u dhaxaysa xaaladaha?
MIG waxay xoojisaa go'doominta qalabka, u hibeynta SM-yada, xaleefyada kaydka L2, iyo xusuusta tusaale kasta si culeyska shaqadu aanu u faragelin karin.
Qaab dhismeedka NVIDIA ee ugu horeysay ee MIG lagu soo bandhigay?
MIG waxa uu ka dooday A100-ku-salaysan Ampere oo uu ku sii socday H100 iyo GPU-yada dambe ee xogta-xarunta.
Waa maxay tirada ugu badan ee tusaalooyinka MIG awooda u leh ee loo qaybin karo?
GPU karti u leh MIG sida A100 waxa loo qaybin karaa ilaa todoba xaaladood oo madax banaan.
Muuqaalka MIG sida '1g.5gb', muxuu '5gb' u taagan yahay?
Profile-ku wuxuu qeexayaa xisaabinta jeexjeexyada (1g) iyo xusuusta u go'an (5GB) ee la siiyay tusaale ahaan.
Shaqadee ayaa MIG gaar ahaan ku habboon?
MIG waxay iftiimaysaa marka qaar badan oo yaryar, culaysyo shaqo oo go'doonsan (sida fikradda) ay wadaagaan hal kaadh oo leh tayada adeeg ee dammaanad qaaday.