{"id":122838,"date":"2025-04-02T12:55:26","date_gmt":"2025-04-02T10:55:26","guid":{"rendered":"https:\/\/prohoster.info\/?p=122838"},"modified":"2025-04-02T12:55:26","modified_gmt":"2025-04-02T10:55:26","slug":"novaya-statya-radeon-rx-9070-xt-protiv-geforce-rtx-5070-ti","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sl\/blog\/news\/novaya-statya-radeon-rx-9070-xt-protiv-geforce-rtx-5070-ti","title":{"rendered":"Novi \u010dlanak: Radeon RX 9070 XT protiv GeForce RTX 5070 Ti","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

Zna\u010dilnosti GPU dru\u017eine Blackwell in testi GeForce RTX 5080<\/a><\/noindex> so pokazali, da so vse pozicije 50. serije, razen GeForce RTX 5090, le nekoliko bolj\u0161i nadgradnji ustreznih modelov prej\u0161nje generacije. Situacijo je \u0161e poslab\u0161al pomanjkanje novih grafi\u010dnih kartic in posledi\u010dno astronomske cene.<\/p>\n

Ve\u010dina kupaca visoke klase ure\u0111aja, kojima monopolo raspola\u017ee NVIDIA, izaziva samo teorijski interes. Dok je na ni\u017eem nivou tehnolo\u0161ki zastoj 'zelene' grafike stvorio situaciju u kojoj bi AMD mogao jednim skokom nadoknaditi zaostatak u problemati\u010dnim aspektima arhitekture RDNA, kao \u0161to su ray tracing i upscaling. Me\u0111utim, proroci da \u0107e AMD ponovo posti\u0107i paritet sa konkurentom (iako bez ambicija za flagship sektor) \u010duju se ve\u0107 dugo, a svaki put zvu\u010de sve manje uverljivo, stoga najava grafi\u010dkih kartica Radeon RX 9000 nije privukla veliku pa\u017enju naspram burne aktivnosti NVIDIA. A ipak, AMD je obavio opse\u017ean rad na gre\u0161kama i, ako nije uklonio, onda je bar donekle kompenzovao nedostatke pro\u0161lih GPU-a.<\/p>\n

Pregled Radeon RX 9070 XT prikazujemo u neobi\u010dnom formatu upore\u0111ivanja za 3DNews, budu\u0107i da su 'crvena' novina i njen glavni konkurent \u2014 GeForce RTX 5070 Ti \u2014 dospjeli u na\u0161e ruke istovremeno, i jedno ure\u0111aj ne mo\u017ee se posmatrati odvojeno od drugog, jer se u Rusiji prodaju za otprilike iste pare.<\/p>\n\n\n\n
\"Novi <\/td>\n <\/td>\n \"Novi <\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

<\/center> \t\t\t\t <\/p>\n

⇡<\/a><\/noindex>#<\/a><\/noindex> Grafi\u010dni procesor Navi 48 in arhitektura RDNA 4<\/strong><\/h2>\n

O tome da je NVIDIA postala monopolista na tr\u017ei\u0161tu visokoperformantnih GPU-a i pona\u0161a se u skladu s tim, nije govorio samo lenj. Re\u0111e se obra\u0107a pa\u017enja na \u010dinjenicu da sve kompanije koje projektuju velike procesore za njihovu proizvodnju moraju da koriste usluge jednog izvo\u0111a\u010da \u2014 tajvanske TSMC. Potra\u017enja za kapacitetima TSMC je sigurno postala jedan od razloga za\u0161to je AMD obustavio razvoj velikih potro\u0161a\u010dkih GPU-a, a i generalno u liniji 'crvenih' grafi\u010dkih procesora nove generacije trenutno postoji samo jedan kristal \u2014 Navi 48.<\/p>\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
Proizvajalec<\/th>\nAMD<\/th>\n<\/tr>\n
Ime<\/td>\nNavi 32<\/td>\nNavi 31<\/td>\nNavi 48<\/td>\n<\/tr>\n
Uporaba<\/td>\n\n

Radeon RX 7700 XT;<\/p>\n

Radeon RX 7800 XT<\/p>\n<\/td>\n

\n

Radeon RX 7900 GRE;<\/p>\n

Radeon RX 7900 XT;<\/p>\n

Radeon RX 7900 XTX<\/p>\n<\/td>\n

\n

Radeon RX 9070;<\/p>\n

Radeon RX 9070 XTX<\/p>\n<\/td>\n<\/tr>\n

Arhitektura<\/td>\nRDNA 3<\/td>\nRDNA 3<\/td>\nRDNA 4<\/td>\n<\/tr>\n
Tehnolo\u0161ki proces, nm<\/td>\nTSMC N5\/N6<\/td>\nTSMC N5\/N6<\/td>\nTSMC N4P<\/td>\n<\/tr>\n
\u0160tevilo tranzistorjev, milijard<\/td>\n28,1<\/td>\n57,7<\/td>\n53,9<\/td>\n<\/tr>\n
Povr\u0161ina \u010dipa, mm2<\/td>\n346<\/td>\n522<\/td>\n357<\/td>\n<\/tr>\n
\u0160tevilo CU\/WGP\/SE<\/td>\n<\/tr>\n
Ra\u010dunske enote (CU)<\/td>\n60<\/td>\n96<\/td>\n64<\/td>\n<\/tr>\n
Procesorji delovnih skupin (WGP)<\/td>\n30<\/td>\n48<\/td>\n32<\/td>\n<\/tr>\n
Polja sen\u010dil (SA)<\/td>\nNe<\/td>\nNe<\/td>\nNe<\/td>\n<\/tr>\n
Motorji sen\u010dil (SE)<\/td>\n4?<\/td>\n6<\/td>\n4<\/td>\n<\/tr>\n
Konfiguracija Compute Unit<\/td>\n<\/tr>\n
Vektorske ALU (FP32)<\/td>\n2 \u00d7 32<\/td>\n2 \u00d7 32<\/td>\n2 \u00d7 32<\/td>\n<\/tr>\n
Vektorske ALU (FP32\/INT32)<\/td>\n2 \u00d7 32<\/td>\n2 \u00d7 32<\/td>\n2 \u00d7 32<\/td>\n<\/tr>\n
Skalarni ALU<\/td>\n2<\/td>\n2<\/td>\n2<\/td>\n<\/tr>\n
ALU za posebne namene (SFU)<\/td>\n2 \u00d7 8<\/td>\n2 \u00d7 8<\/td>\n2 \u00d7 8<\/td>\n<\/tr>\n
Blokovi trazenja zraka (Ray Accelerators)<\/td>\n1<\/td>\n1<\/td>\n1<\/td>\n<\/tr>\n
Blokovi teksturiranja (TMU)<\/td>\n4<\/td>\n4<\/td>\n4<\/td>\n<\/tr>\n
Vektorski\/skalarni registri, Kb<\/td>\n384\/10<\/td>\n384\/10<\/td>\n384\/16<\/td>\n<\/tr>\n
Kapacitet L0 ke\u0161a, Kb<\/td>\n32<\/td>\n32<\/td>\n32<\/td>\n<\/tr>\n
Konfiguracija Workgroup Processora (WGP)<\/td>\n<\/tr>\n
Kapacitet Shared Memory, Kb<\/td>\n128<\/td>\n128<\/td>\n128<\/td>\n<\/tr>\n
Kapacitet instrukcijskog ke\u0161a, Kb<\/td>\n32<\/td>\n32<\/td>\n32<\/td>\n<\/tr>\n
Kapacitet skalarnih ke\u0161eva, Kb<\/td>\n16<\/td>\n16<\/td>\n16<\/td>\n<\/tr>\n
Ra\u010dunarski blokovi GPU<\/td>\n<\/tr>\n
Vektorske ALU (FP32)<\/td>\n3 840 (7 680)<\/td>\n6 144 (12 288)<\/td>\n4 096 (8 192)<\/td>\n<\/tr>\n
Blokovi teksturiranja (TMU)<\/td>\n240<\/td>\n384<\/td>\n256<\/td>\n<\/tr>\n
Blokovi rasterizacije (ROP)<\/td>\n96<\/td>\n192<\/td>\n128<\/td>\n<\/tr>\n
Blokovi trazenja zraka (Ray Accelerators)<\/td>\n60<\/td>\n96<\/td>\n64<\/td>\n<\/tr>\n
Konfiguracija memorije<\/td>\n<\/tr>\n
Kapacitet L1 ke\u0161a, Kb<\/td>\n2 048<\/td>\n3 072<\/td>\n2 048<\/td>\n<\/tr>\n
Kapacitet L2 ke\u0161a, Mb<\/td>\n4<\/td>\n6<\/td>\n8<\/td>\n<\/tr>\n
Kapacitet Infinity Cache, Mb<\/td>\n64<\/td>\n96<\/td>\n64<\/td>\n<\/tr>\n
\u0160irina VRAM sabirnice, bit<\/td>\n256<\/td>\n384<\/td>\n256<\/td>\n<\/tr>\n
Tip VRAM \u010dipova<\/td>\nGDDR6<\/td>\nGDDR6<\/td>\nGDDR6<\/td>\n<\/tr>\n
<\/td>\n<\/tr>\n
PCI Express interfejs<\/td>\n4.0 x16<\/td>\n4.0 x16<\/td>\n5.0 x16<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

Me\u0111u starijim \u010dipovima najvi\u0161e podse\u0107a na model drugog nivoa, Navi 32, po osnovnim karakteristikama, ali sadr\u017ei gotovo jednaku koli\u010dinu tranzistora kao Navi 31 (53,9 i 57,7 milijardi respektivno). Navi 48 se proizvodi prema TSMC N4P tehnologiji, koja pripada istom litografskom procesu od 5 nm, koji je razvijen pre dve godine. Ipak, Navi 48 zauzima samo 68% povr\u0161ine Navi 31 zahvaljuju\u0107i tome \u0161to je AMD napustio \u010diplet dizajn, jer se MCD \u010dipleti, koji sadr\u017ee tre\u0107i nivo ke\u0161a i kontrolere memorije, proizvode prema grubljoj N6 tehnologiji.<\/p>\n

\"Novi <\/p>\n

Arhitektura nove grafi\u010dne enote, RDNA 4, se strukturalno ne razlikuje od prej\u0161njih iteracij. Navi 48 vsebuje 64 enot za izra\u010dun, ali 4.096 FP32 zdru\u017eljivih sen\u010dilnih ALU, kar je povsem tipi\u010dna formula za GPU srednjega razreda - prav tako kot 256-bitna vodila za video pomnilnik. Vendar pa AMD v ocenah nezmo\u017enosti postavlja Navi 48 nasproti vodilnemu Navi 31, ki ima 6.144 \u0161tevilsko natan\u010dnih sen\u010dilnih ALU. Razliko bi morale v veliki meri nadomestiti \u0161tevilne izbolj\u0161ave logike, za kar priporo\u010damo osve\u017eitev spomina na arhitekturo RDNA 3, ki smo jo podrobno obravnavali v pregledu. Radeon RX 7900 XTX<\/a><\/noindex>.<\/p>\n

V primerjavi z 'zeleno' arhitekturo Blackwell<\/a><\/noindex> RDNA 4 predstavlja \u0161iri i dublji unapre\u0111enje prethodnih koncepata. Me\u0111utim, opis svih novina na \u010dudan na\u010din se svodi na kratak spisak zbog toga \u0161to NVIDIA tradicionalno posve\u0107uje ve\u0107u pa\u017enju slo\u017eenim softverskim ili polu-softverskim re\u0161enjima, koja zahtevaju obimne komentare, dok inovacije AMD-a obi\u010dno po\u010dinju i zavr\u0161avaju u hardveru.<\/p>\n

\"Novi <\/p>\n

Temeljne spremembe so se zgodile v napravi enote za izra\u010dun - nedeljivem ra\u010dunalni\u0161kem bloku, ki je analog SM pri NVIDIA in jeder CPU. Tako zdaj na\u010drtovalnik ukazov podpira t. i. deljene ovire, ki omogo\u010dajo oddajo ukazov v izvajanje med njihovim trajanjem. Prav tako je bila optimizirana predizbira ukazov iz predpomnilnika, medtem ko so skalarni ALU, ki se uporabljajo za ekonomi\u010dno izvajanje operacij pogojevanja in razvejitve, podedovali sposobnost dela s \u0161tevilskimi podatki iz mobilne arhitekture RDNA 3.5. Hkrati so pove\u010dali tudi obseg skalarnih registrov: s 10 na 16 KB.<\/p>\n

\"Novi <\/p>\n

One of the most significant innovations of RDNA 4 is the doubled execution rate of matrix instructions WMMA, which are used in machine learning tasks and data processing with neural networks. Moreover, the RDNA 4 instruction set includes commands for calculations with structurally sparse matrices and 8-bit floating-point data (FP8 and BF8).<\/p>\n

\"Novi <\/p>\n

Kljub temu pa prepustnost 1.024 FLOP na takt pomeni, da so \u010dipi Navi le dosegli raven predprej\u0161nje 'zeleno' arhitekture - Ampere. Poleg tega, \u010de \u010dipi NVIDIA in Intel uporabljajo lo\u010dene mno\u017ei\u010dne matrice ALU za matri\u010dne operacije, pa v arhitekturi RDNA 4 to delo \u0161e vedno pade na sen\u010dilne SIMD.<\/p>\n

Vendar ne smemo pozabiti, da se grafi\u010dni procesorji, temelje\u010di na arhitekturi RDNA 3, nikoli ne pribli\u017eujejo polni obremenitvi v realnih aplikacijah. Oglasi terefopsi v izra\u010dunih FP32 za splo\u0161no rabo so dosegljivi le pod pogojem, da se uporablja format VOPD (Vector Operation Dual), ki omogo\u010da zdru\u017eevanje dveh neodvisnih vektorskih navodil. Ta sam po sebi postavlja \u0161tevilne omejitve, zato polovica 32-bitnih ALU znotraj Compute Unit'a pogosto ostaja neaktivna. Pri me\u0161anih obremenitvah (kot je igra z upscalingom) bodo matrike seveda odvzele del izra\u010dunskih virov sen\u010dilom, vendar bodo te izgube manj\u0161e, kot se zdi na prvi pogled.<\/p>\n\n\n\n\n\n\n
<\/th>\nCompute Unit (AMD RDNA 3)<\/th>\nCompute Unit (AMD RDNA 4)<\/th>\nStreaming Multiprocessor (NVIDIA Ada Lovelace)<\/th>\nStreaming Multiprocessor (NVIDIA Blackwell)<\/th>\nXe-core (Intel Xe-HPG)<\/th>\nXe-core (Intel Xe2)<\/th>\n<\/tr>\n
Execution Units<\/td>\n\n

2 \u00d7 SIMD32 (FP32\/INT32);<\/p>\n

2 \u00d7 SIMD32 (FP32);<\/p>\n

2 \u00d7 SIMD2 (FP64);<\/p>\n

2 \u00d7 SIMD8 (SFU);<\/p>\n

2 \u00d7 skalarnh ALU;<\/p>\n<\/td>\n

\n

2 \u00d7 SIMD32 (FP32\/INT32);<\/p>\n

2 \u00d7 SIMD32 (FP32);<\/p>\n

2 \u00d7 SIMD2 (FP64);<\/p>\n

2 \u00d7 SIMD8 (SFU);<\/p>\n

2 \u00d7 skalarnh ALU;<\/p>\n<\/td>\n

\n

4 \u00d7 SIMD16 (FP32\/INT32);<\/p>\n

4 \u00d7 SIMD16 (FP32);<\/p>\n

2 \u00d7 SISD? (FP64);<\/p>\n

4 \u00d7 SIMD4 (SFU);<\/p>\n

4 \u00d7 skalarnh ALU;<\/p>\n

4 \u00d7 tensornih jeder;<\/p>\n<\/td>\n

\n

8 \u00d7 SIMD16 (FP32\/INT32);<\/p>\n

2 \u00d7 SISD? (FP64);<\/p>\n

4 \u00d7 SIMD4 (SFU);<\/p>\n

4 \u00d7 skalarnh ALU;<\/p>\n

4 \u00d7 tensornih jeder;<\/p>\n<\/td>\n

\n

16 \u00d7 SIMD8 (FP32);<\/p>\n

16 \u00d7 SIMD8 (INT32);<\/p>\n

16 \u00d7 SISD (FP64);<\/p>\n

16 \u00d7 SIMD2 (SFU);<\/p>\n

16 \u00d7 XMX;<\/p>\n<\/td>\n

\n

8 \u00d7 SIMD16 (FP32);<\/p>\n

8 \u00d7 SIMD16 (INT32);<\/p>\n

8 \u00d7 SIMD2 (FP64);<\/p>\n

8 \u00d7 SIMD4 (SFU);<\/p>\n

8 \u00d7 XMX;<\/p>\n<\/td>\n<\/tr>\n

SIMD operations per cycle<\/td>\n\n

128 \u00d7 FP32;<\/p>\n

64 \u00d7 INT32;<\/p>\n

256 \u00d7 FP16;<\/p>\n

4 \u00d7 FP64;<\/p>\n

16 \u00d7 transcendentalne funkcije;<\/p>\n<\/td>\n

\n

128 \u00d7 FP32;<\/p>\n

64 \u00d7 INT32;<\/p>\n

256 \u00d7 FP16;<\/p>\n

4 \u00d7 FP64;<\/p>\n

16 \u00d7 transcendentalne funkcije;<\/p>\n<\/td>\n

\n

128 \u00d7 FP32;<\/p>\n

64 \u00d7 INT32;<\/p>\n

128 \u00d7 FP16;<\/p>\n

2 \u00d7 FP64;<\/p>\n

16 \u00d7 transcendentalne funkcije;<\/p>\n<\/td>\n

\n

128 \u00d7 FP32;<\/p>\n

128 \u00d7 INT32;<\/p>\n

128 \u00d7 FP16;<\/p>\n

2 \u00d7 FP64;<\/p>\n

16 \u00d7 transcendentalne funkcije;<\/p>\n<\/td>\n

\n

128 \u00d7 FP32;<\/p>\n

128 \u00d7 INT32;<\/p>\n

256 \u00d7 FP16;<\/p>\n

16 \u00d7 FP64;<\/p>\n

32 \u00d7 transcendentalne funkcije;<\/p>\n<\/td>\n

\n

128 \u00d7 FP32;<\/p>\n

128 \u00d7 INT32;<\/p>\n

256 \u00d7 FP16;<\/p>\n

16 \u00d7 FP64;<\/p>\n

32 \u00d7 transcendentalne funkcije;<\/p>\n<\/td>\n<\/tr>\n

Matrix operations, FLOP per cycle (FP16)<\/td>\n512<\/td>\n1 024<\/td>\n2 048<\/td>\n2 048<\/td>\n2 048<\/td>\n2 048<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

Spreminjanje slik s pomo\u010djo nevronskih mre\u017e ni ve\u010d teoreti\u010dna situacija za pospe\u0161evalce Radeon. Nova, \u010detrta razli\u010dica FSR kon\u010dno temelji na strojni inteligenci, vklju\u010dno za generiranje slik, uporabljeni modeli pa so hibridna razli\u010dica konvolucijskih omre\u017eij in transformatorjev. \u017dal, za razliko od prej\u0161njih razli\u010dic upscalerja, ki so delovale prakti\u010dno na vsem hardveru, FSR 4 podpirajo le grafi\u010dni procesorji arhitekture RDNA 4 (v povezavi z raz\u0161irjenim naborom navodil WMMA za podatke FP8). Univerzalni na\u010din z poenostavljenimi izra\u010dunskimi jedri, kot pri Intelu XeSS, ni predviden.<\/p>\n

\"Novi <\/p>\n

Drugotna smer reforme arhitekture RDNA je postala sledenje \u017earkov. Hitrost testiranja presekov \u017earka z BVH in trikotnikom se je podvojila na en Ray Accelerator (torej tudi en Compute Unit): iz 4 in 1 za takt na 8 in 2. Kljub temu RT-blok \u010dipov Intel Battlemage deluje s hitrostjo 18 in 2, medtem ko grafi\u010dni procesorji Blackwell hkrati najdejo \u0161tiri pregrade \u017earka s trikotnikom na takt in neobjavljeno \u0161tevilo presekov \u017earka z BVH. Pri tekmecih se testi dveh vrst (proti BVH in trikotniku) izvajajo hkrati, pri AMD pa \u0161e vedno ne. Nekoliko paralelizma je mo\u017eno, saj vsak Ray Accelerator sedaj vsebuje dva neodvisna 'testiralnika', vendar ne pri polni hitrosti.<\/p>\n

\"Novi <\/p>\n

Additionally, the Ray Accelerator has gained the ability to process rays out of order. Rays that do not require access to slow memory stack levels are processed first, and the testing results are stored so that the shader program receives them in the expected order.<\/p>\n

Logika raytracinga sada koristi gusto strukturiran BVH8 umjesto BVH4 (osam umjesto \u010detiri \u010dvora za svaki granu hijerarhije volumena), \u0161to pove\u0107ava brzinu prolaska i istovremeno smanjuje pritisak na video memoriju. Veli\u010dina BVH-a u memoriji smanjena je i zahvaljuju\u0107i novom algoritmu kompresije primitivi. Napokon, AMD nudi odvezivanje orijentacije BVH kutija od apsolutnih koordinata scene. Rotacija kutija u skladu s geometrijom objekata smanjuje broj koraka u pretra\u017eivanju.<\/p>\n

\"Novi <\/p>\n

\u017dal, da RDNA 4 kljub vsem navedenim spremembam \u0161e vedno izvaja prehode BVH na \u0161ejdernih ALU in ne s pomo\u010djo specializirane logike. Glede na razvrstitev sredstev za strojno sledenje \u017earkom, ki jo je predlagalo podjetje Imagination Technologies, so re\u0161itve AMD obstale na drugem nivoju, grafi\u010dni procesorji Intel so na tretjem, medtem ko se NVIDIA's \u017eelezo \u017ee premika od nivoja 3 proti nivoju 4 s pomo\u010djo funkcije SER \u2014 dinami\u010dno grupiranje podatkovnih tokov z namenom pove\u010dati koherentnost dostopov do pomnilnika in posledi\u010dno zmanj\u0161ati latenco.<\/p>\n

\"Novi <\/p>\n

Analoga SER v arhitekturi RDNA 4 prav tako ni, a je bila uvedena funkcija Out of Order Memory, ki na drug na\u010din pomaga re\u0161iti isto nalogo. Prej\u0161nje razli\u010dice RDNA, ki delujejo z ve\u010d skupinami ukazov (wavefronti v terminologiji AMD), vra\u010dajo podatke iz pomnilnika strogo v istem vrstnem redu, kot so bile narejene zahteve. Ko so podatki, ki jih zahteva \u0161ejder, v hitrem predpomnilniku, se lahko vra\u010danje \u0161e vedno zavle\u010de, \u010de je predhodno svojo zahtevo podal drug \u0161ejder, katerih podatki le\u017eijo v relativno po\u010dasni VRAM. Odkrito povedano, to omejitev se zdi popolnoma nesmiselna v kontekstu ve\u010dnitnih izra\u010dunov, a se je zagotovo pojavila z namenom poenostaviti organizacijo GPU in do takrat (beri: v rasto\u010dih igrah) ostala razumljiv kompromis. RDNA 4 omogo\u010da skupinam ukazov, da vra\u010dajo podatke neodvisno od vrstnega reda zahtev.<\/p>\n

\"Novi <\/p>\n

Neispravna memorija \u2014 samo jedna od promena u arhitekturi RDNA 4. Dodeljivanje registara Compute Unit shader programima je ranije obavljano stati\u010dki, uz ra\u010dunanje na maksimum potro\u0161nje tokom \u017eivotnog ciklusa \u0161ejdera, \u0161to zna\u010di da je odre\u0111eni deo registarskog fajla zapravo ostajao prazan. RDNA 4 omogu\u0107ava \u0161ejderima da dinami\u010dki tra\u017ee i osloba\u0111aju registre. Tako je volumen registarskog fajla kori\u0161\u0107en efektnije.<\/p>\n

Zanimljivo je da je Shader Engine \u2014 velika struktura GPU-a koja kombinira niz CU, rasterizator, blok primitiva i blokove operacija rasterizacije \u2014 u RDNA 4 izgubila svoj vlastiti L1 ke\u0161. Me\u0111utim, svaki segment L2 ke\u0161a sadr\u017ei dvostruko vi\u0161e banaka. Tako je pove\u0107an ne samo njegov kapacitet, ve\u0107 i propusnost, u oba pravca \u2014 prema Shader Engine-u i prema Infinity Cache-u (drugim re\u010dima, L3). Kapacitet samog Infinity Cache-a ostao je isti kao kod Navi 32 \u2014 64 MB.<\/p>\n

Navi 48 ima 16 linija PCI Express 5. generacije, ali u igrama, naravno, ne mo\u017ee potpuno iskoristiti tako brzi su\u010delje.<\/p>\n

\"Novi <\/p>\n

Finalmente, AMD javlja da hardverski video kodek u GPU-u nove generacije osigurava pobolj\u0161anu kvalitetu kodiranja s popularnim niskim bitnim tokom za streaming, ali ne vi\u0161e od toga. To zna\u010di da podr\u0161ka za kolor subuzorkovanje YUV 4:2:2, koju imaju NVIDIA i Intel grafi\u010dki procesori, nije uvedena.<\/p>\n

\u0160to se ti\u010de sredstava za ispitivanje slike, kontroler prikaza Navi 48 kompatibilan je s najnovijim verzijama su\u010delja DisplayPort i HDMI: 2.1a i 2.1b, redom.<\/p>\n

⇡<\/a><\/noindex>#<\/a><\/noindex>Specifications, prices<\/strong><\/h2>\n

Radeon RX 9070 XT dolazi sa potpuno funkcionalnim \u010dipom Navi 48 sa 64 Compute Unit-a, \u0161to zna\u010di 4.096 aritmeti\u010dkih ALU sa standardnom precizno\u0161\u0107u. Te\u0161ko je prona\u0107i formalnog prethodnika noviteta me\u0111u 'crvenim' akceleratorima pro\u0161le generacije, jer su se promenili i modelna mre\u017ea i na\u010din numerisanja GPU-a (\u010dipovi Navi nikada ranije nisu dostigli broj osam). Po formuli ra\u010dunskih blokova i karakteristikama steka memorije, Radeon RX 9070 XT podse\u0107a na Radeon RX 7800 XT, ali AMD ga pozicionira kao zamenu za Radeon RX 7900 XTX.<\/p>\n

Game Clock in Boost Clock (target frequency in games and clock limit) of the graphics processor have significantly increased compared to the flagship model of the previous generation, reaching 2.4 and 2.97 GHz respectively. The theoretical performance of the Radeon RX 7900 XTX at maximum clock frequency is 26% higher, but AMD plans to compensate for this difference with enhancements to the RDNA 4 architecture.<\/p>\n

The Radeon RX 9070 XT features 16 GB of GDDR6 memory with a bandwidth of 20 Gbps over a 256-bit bus. Rumor has it that AMD considered increasing the VRAM to 32 GB, but ultimately deemed such a configuration excessive for consumer solutions. It\u2019s unfortunate, as the demands of games are steadily increasing, and the amount of video memory has been a crucial competitive advantage for many previous AMD devices.<\/p>\n

Referentna potro\u0161nja energije Radeon RX 9070 XT iznosi 304 W, dok je preporu\u010dena cena \u2014 599 $ (100 $ vi\u0161e nego kod Radeon RX 7800 XT).<\/p>\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
Proizvajalec<\/th>\nAMD<\/th>\nNVIDIA<\/th>\n<\/tr>\n
Model<\/td>\nRadeon RX 7900 XT<\/td>\nRadeon RX 7900 XTX<\/td>\nRadeon RX 9070<\/td>\nRadeon RX 9070 XT<\/td>\nGeForce RTX 4070 Ti SUPER<\/td>\nGeForce RTX 4080 SUPER<\/td>\nGeForce RTX 5070 Ti<\/td>\nGeForce RTX 5080<\/td>\n<\/tr>\n
Graphics Processor<\/td>\n<\/tr>\n
Ime<\/td>\nNavi 31 XT<\/td>\nNavi 31 XTX<\/td>\nNavi 48 XT<\/td>\nNavi 48 XTX<\/td>\nAD103<\/td>\nAD103<\/td>\nGB203<\/td>\nGB203<\/td>\n<\/tr>\n
Arhitektura<\/td>\nRDNA 3<\/td>\nRDNA 3<\/td>\nRDNA 4<\/td>\nRDNA 4<\/td>\nAda Lovelace<\/td>\nAda Lovelace<\/td>\nBlackwell<\/td>\nBlackwell<\/td>\n<\/tr>\n
Technology Process<\/td>\nTSMC N5\/N6<\/td>\nTSMC N5\/N6<\/td>\nTSMC N4C<\/td>\nTSMC N4C<\/td>\nTSMC 4N<\/td>\nTSMC 4N<\/td>\nTSMC 4NP<\/td>\nTSMC 4NP<\/td>\n<\/tr>\n
\u0160tevilo tranzistorjev, milijard<\/td>\n57,7<\/td>\n57,7<\/td>\n53,9<\/td>\n53,9<\/td>\n45,9<\/td>\n45,9<\/td>\n45,6<\/td>\n45,6<\/td>\n<\/tr>\n
Clock Frequency (Base Clock\/Game Clock\/Boost Clock), MHz<\/td>\n1 500 \/2 025\/2 394<\/td>\n1 855 \/2 269\/2 499<\/td>\n1 330 \/2 070\/2 520<\/td>\n1 660 \/2 400\/2 970<\/td>\n2 310\/2 610<\/td>\n2 205\/2 550<\/td>\n2 295\/2 452<\/td>\n2 295\/2 617<\/td>\n<\/tr>\n
Shader ALUs (FP32)<\/td>\n5 376 (10 752)<\/td>\n6 144 (12 228)<\/td>\n3 584 (7 168)<\/td>\n4 096 (8 192)<\/td>\n8 448<\/td>\n10 240<\/td>\n8 960<\/td>\n10 752<\/td>\n<\/tr>\n
Blokovi teksturiranja (TMU)<\/td>\n336<\/td>\n384<\/td>\n224<\/td>\n256<\/td>\n264<\/td>\n320<\/td>\n280<\/td>\n336<\/td>\n<\/tr>\n
Blokovi rasterizacije (ROP)<\/td>\n192<\/td>\n192<\/td>\n128<\/td>\n128<\/td>\n112<\/td>\n112<\/td>\n96<\/td>\n112<\/td>\n<\/tr>\n
Tensor Cores<\/td>\nNe<\/td>\nNe<\/td>\nNe<\/td>\nNe<\/td>\n264<\/td>\n320<\/td>\n280<\/td>\n336<\/td>\n<\/tr>\n
RT Cores<\/td>\n84<\/td>\n96<\/td>\n112<\/td>\n128<\/td>\n66<\/td>\n80<\/td>\n70<\/td>\n84<\/td>\n<\/tr>\n
Last Level Cache Size, MB<\/td>\n80<\/td>\n96<\/td>\n64<\/td>\n64<\/td>\n48<\/td>\n64<\/td>\n64<\/td>\n64<\/td>\n<\/tr>\n
Operativna pomnilnika<\/td>\n<\/tr>\n
Bus Width, bits<\/td>\n320<\/td>\n384<\/td>\n256<\/td>\n256<\/td>\n256<\/td>\n256<\/td>\n256<\/td>\n256<\/td>\n<\/tr>\n
Chip Type<\/td>\nGDDR6 SGRAM<\/td>\nGDDR6 SGRAM<\/td>\nGDDR6 SGRAM<\/td>\nGDDR6 SGRAM<\/td>\nGDDR6X SGRAM<\/td>\nGDDR6X SGRAM<\/td>\nGDDR7 SGRAM<\/td>\nGDDR7 SGRAM<\/td>\n<\/tr>\n
Bandwidth per Contact, Gbps<\/td>\n20<\/td>\n20<\/td>\n20<\/td>\n20<\/td>\n21<\/td>\n23<\/td>\n28<\/td>\n30<\/td>\n<\/tr>\n
Total Bandwidth, GB\/s<\/td>\n800<\/td>\n960<\/td>\n640<\/td>\n640<\/td>\n672<\/td>\n736<\/td>\n896<\/td>\n960<\/td>\n<\/tr>\n
Size, GB<\/td>\n20<\/td>\n24<\/td>\n16<\/td>\n16<\/td>\n16<\/td>\n16<\/td>\n16<\/td>\n16<\/td>\n<\/tr>\n
Performansa<\/td>\n<\/tr>\n
Peak Performance FP32, TFLOPS<\/td>\n25,7 (51,5)<\/td>\n30,7 (61,1)<\/td>\n18,1 (36,1)<\/td>\n24,3 (48,7)<\/td>\n44,1<\/td>\n52,2<\/td>\n43,9<\/td>\n56,3<\/td>\n<\/tr>\n
FP64\/FP32 Performance<\/td>\n1\/32 (1\/64)<\/td>\n1\/32 (1\/64)<\/td>\n1\/32 (1\/64)<\/td>\n1\/32 (1\/64)<\/td>\n1\/64<\/td>\n1\/64<\/td>\n1\/64<\/td>\n1\/64<\/td>\n<\/tr>\n
FP16\/FP32 Performance<\/td>\n2\/1<\/td>\n2\/1<\/td>\n2\/1<\/td>\n2\/1<\/td>\n1\/1<\/td>\n1\/1<\/td>\n1\/1<\/td>\n1\/1<\/td>\n<\/tr>\n
Ostalo<\/td>\n<\/tr>\n
PCI Express Bus<\/td>\nPCI Express 4.0 x16<\/td>\nPCI Express 4.0 x16<\/td>\nPCI Express 5.0 x16<\/td>\nPCI Express 5.0 x16<\/td>\nPCI Express 4.0 x16<\/td>\nPCI Express 4.0 x16<\/td>\n5.0 x16<\/td>\n5.0 x16<\/td>\n<\/tr>\n
Image Output Interfaces<\/td>\nDisplayPort 2.1, HDMI 2.1a<\/td>\nDisplayPort 2.1, HDMI 2.1a<\/td>\nDisplayPort 2.1a, HDMI 2.1b<\/td>\nDisplayPort 2.1a, HDMI 2.1b<\/td>\nDisplayPort 1.4a, HDMI 2.1<\/td>\nDisplayPort 1.4a, HDMI 2.1<\/td>\nDisplayPort 2.1b, HDMI 2.1b<\/td>\nDisplayPort 2.1b, HDMI 2.1b<\/td>\n<\/tr>\n
TDP\/TBP, W<\/td>\n315<\/td>\n335<\/td>\n220<\/td>\n304<\/td>\n285<\/td>\n320<\/td>\n300<\/td>\n360<\/td>\n<\/tr>\n
Retail Price (USA), $<\/td>\n899 (recommended at launch)<\/td>\n999 (recommended at launch)<\/td>\n549 (recommended at launch)<\/td>\n599 (recommended at launch)<\/td>\n799 (recommended at launch)<\/td>\n999 (recommended at launch)<\/td>\n749 (recommended at launch)<\/td>\n999 (recommended at launch)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

S druge strane, GeForce RTX 5070 Ti se zasniva na istom grafi\u010dkom procesoru GB203, koji se koristi u GeForce RTX 5080, ali je mla\u0111i model izgubio 14 multiprocesora, odnosno 1.792 FP32-kompatibilnih \u0161ejderskih ALU, a radna taktok frekvencija je smanjena. Kao rezultat, GeForce RTX 5080 se od RTX 5070 Ti odvaja sa 27% teoretskog ubrzanja, a sam GeForce RTX 5070 Ti se po ovom kriterijumu ne razlikuje od RTX 4070 Ti SUPER. Kao i u slu\u010daju Radeon RX 9070 XT, nedostatak sirove ra\u010dunarske snage treba da nadoknadi nova arhitektura GPU-a, a pre svega \u2014 funkcija generisanja vi\u0161e frejmova putem DLSS (do tri generisana frejma izme\u0111u dva 'stvarna'). <\/p>\n

GeForce RTX 5070 Ti, 300 W g\u00fc\u00e7 t\u00fcketimi i\u00e7in tasarlanm\u0131\u015f olup MSRP $749 ile GeForce RTX 4070 Ti SUPER'den $50 daha ucuzdur.<\/p>\n

Razlika od $150 preporu\u010dene valute izme\u0111u Radeon RX 9070 XT i GeForce RTX 5070 Ti nije primetna prema trenutnim ruskim cenama: najpristupa\u010dnije partnerske modele prona\u0161li smo u DNS-u po cijenama od 99.999 i 104.999 rubalja, redom. Upravo \u0107emo njih testirati u benchmarkovima. Radeon RX 9070 XT predstavlja grafi\u010dka kartica GIGABYTE GAMING OC sa fabri\u010dki overklokiranim GPU do 2,52\/3,06 GHz, dok je GeForce RTX 5070 Ti \u2013 ubrziva\u010d Palit GamingPro sa referentnim specifikacijama.<\/p>\n

⇡<\/a><\/noindex>#<\/a><\/noindex>GIGABYTE Radeon RX 9070 XT GAMING OC: Tasar\u0131m<\/strong><\/h2>\n

Radeon RX 9070 XT u izdanju GIGABYTE ima prili\u010dno skromne dimenzije za grafi\u010dku karticu sa potro\u0161njom energije ne manju (i, kako \u0107e pokazati testovi, mnogo ve\u0107u) od 300 W: 288 mm du\u017eine i tri slota \u0161irine. Jedini element LED osvetljenja je kratka traka sa strane. Oznaka sa logotipom proizvo\u0111a\u010da \u2013 pomi\u010dna: ako je pomerite na LED trakicu, logotip po\u010dinje da svetli.<\/p>\n

\"Novi <\/p>\n

So\u011futma sistemini 87 mm \u00e7ap\u0131nda \u00fc\u00e7 fan hizmet ediyor. K\u00fc\u00e7\u00fck bir radyat\u00f6r b\u00f6l\u00fcm\u00fc arac\u0131l\u0131\u011f\u0131yla hava, arka plakan\u0131n kesik k\u0131sm\u0131na kadar ge\u00e7er.<\/p>\n

\"Novi <\/p>\n

Grafi\u010dki procesor i \u010dipovi video memorije osloba\u0111aju toplotu isparavaju\u0107oj komori, ali prvi \u2013 direktno, dok drugi \u2013 preko U-oblikovane metalne podlo\u0161ke. Na lamelama hladnjaka su pri\u010dvr\u0161\u0107ene odvojene plo\u010de koje slu\u017ee za odvo\u0111enje toplote komponenti VRM-a. Toplotnih cevi ima ukupno sedam, a njihov pre\u010dnik je 6 mm. Zapa\u017eamo da su GIGABYTE odabrali te\u010dne termalne podlo\u0161ke umesto klasi\u010dnih. Mo\u017eda je upravo to razlog za\u0161to se chipovi VRAM-a pregrevaju pod optere\u0107enjem (vidi empirijski deo pregleda).<\/p>\n

\"Novi <\/p>\n

Zadnja plo\u0161\u010da sodeluje pri odvajanju toplote s tiskanega vezja s pomo\u010djo ve\u010d termopadov (tukaj obi\u010dajnih, ne teko\u010dih).<\/p>\n

\"Novi <\/p>\n

⇡<\/a><\/noindex>#<\/a><\/noindex>GIGABYTE Radeon RX 9070 XT GAMING OC: tiskan vezje<\/strong><\/h2>\n

Neki varijante Radeon RX 9070 XT napajaju se putem konektora 12V-2\u00d76, ali GIGABYTE je odabrala tri univerzalna osmostruka konektora. S obzirom na visoku potro\u0161nju energije grafi\u010dke kartice i neprekidne pritu\u017ebe na novi konektor, to je pouzdanije re\u0161enje, iako nije tako prakti\u010dno i estetski privla\u010dno. Od 17 faza sistema napajanja, 14, pod upravom PWM kontrolera Monolithic Power Systems MP2868A, dodelju se grafi\u010dkom procesoru, dok su tri rezervisane za video memoriju. Oba VRM-a opremljena su naponskim kaskadama MP87993 sa pretpostavljenim nominalnim strujnim kapacitetom od 90A.<\/p>\n

VRAM je sestavljen iz \u010dipov SK hynix, v njeni oznaki H56G42AS8DX-014 je kodirana propustnost 20 Gbit\/s.<\/p>\n

Pored kopije BIOS-a, koja je aktivna po defaultu, grafi\u010dka kartica ima rezervnu 'tihu' verziju firmvera.<\/p>\n

\"Novi \"Novi <\/p>\n

⇡<\/a><\/noindex>#<\/a><\/noindex>Palit GeForce RTX 5070 Ti GamingPro: konstrukcija<\/strong><\/h2>\n

Palit GeForce RTX 5070 Ti GamingPro zauzima 331,9 mm du\u017eine, ali samo tri pro\u0161irena slota. Ispod logotipa GamingPro i susednog segmenta prednje panele su RGB LED svetla. Rad LED svetla mo\u017ee se sinhronizovati sa mati\u010dnom plo\u010dom putem standardnog ARGB konektora koji se nalazi pored ulaza za napajanje 12V-2\u00d76.<\/p>\n

\"Novi <\/p>\n

Stranske povr\u0161ine ohi\u0161ja tvori lit aluminijasta okvirja z ventilacijskimi re\u017eami na dolgih straneh. V za\u0161\u010ditni plo\u0161\u010di na zadnji strani PCB je obi\u010dajna mre\u017ea za pre\u010drpavanje zraka iz hladilnika. Impellerji ventilatorjev imajo premer 94 mm.<\/p>\n

\"Novi <\/p>\n

\u041d\u0430 \u0434nu \u043a\u0443\u043b\u0435\u0440\u0430 \u043d\u0430\u0445\u043e\u0434\u0438\u0442\u0441\u044f \u0438\u0441\u043f\u0430\u0440\u0438\u0442\u0435\u043b\u044c\u043d\u0430\u044f \u043a\u0430\u043c\u0435\u0440\u0430 \u0434\u043e\u0441\u0442\u0430\u0442\u043e\u0447\u043d\u043e\u0439 \u043f\u043b\u043e\u0449\u0430\u0434\u0438, \u0447\u0442\u043e\u0431\u044b \u043f\u043e\u043a\u0440\u044b\u0432\u0430\u0442\u044c \u043a\u0440\u0438\u0441\u0442\u0430\u043b\u043b GPU \u0438 \u043e\u043a\u0440\u0443\u0436\u0430\u044e\u0449\u0438\u0435 \u0435\u0433\u043e \u043c\u0438\u043a\u0440\u043e\u0441\u0445\u0435\u043c\u044b \u0432\u0438\u0434\u0435\u043e\u043f\u0430\u043c\u044f\u0442\u0438, \u0430 \u0434\u043b\u044f \u043a\u043e\u043c\u043f\u043e\u043d\u0435\u043d\u0442\u043e\u0432 VRM \u043f\u0440\u0435\u0434\u0443\u0441\u043c\u043e\u0442\u0440\u0435\u043d\u044b \u043e\u0442\u0434\u0435\u043b\u044c\u043d\u044b\u0435 \u043f\u043b\u0430\u0441\u0442\u0438\u043d\u044b \u0434\u043b\u044f \u0442\u0435\u043f\u043b\u043e\u043e\u0442\u0432\u0435\u0434\u0435\u043d\u0438\u044f. \u041b\u0430\u043c\u0435\u043b\u0438 \u0440\u0430\u0434\u0438\u0430\u0442\u043e\u0440\u0430 \u043d\u0430\u043d\u0438\u0437\u0430\u043d\u044b \u043d\u0430 \u0432\u043e\u0441\u0435\u043c\u044c \u0442\u0435\u043f\u043b\u043e\u0432\u044b\u0445 \u0442\u0440\u0443\u0431\u043e\u043a \u0434\u0438\u0430\u043c\u0435\u0442\u0440\u043e\u043c 6 \u043c\u043c.<\/p>\n

\"Novi <\/p>\n

\u041c\u0435\u0442\u0430\u043b\u043b\u0438\u0447\u0435\u0441\u043a\u0430\u044f \u0437\u0430\u0434\u043d\u044f\u044f \u043f\u0430\u043d\u0435\u043b\u044c \u0432\u044b\u043f\u043e\u043b\u043d\u044f\u0435\u0442 \u0442\u043e\u043b\u044c\u043a\u043e \u0437\u0430\u0449\u0438\u0442\u043d\u0443\u044e \u0438 \u0434\u0435\u043a\u043e\u0440\u0430\u0442\u0438\u0432\u043d\u0443\u044e \u0444\u0443\u043d\u043a\u0446\u0438\u044e, \u0442\u0430\u043a \u043a\u0430\u043a \u043c\u0435\u0436\u0434\u0443 \u043d\u0435\u0439 \u0438 PCB \u043d\u0435\u0442 \u043d\u0438 \u043e\u0434\u043d\u043e\u0439 \u0442\u0435\u0440\u043c\u043e\u043f\u0440\u043e\u043a\u043b\u0430\u0434\u043a\u0438.<\/p>\n

\"Novi <\/p>\n

Grafi\u010dnoj kartici prilo\u017een je adapter sa dva osmo-pin konektora za napajanje na 12V-2\u00d76 i mali tkaninski podlog za mi\u0161.<\/p>\n

\"Novi <\/p>\n

⇡<\/a><\/noindex>#<\/a><\/noindex>Palit GeForce RTX 5070 Ti GamingPro: \u043f\u0435\u0447\u0430\u0442\u043d\u0430\u044f \u043f\u043b\u0430\u0442\u0430<\/strong><\/h2>\n

\u0412 \u043e\u0442\u043b\u0438\u0447\u0438\u0435 \u043e\u0442 \u0432\u0438\u0434\u0435\u043e\u043a\u0430\u0440\u0442\u044b Palit GameRock<\/a><\/noindex>, koju smo pregledali u nedavnom pregledu GeForce RTX 5080, slede\u0107i model serije sastavljen je na prostornom PCB-u, ali je njen napajanje jednostavnije. Grafi\u010dkom procesoru pripada 15-fazni VRM pod upravom PWM kontrolera Monolithic Power Systems MP29816. Napajanje memorijskih \u010dipova GDDR7 \u2013 trofazno, na bazi kontrolera MP2988. Oba regulatora opremljena su tranzistorskim sklopovima MP87993 sa pretpostavljenim nominalnim strujnim optere\u0107enjem od 90 A.<\/p>\n

\u041c\u0430\u0440\u043a\u0438\u0440\u043e\u0432\u043a\u0430 \u0447\u0438\u043f\u043e\u0432 \u0432\u0438\u0434\u0435\u043e\u043f\u0430\u043c\u044f\u0442\u0438 Samsung K4VAF325ZC-SC28 \u0443\u043a\u0430\u0437\u044b\u0432\u0430\u0435\u0442 \u043d\u0430 \u043f\u0440\u043e\u043f\u0443\u0441\u043a\u043d\u0443\u044e \u0441\u043f\u043e\u0441\u043e\u0431\u043d\u043e\u0441\u0442\u044c 28 \u0413\u0431\u0438\u0442\/\u0441.<\/p>\n

Kao i Radeon RX 9070 XT u izvedbi GIGABYTE, akcelerator Palit ima dve verzije BIOS-a \u2013 \"performans\" i \"tiha\".<\/p>\n

\"Novi \"Novi <\/p>\n

⇡<\/a><\/noindex>#<\/a><\/noindex>\u0422\u0435\u0441\u0442\u043e\u0432\u0430\u044f \u043f\u043b\u0430\u0442\u0444\u043e\u0440\u043c\u0430, \u043c\u0435\u0442\u043e\u0434\u043e\u043b\u043e\u0433\u0438\u044f \u0442\u0435\u0441\u0442\u0438\u0440\u043e\u0432\u0430\u043d\u0438\u044f<\/strong><\/h2>\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
Testni postavitev<\/th>\n<\/tr>\n
CPU<\/td>\nAMD Ryzen 9 7950X3D (PBO +150 \u041c\u0413\u0446, CU -20)<\/td>\n<\/tr>\n
\u041c\u0430\u0442\u0435\u0440\u0438\u043d\u0441\u043a\u0430\u044f \u043f\u043b\u0430\u0442\u0430<\/td>\nASUS ROG Crosshair X670E Hero<\/td>\n<\/tr>\n
Operativna pomnilnika<\/td>\nG.Skill Trident Z5 Neo RGB (F5-6000J3040G32GX2-TZ5NR), 2 \u00d7 32 GB (6200 MT\/s, CL30)<\/td>\n<\/tr>\n
\u041f\u0417\u0423<\/td>\nSolidigm P44 Pro, 2 \u0422\u0431\u0430\u0439\u0442<\/td>\n<\/tr>\n
\u0411\u043b\u043e\u043a \u043f\u0438\u0442\u0430\u043d\u0438\u044f<\/td>\nCorsair AX1600i, 1600 \u0412\u0442<\/td>\n<\/tr>\n
\u0421\u0438\u0441\u0442\u0435\u043c\u0430 \u043e\u0445\u043b\u0430\u0436\u0434\u0435\u043d\u0438\u044f \u0426\u041f<\/td>\nKustomni WC (EK-Quantum Velocity\u00b2 DDC 4.2 PWM D-RGB + EK-Quantum Surface X280M)<\/td>\n<\/tr>\n
\u041a\u043e\u0440\u043f\u0443\u0441<\/td>\n\u041e\u0442\u043a\u0440\u044b\u0442\u044b\u0439 \u0441\u0442\u0435\u043d\u0434<\/td>\n<\/tr>\n
Operacijski sistem<\/td>\nWindows 11 Pro<\/td>\n<\/tr>\n
Programska oprema za GPU AMD<\/td>\n<\/tr>\n
Vse grafi\u010dne kartice<\/td>\nAMD Software Adrenalin Edition 25.2.1\/25.3.1<\/td>\n<\/tr>\n
\u041f\u041e \u0434\u043b\u044f GPU NVIDIA<\/td>\n<\/tr>\n
Vse grafi\u010dne kartice<\/td>\nNVIDIA GeForce Game Ready Driver 532.60\/532.70<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
\u0418\u0433\u0440\u044b \u0431\u0435\u0437 \u0442\u0440\u0430\u0441\u0441\u0438\u0440\u043e\u0432\u043a\u0438 \u043b\u0443\u0447\u0435\u0439<\/th>\n<\/tr>\n
\u0418\u0433\u0440\u0430<\/td>\nAPI<\/td>\n\u041c\u0435\u0442\u043e\u0434 \u0442\u0435\u0441\u0442\u0438\u0440\u043e\u0432\u0430\u043d\u0438\u044f<\/td>\n\u041d\u0430\u0441\u0442\u0440\u043e\u0439\u043a\u0438 \u0433\u0440\u0430\u0444\u0438\u043a\u0438<\/td>\n<\/tr>\n
Alan Wake 2<\/td>\nDirectX 12<\/td>\nOCAT, \u043b\u043e\u043a\u0430\u0446\u0438\u044f Bright Falls<\/td>\n\u041c\u0430\u043a\u0441. \u043a\u0430\u0447\u0435\u0441\u0442\u0432\u043e \u0433\u0440\u0430\u0444\u0438\u043a\u0438<\/td>\n<\/tr>\n
Black Myth: Wukong<\/td>\nDirectX 12<\/td>\n\u0412\u0441\u0442\u0440\u043e\u0435\u043d\u043d\u044b\u0439 \u0431\u0435\u043d\u0447\u043c\u0430\u0440\u043a<\/td>\n\u041c\u0430\u043a\u0441. \u043a\u0430\u0447\u0435\u0441\u0442\u0432\u043e \u0433\u0440\u0430\u0444\u0438\u043a\u0438<\/td>\n<\/tr>\n
Cyberpunk 2077<\/td>\nDirectX 12<\/td>\n\u0412\u0441\u0442\u0440\u043e\u0435\u043d\u043d\u044b\u0439 \u0431\u0435\u043d\u0447\u043c\u0430\u0440\u043a<\/td>\n\u041c\u0430\u043a\u0441. \u043a\u0430\u0447\u0435\u0441\u0442\u0432\u043e \u0433\u0440\u0430\u0444\u0438\u043a\u0438<\/td>\n<\/tr>\n
F1 24<\/td>\nDirectX 12<\/td>\n\u0412\u0441\u0442\u0440\u043e\u0435\u043d\u043d\u044b\u0439 \u0431\u0435\u043d\u0447\u043c\u0430\u0440\u043a, \u0442\u0440\u0430\u0441\u0441\u0430 Monaco (\u0434\u043e\u0436\u0434\u044c)<\/td>\n\u041c\u0430\u043a\u0441. \u043a\u0430\u0447\u0435\u0441\u0442\u0432\u043e \u0433\u0440\u0430\u0444\u0438\u043a\u0438<\/td>\n<\/tr>\n
Hogwarts Legacy<\/td>\nDirectX 12<\/td>\nOCAT, \u043f\u043e\u0435\u0437\u0434\u043a\u0430 \u043d\u0430 \u0442\u0435\u043b\u0435\u0436\u043a\u0435 \u0432 Path to Hogwarts<\/td>\n\u041c\u0430\u043a\u0441. \u043a\u0430\u0447\u0435\u0441\u0442\u0432\u043e \u0433\u0440\u0430\u0444\u0438\u043a\u0438<\/td>\n<\/tr>\n
Horizon Zero Dawn Remastered<\/td>\nDirectX 12<\/td>\n\u0412\u0441\u0442\u0440\u043e\u0435\u043d\u043d\u044b\u0439 \u0431\u0435\u043d\u0447\u043c\u0430\u0440\u043a<\/td>\n\u041c\u0430\u043a\u0441. \u043a\u0430\u0447\u0435\u0441\u0442\u0432\u043e \u0433\u0440\u0430\u0444\u0438\u043a\u0438<\/td>\n<\/tr>\n
Metro Exodus<\/td>\nDirectX 12<\/td>\n\u0412\u0441\u0442\u0440\u043e\u0435\u043d\u043d\u044b\u0439 \u0431\u0435\u043d\u0447\u043c\u0430\u0440\u043a<\/td>\n\u041c\u0430\u043a\u0441. \u043a\u0430\u0447\u0435\u0441\u0442\u0432\u043e \u0433\u0440\u0430\u0444\u0438\u043a\u0438; Shading Rate: 100%<\/td>\n<\/tr>\n
Red Dead Redemption 2<\/td>\nVulkan<\/td>\n\u0412\u0441\u0442\u0440\u043e\u0435\u043d\u043d\u044b\u0439 \u0431\u0435\u043d\u0447\u043c\u0430\u0440\u043a<\/td>\n\u041c\u0430\u043a\u0441. \u043a\u0430\u0447\u0435\u0441\u0442\u0432\u043e \u0433\u0440\u0430\u0444\u0438\u043a\u0438<\/td>\n<\/tr>\n
Returnal<\/td>\nDirectX 12<\/td>\n\u0412\u0441\u0442\u0440\u043e\u0435\u043d\u043d\u044b\u0439 \u0431\u0435\u043d\u0447\u043c\u0430\u0440\u043a<\/td>\n\u041c\u0430\u043a\u0441. \u043a\u0430\u0447\u0435\u0441\u0442\u0432\u043e \u0433\u0440\u0430\u0444\u0438\u043a\u0438<\/td>\n<\/tr>\n
Total War: WARHAMMER III<\/td>\nDirectX 11<\/td>\n\u0412\u0441\u0442\u0440\u043e\u0435\u043d\u043d\u044b\u0439 \u0431\u0435\u043d\u0447\u043c\u0430\u0440\u043a (Mirrors of Madness Benchmark)<\/td>\n\u041c\u0430\u043a\u0441. \u043a\u0430\u0447\u0435\u0441\u0442\u0432\u043e \u0433\u0440\u0430\u0444\u0438\u043a\u0438<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n\n\n\n\n\n\n\n\n\n\n\n\n\n
\u0418\u0433\u0440\u044b \u0441 \u0442\u0440\u0430\u0441\u0441\u0438\u0440\u043e\u0432\u043a\u043e\u0439 \u043b\u0443\u0447\u0435\u0439<\/th>\n<\/tr>\n
\u0418\u0433\u0440\u0430<\/td>\nAPI<\/td>\n\u041c\u0435\u0442\u043e\u0434 \u0442\u0435\u0441\u0442\u0438\u0440\u043e\u0432\u0430\u043d\u0438\u044f<\/td>\n\u041d\u0430\u0441\u0442\u0440\u043e\u0439\u043a\u0438 \u0433\u0440\u0430\u0444\u0438\u043a\u0438<\/td>\n\u041c\u0430\u0441\u0448\u0442\u0430\u0431\u0438\u0440\u043e\u0432\u0430\u043d\u0438\u0435 \u043a\u0430\u0434\u0440\u043e\u0432<\/td>\n<\/tr>\n
AMD<\/td>\nIntel<\/td>\nNVIDIA<\/td>\n<\/tr>\n
Alan Wake 2<\/td>\nDirectX 12<\/td>\nOCAT, \u043b\u043e\u043a\u0430\u0446\u0438\u044f Bright Falls<\/td>\n\u041c\u0430\u043a\u0441. \u043a\u0430\u0447\u0435\u0441\u0442\u0432\u043e \u0433\u0440\u0430\u0444\u0438\u043a\u0438, \u0432\u044b\u0441\u043e\u043a\u043e\u0435 \u043a\u0430\u0447\u0435\u0441\u0442\u0432\u043e \u0442\u0440\u0430\u0441\u0441\u0438\u0440\u043e\u0432\u043a\u0438 \u043b\u0443\u0447\u0435\u0439<\/td>\nFSR Balanced<\/td>\nFSR Balanced<\/td>\nDLSS Balanced + Ray Reconstruction (+ Frame Generation)<\/td>\n<\/tr>\n
Black Myth: Wukong<\/td>\n\u0412\u0441\u0442\u0440\u043e\u0435\u043d\u043d\u044b\u0439 \u0431\u0435\u043d\u0447\u043c\u0430\u0440\u043a<\/td>\n\u041c\u0430\u043a\u0441. \u043a\u0430\u0447\u0435\u0441\u0442\u0432\u043e \u0433\u0440\u0430\u0444\u0438\u043a\u0438 \u0438 \u0442\u0440\u0430\u0441\u0441\u0438\u0440\u043e\u0432\u043a\u0438 \u043b\u0443\u0447\u0435\u0439 (Path Tracing)<\/td>\nFSR Balanced (+ Frame Generation)<\/td>\nXeSS Balanced\/FSR Balanced + Frame Generation<\/td>\nDLSS Balanced (+ Frame Generation)<\/td>\n<\/tr>\n
Cyberpunk 2077<\/td>\n\u0412\u0441\u0442\u0440\u043e\u0435\u043d\u043d\u044b\u0439 \u0431\u0435\u043d\u0447\u043c\u0430\u0440\u043a (OCAT \u0434\u043b\u044f \u0433\u0435\u043d\u0435\u0440\u0430\u0446\u0438\u0438 \u043a\u0430\u0434\u0440\u043e\u0432)<\/td>\n\u041c\u0430\u043a\u0441. \u043a\u0430\u0447\u0435\u0441\u0442\u0432\u043e \u0433\u0440\u0430\u0444\u0438\u043a\u0438 \u0438 \u0442\u0440\u0430\u0441\u0441\u0438\u0440\u043e\u0432\u043a\u0438 \u043b\u0443\u0447\u0435\u0439 (Path Tracing)<\/td>\nFSR Balanced (+ Frame Generation)<\/td>\nXeSS Balanced\/FSR Balanced + Frame Generation<\/td>\nDLSS Balanced (Transformer Model) + Ray Reconstruction (+ Frame Generation)<\/td>\n<\/tr>\n
F1 24<\/td>\n\u0412\u0441\u0442\u0440\u043e\u0435\u043d\u043d\u044b\u0439 \u0431\u0435\u043d\u0447\u043c\u0430\u0440\u043a, \u0442\u0440\u0430\u0441\u0441\u0430 Monaco (\u0434\u043e\u0436\u0434\u044c)<\/td>\n\u041c\u0430\u043a\u0441. \u043a\u0430\u0447\u0435\u0441\u0442\u0432\u043e \u0433\u0440\u0430\u0444\u0438\u043a\u0438 \u0438 \u0442\u0440\u0430\u0441\u0441\u0438\u0440\u043e\u0432\u043a\u0438 \u043b\u0443\u0447\u0435\u0439<\/td>\nFSR Balanced (+ Frame Generation)<\/td>\nXeSS Balanced\/FSR Balanced + Frame Generation<\/td>\nDLSS Balanced (+ Frame Generation)<\/td>\n<\/tr>\n
Hogwarts Legacy<\/td>\nOCAT, \u043f\u043e\u0435\u0437\u0434\u043a\u0430 \u043d\u0430 \u0442\u0435\u043b\u0435\u0436\u043a\u0435 \u0432 Path to Hogwarts<\/td>\n\u041c\u0430\u043a\u0441. \u043a\u0430\u0447\u0435\u0441\u0442\u0432\u043e \u0433\u0440\u0430\u0444\u0438\u043a\u0438 \u0438 \u0442\u0440\u0430\u0441\u0441\u0438\u0440\u043e\u0432\u043a\u0438 \u043b\u0443\u0447\u0435\u0439<\/td>\nFSR Balanced<\/td>\nXeSS Balanced<\/td>\nDLSS Balanced + Ray Reconstruction (+ Frame Generation)<\/td>\n<\/tr>\n
Indiana Jones in the Great Circle<\/td>\nOCAT, location Sukhothai<\/td>\nMax. graphics quality, high-quality ray tracing (Path Tracing); Ray Traced Shadows: Only Sunlight<\/td>\nFSR Balanced (+ Frame Generation)<\/td>\nXeSS Balanced\/FSR Balanced + Frame Generation<\/td>\nDLSS Balanced (Transformer Model) + Ray Reconstruction (+ Frame Generation)<\/td>\n<\/tr>\n
Metro Exodus Enhanced Edition<\/td>\n\u0412\u0441\u0442\u0440\u043e\u0435\u043d\u043d\u044b\u0439 \u0431\u0435\u043d\u0447\u043c\u0430\u0440\u043a<\/td>\n\u041c\u0430\u043a\u0441. \u043a\u0430\u0447\u0435\u0441\u0442\u0432\u043e \u0433\u0440\u0430\u0444\u0438\u043a\u0438 \u0438 \u0442\u0440\u0430\u0441\u0441\u0438\u0440\u043e\u0432\u043a\u0438 \u043b\u0443\u0447\u0435\u0439<\/td>\nN\/A<\/td>\nN\/A<\/td>\nDLSS Balanced<\/td>\n<\/tr>\n
Returnal<\/td>\n\u0412\u0441\u0442\u0440\u043e\u0435\u043d\u043d\u044b\u0439 \u0431\u0435\u043d\u0447\u043c\u0430\u0440\u043a (OCAT \u0434\u043b\u044f \u0433\u0435\u043d\u0435\u0440\u0430\u0446\u0438\u0438 \u043a\u0430\u0434\u0440\u043e\u0432)<\/td>\n\u041c\u0430\u043a\u0441. \u043a\u0430\u0447\u0435\u0441\u0442\u0432\u043e \u0433\u0440\u0430\u0444\u0438\u043a\u0438 \u0438 \u0442\u0440\u0430\u0441\u0441\u0438\u0440\u043e\u0432\u043a\u0438 \u043b\u0443\u0447\u0435\u0439<\/td>\nFSR Balanced (+ Frame Generation)<\/td>\nXeSS Balanced\/FSR Balanced + Frame Generation<\/td>\nDLSS Balanced (+ Frame Generation)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

In most games, the average and minimum frame rates (we indicate the 1st percentile of the distribution) are derived from the array of rendering times of individual frames or the instantaneous framerate obtained using the built-in benchmark. Exceptions are games that do not have a built-in benchmark and tests using frame generation: in these cases, we use the OCAT program to capture inter-frame intervals.<\/p>\n\n\n\n\n\n\n\n\n
Delovne aplikacije<\/th>\n<\/tr>\n
Aplikacija<\/td>\nBenchmark<\/td>\nNastavitve<\/td>\n<\/tr>\n
Adobe Premiere Pro 25.x<\/td>\nPugetBench for Premiere Pro 1.1.1 (test composition linked)<\/a><\/noindex><\/td>\nStandard (4K)<\/td>\n<\/tr>\n
Blender 4.x<\/td>\nAgent 327 Barbershop demo from the Blender website<\/td>\nRenderer Cycles<\/td>\n<\/tr>\n
Blackmagic Design DaVinci Resolve Studio 19.x<\/td>\nPugetBench for DaVinci Resolve 1.1.1 (test composition linked)<\/a><\/noindex><\/td>\nStandard (4K); H.264\/HEVC Encoding Mode: Auto<\/td>\n<\/tr>\n
CAD applications<\/td>\nSPECviewperf 2020 v3.1<\/td>\nRezolucija ekrana: 3840 \u00d7 2160<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
Video decoding (ffmpeg 5.x)<\/th>\n<\/tr>\n
Format<\/td>\nLo\u010dljivost<\/td>\nEncoding options<\/td>\nAPI<\/td>\n<\/tr>\n
H.264 (YUV 4:2:0, 8 bits\/channel)<\/td>\n1920 \u00d7 1080<\/td>\nHigh Profile, L4.1<\/td>\nD3D11VA<\/td>\n<\/tr>\n
3840 \u00d7 2160<\/td>\nHigh Profile, L5.1<\/td>\n<\/tr>\n
HEVC (YUV 4:2:0, 8 bits\/channel)<\/td>\n1920 \u00d7 1080<\/td>\nMain Profile, L4.0<\/td>\n<\/tr>\n
3840 \u00d7 2160<\/td>\nMain Profile, L5.0<\/td>\n<\/tr>\n
7680 \u00d7 4320<\/td>\nMain Profile, L6.0<\/td>\n<\/tr>\n
VP9 (YUV 4:2:0, 8 bits\/channel)<\/td>\n1920 \u00d7 1080<\/td>\nN\/A<\/td>\n<\/tr>\n
3840 \u00d7 2160<\/td>\n<\/tr>\n
7680 \u00d7 4320<\/td>\n<\/tr>\n
AV1 (YUV 4:2:0, 8 bits\/channel)<\/td>\n1920 \u00d7 1080<\/td>\nMain Profile, L4.0<\/td>\n<\/tr>\n
3840 \u00d7 2160<\/td>\nMain Profile, L5.0<\/td>\n<\/tr>\n
7680 \u00d7 4320<\/td>\nMain Profile, L6.0<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n\n\n\n\n\n\n\n\n\n\n\n\n\n
Video encoding (ffmpeg 5.x)<\/th>\n<\/tr>\n
Format<\/td>\nLo\u010dljivost<\/td>\nEncoding options<\/td>\nAPI<\/td>\n<\/tr>\n
AMD<\/td>\nIntel<\/td>\nNVIDIA<\/td>\nAMD<\/td>\nIntel<\/td>\nNVIDIA<\/td>\n<\/tr>\n
H.264 (YUV 4:2:0, 8 bits\/channel)<\/td>\n1920 \u00d7 1080<\/td>\n-c:v h264_amf -quality speed -coder cabac -refs 1 -b:v 3M<\/td>\n-c:v h264_qsv -preset veryfast -profile:v main -level 4.1 -b:v 3M<\/td>\n-c:v h264_nvenc -preset fast -coder cabac -refs 1 -b:v 3M<\/td>\nAMF<\/td>\noneVPL<\/td>\nNVENC<\/td>\n<\/tr>\n
3840 \u00d7 2160<\/td>\n-c:v h264_amf -quality speed -coder cabac -refs 1 -b:v 7.5M<\/td>\n-c:v h264_qsv -preset veryfast -profile:v main -level 5.1 -b:v 7.5M<\/td>\n-c:v h264_nvenc -preset fast -coder cabac -refs 1 -b:v 7.5M<\/td>\n<\/tr>\n
HEVC (YUV 4:2:0, 8 bits\/channel)<\/td>\n1920 \u00d7 1080<\/td>\n-c:v hevc_amf -quality speed -b:v 3M<\/td>\n-c:v hevc_qsv -preset veryfast -tier main -b:v 3M<\/td>\n-c:v hevc_nvenc -preset fast -b:v 3M<\/td>\n<\/tr>\n
3840 \u00d7 2160<\/td>\n-c:v hevc_amf -quality speed -b:v 7.5M<\/td>\n-c:v hevc_qsv -preset veryfast -tier main -b:v 7.5M<\/td>\n-c:v hevc_nvenc -preset fast -b:v 7.5M<\/td>\n<\/tr>\n
7680 \u00d7 4320<\/td>\n-c:v hevc_amf -quality speed -b:v 20M<\/td>\n-c:v hevc_qsv -preset veryfast -tier main -b:v 20M<\/td>\n-c:v hevc_nvenc -preset fast -b:v 20M<\/td>\n<\/tr>\n
AV1 (YUV 4:2:0, 8 bits\/channel)<\/td>\n1920 \u00d7 1080<\/td>\n-c:v hevc_amf -quality speed -b:v 3M<\/td>\n-c:v av1_qsv -preset veryfast -profile main -b:v 3M<\/td>\n-c:v hevc_nvenc -preset fast -b:v 3M<\/td>\n<\/tr>\n
3840 \u00d7 2160<\/td>\n-c:v hevc_amf -quality speed -b:v 7.5M<\/td>\n-c:v av1_qsv -preset veryfast -profile main -b:v 7.5M<\/td>\n-c:v hevc_nvenc -preset fast -b:v 7.5M<\/td>\n<\/tr>\n
7680 \u00d7 4320<\/td>\n-c:v hevc_amf -quality speed -b:v 20M<\/td>\n-c:v av1_qsv -preset veryfast -profile main -b:v 20M<\/td>\n-c:v hevc_nvenc -preset fast -b:v 20M<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

Mo\u0107 grafi\u010dkih kartica registrovana je odvojeno od CPU-a i drugih komponenti PC-a pomo\u0107u ure\u0111aja NVIDIA PCAT. Kao optere\u0107enje za testove mo\u0107i i nivoa buke koristi se igra Cyberpunk 2077 pri rezoluciji 3840 \u00d7 2160 i maksimalnim parametrima kvaliteta grafike (bez ray tracing-a), kao i stres test FurMark s najagresivnijim postavkama (rezolucija 3840 \u00d7 2160, MSAA 8x). Merenja svih parametara obavljaju se nakon zagrevanja grafi\u010dke kartice, kada se temperatura GPU-a i taktok frekvencije stabilizuju.<\/p>\n

⇡<\/a><\/noindex>#<\/a><\/noindex>\u0423\u0447\u0430\u0441\u0442\u043d\u0438\u043a\u0438 \u0442\u0435\u0441\u0442\u0438\u0440\u043e\u0432\u0430\u043d\u0438\u044f<\/strong><\/h2>\n

\u0412 \u0442\u0435\u0441\u0442\u0438\u0440\u043e\u0432\u0430\u043d\u0438\u0438 \u043f\u0440\u043e\u0438\u0437\u0432\u043e\u0434\u0438\u0442\u0435\u043b\u044c\u043d\u043e\u0441\u0442\u0438 \u0443\u0447\u0430\u0441\u0442\u0432\u043e\u0432\u0430\u043b\u0438 \u0441\u043b\u0435\u0434\u0443\u044e\u0449\u0438\u0435 \u0432\u0438\u0434\u0435\u043e\u043a\u0430\u0440\u0442\u044b:<\/p>\n