Ƙayyadaddun Iyali na Blackwell GPU da Alamomi ya bayyana a sarari cewa duk matsayi na 50, ban da GeForce RTX 5090, suna wakiltar ƙarancin haɓakawa mai laushi daga daidaitattun samfuran ƙarshe. Lamarin ya kara tabarbare sakamakon karancin sabbin katunan bidiyo da kuma sakamakon tsadar kayayyaki.
Ga yawancin masu siye, manyan na'urori na ƙarshe waɗanda NVIDIA ke da keɓantacce akan su suna da sha'awa kawai. Amma sau ɗaya kawai, ƙarancin fasaha na zane-zane na "kore" ya haifar da yanayi inda AMD zai iya, a cikin tsalle ɗaya, cim ma irin waɗannan matsaloli na gine-ginen RDNA kamar ganowa da haɓakawa. Koyaya, mun kasance muna jin annabcin cewa AMD za ta dawo da daidaito tare da mai fafatawa (ko da ba tare da da'awar sashin flagship ba) na dogon lokaci, kuma duk lokacin da ba ta da ƙarfi sosai, don haka sanarwar katunan bidiyo na Radeon RX 9000 bai jawo hankali sosai ba a kan bangon aikin NVIDIA mai ƙarfi. Amma AMD ya yi aiki mai yawa akan kurakuran kuma, idan ba a kawar da su ba, to aƙalla an biya diyya ga gazawar GPUs da suka gabata.
Muna gabatar da bita na Radeon RX 9070 XT a cikin wani sabon tsarin kwatancen don 3DNews, tun da sabon samfurin "ja" da babban abokin hamayyarsa, GeForce RTX 5070 Ti, ya fada hannunmu a lokaci guda, kuma ba za a iya la'akari da na'urar ɗaya a ware daga ɗayan ba, tunda a Rasha ana siyar da su kusan kuɗi ɗaya.
 |
 |
Navi 48 GPU da RDNA 4 gine
Lalaci ne kawai ba su ce NVIDIA ta zama mai cin gashin kanta ba a cikin babban kasuwar GPU kuma tana yin hakan. Kadan sau da yawa, ana ba da hankali ga gaskiyar cewa duk kamfanonin da ke zayyana manyan na'urori suna tilasta yin amfani da sabis na ɗan kwangila ɗaya don samar da su: Taiwan TSMC. Buƙatar ƙarfin TSMC wataƙila shine ɗayan dalilan da yasa AMD ta daina haɓaka manyan GPUs masu amfani, kuma gabaɗaya, layin "ja" na'urori masu haɓakawa na gaba a halin yanzu suna da crystal guda ɗaya kawai - Navi 48.
| Manufacturer | AMD | ||
|---|---|---|---|
| Title | Navi 32 | Navi 31 | Navi 48 |
| Inda ake amfani da shi |
Radeon RX 7700 XT; Radeon RX 7800 XT |
Radeon RX 7900 GRE; Radeon RX 7900 XT; Radeon RX 7900 XTX |
Radeon RX 9070; Radeon RX 9070 XTX |
| gine | RDNA3 | RDNA3 | RDNA4 |
| Fasahar aiwatarwa, nm | TSMC N5/N6 | TSMC N5/N6 | Farashin TSMC N4P |
| Yawan transistor, biliyan | 28,1 | 57,7 | 53,9 |
| Yankin guntu, mm2 | 346 | 522 | 357 |
| Adadin CU/WGP/SE | |||
| Ƙididdigar Ƙididdigar (CU) | 60 | 96 | 64 |
| Masu Gudanar da Ƙungiyar Aiki (WGP) | 30 | 48 | 32 |
| Shader Arrays (SA) | Babu | Babu | Babu |
| Injin Shader (SE) | 4? | 6 | 4 |
| Kanfigareshan Lissafi | |||
| Vector ALUs (FP32) | 2 × 32 | 2 × 32 | 2 × 32 |
| Vector ALUs (FP32/INT32) | 2 × 32 | 2 × 32 | 2 × 32 |
| Scalar ALU | 2 | 2 | 2 |
| Sashin Manufar Musamman (SFU) ALU | 2 × 8 | 2 × 8 | 2 × 8 |
| Ray Accelerators | 1 | 1 | 1 |
| Rukunin Taswirar Rubutu (TMUs) | 4 | 4 | 4 |
| Rijistar Vector/scalar, KB | 384/10 | 384/10 | 384/16 |
| Girman cache L0, KB | 32 | 32 | 32 |
| Kanfigareshan Mai Gudanar da Ƙungiyar Aiki (WGP). | |||
| Ƙwaƙwalwar Ƙwaƙwalwar Ƙwaƙwalwa, KB | 128 | 128 | 128 |
| Girman cache umarni, KB | 32 | 32 | 32 |
| Girman cache scalar, KB | 16 | 16 | 16 |
| GPU Computing Units | |||
| Vector ALUs (FP32) | 3 (840) | 6 (144) | 4 (096) |
| Rukunin Taswirar Rubutu (TMUs) | 240 | 384 | 256 |
| Raster Operation Blocks (ROPs) | 96 | 192 | 128 |
| Ray Accelerators | 60 | 96 | 64 |
| Tsarin ƙwaƙwalwar ajiya | |||
| Girman cache L1, KB | 2 048 | 3 072 | 2 048 |
| Girman cache L2, MB | 4 | 6 | 8 |
| Infinity Cache Volume, MB | 64 | 96 | 64 |
| Faɗin bas na VRAM, rago | 256 | 384 | 256 |
| Nau'in guntu VRAM | GDDR6 | GDDR6 | GDDR6 |
| PCI Express dubawa | 4.0 x16 | 4.0 x16 | 5.0 x16 |
Daga cikin tsofaffin kwakwalwan kwamfuta, ya fi kama da samfurin na biyu, Navi 32, a cikin ainihin ƙayyadaddun bayanai, amma ya ƙunshi kusan transistor kamar Navi 31 (53,9 da biliyan 57,7, bi da bi). An ƙirƙira Navi 48 ta amfani da fasahar TSMC N4P, kuma tana cikin kullin hoto na nm guda 5 wanda aka ƙware shekaru biyu da suka gabata. Koyaya, Navi 48 shine kawai 68% girman Navi 31, godiya ga ƙawar AMD daga ƙirar chiplet, tare da guntuwar MCD waɗanda ke ɗaukar cache na L6 da masu sarrafa ƙwaƙwalwar ajiya akan tsarin NXNUMX mai ƙarfi.
Gine-ginen sabon GPU, RDNA 4, ba ya canzawa daga abubuwan da suka gabata. Navi 48 ya ƙunshi 64 Compute Units, ko 4 FP096 mai iya shader ALUs, wanda shine madaidaicin tsari na GPU na biyu - kamar yadda yake da bas ɗin ƙwaƙwalwar bidiyo 32-bit. Koyaya, AMD kuma yana kwatanta Navi 256 a cikin ƙimar aiki zuwa flagship Navi 48, wanda ke da madaidaicin madaidaicin madaidaicin 31 ALUs. Bambancin ya kamata ya zama fiye da ramawa ta hanyar haɓakawa da yawa a cikin dabaru, kuma don fahimtar su, muna ba da shawarar sabunta ƙwaƙwalwar ajiyar ku na gine-ginen RDNA 6, wanda muka tattauna dalla-dalla a cikin bita. .
Idan aka kwatanta da koren gine-gine RDNA 4 yana wakiltar haɓaka mafi girma da zurfi na abubuwan da suka gabata. Koyaya, bayanin duk sabbin abubuwa da mamaki yana ninka cikin ɗan gajeren jeri saboda gaskiyar cewa NVIDIA a al'adance tana ba da ƙarin kulawa ga hadaddun software ko mafita na software waɗanda ke buƙatar cikakkun bayanai, yayin da sabbin abubuwan AMD, a matsayin mai mulkin, farawa da ƙarewa cikin kayan masarufi.
Canje-canje na asali sun faru a cikin ƙira na Compute Unit, katangar lissafin da ba za a iya ganuwa ba wanda yayi kama da NVIDIA's SM and CPU cores. Don haka, mai tsara umarni yanzu yana goyan bayan abin da ake kira shingen rabe-raben da ke ba da izini don aiwatarwa yayin lokacin aiki. An inganta ƙaddamar da umarni daga cache, kuma scalar ALUs, waɗanda ake amfani da su don aiwatar da tsalle-tsalle na tattalin arziki da ayyukan reshe, sun gaji ikon yin aiki tare da mahimman bayanai daga gine-ginen wayar hannu na RDNA 3.5. A lokaci guda kuma, an ƙara ƙarar rajista na scalar: daga 10 zuwa 16 KB.
Ɗaya daga cikin mahimman abubuwan ƙirƙira na RDNA 4 shine ninki biyu na kisa na umarnin matrix WMMA, waɗanda ake amfani da su a cikin koyan na'ura da ayyukan sarrafa bayanai ta amfani da hanyoyin sadarwa na jijiyoyi. Bugu da ƙari, saitin koyarwa na RDNA 4 ya ƙunshi umarni don ƙididdigewa akan matrices marasa ƙarfi da bayanai masu iyo 8-bit (FP8 da BF8).
Koyaya, 1 FLOPs a kowace kayan aiki na agogo yana nufin cewa kwakwalwan kwamfuta na Navi suna isa kawai matakin ginin “kore” na baya, Ampere. Bugu da kari, yayin da NVIDIA da kwakwalwan kwamfuta na Intel ke amfani da tsararrun ALU daban-daban don ayyukan matrix, a cikin gine-ginen RNDA 024 wannan aikin har yanzu yana kan inuwar SIMDs.
Koyaya, kar mu manta cewa a cikin aikace-aikacen ainihin duniya, GPUs na tushen RDNA 3 ba su taɓa kusantar cikakken kaya ba. Da'awar teraflops a cikin ƙididdiga na FP32 gabaɗaya ba za a iya cimma su ba idan aka yi amfani da tsarin VOPD (Vector Operation Dual), wanda ke ba da izinin tattara umarnin vector masu zaman kansu guda biyu tare. Kuma yana sanya ƙuntatawa da yawa a cikin kanta, don haka rabin 32-bit ALUs a cikin Rukunin Lissafi galibi suna aiki. A cikin kayan aiki mai gauraye (kamar wasa tare da haɓakawa), ayyukan matrix, ba shakka, za su cire wasu albarkatun ƙididdiga daga shaders, amma waɗannan asarar ba za su yi girma kamar yadda ake gani a kallon farko ba.
| Rukunin Lissafi (AMD RDNA 3) | Rukunin Lissafi (AMD RDNA 4) | Multiprocessor Mai Yawo (NVIDIA Ada Lovelace) | Multiprocessor Mai Yawo (NVIDIA Blackwell) | Xe-core (Intel Xe-HPG) | Xe-core (Intel Xe2) | |
|---|---|---|---|---|---|---|
| Babban tubalan |
2 × SIMD32 (FP32/INT32); 2 × SIMD32 (FP32); 2 × SIMD2 (FP64); 2 × SIMD8 (SFU); 2 × scalar ALUs |
2 × SIMD32 (FP32/INT32); 2 × SIMD32 (FP32); 2 × SIMD2 (FP64); 2 × SIMD8 (SFU); 2 × scalar ALUs |
4 × SIMD16 (FP32/INT32); 4 × SIMD16 (FP32); 2×SISD? (FP64); 4 × SIMD4 (SFU); 4 × scalar ALUs; 4 × tensor cores |
8 × SIMD16 (FP32/INT32); 2×SISD? (FP64); 4 × SIMD4 (SFU); 4 × scalar ALUs; 4 × tensor cores |
16 × SIMD8 (FP32); 16 × SIMD8 (INT32); 16 × SISD (FP64); 16 × SIMD2 (SFU); 16×XXX |
8 × SIMD16 (FP32); 8 × SIMD16 (INT32); 8 × SIMD2 (FP64); 8 × SIMD4 (SFU); 8×XXX |
| Ayyukan Layin SIMD a kowace agogo |
128×FP32; 64×INT32; 256×FP16; 4×FP64; 16 × ayyukan trans |
128×FP32; 64×INT32; 256×FP16; 4×FP64; 16 × ayyukan trans |
128×FP32; 64×INT32; 128×FP16; 2×FP64; 16 × ayyukan trans |
128×FP32; 128×INT32; 128×FP16; 2×FP64; 16 × ayyukan trans |
128×FP32; 128×INT32; 256×FP16; 16×FP64; 32 × ayyukan trans |
128×FP32; 128×INT32; 256×FP16; 16×FP64; 32 × ayyukan trans |
| Ayyukan Matrix, FLOPs kowane agogo (FP16) | 512 | 1 024 | 2 048 | 2 048 | 2 048 | 2 048 |
Ƙimar firam tare da hanyar sadarwar jijiyoyi ba halin da ake ciki ba ne ga masu haɓaka Radeon. Sabuwar, sigar ta huɗu ta FSR a ƙarshe ta dogara da koyan na'ura, gami da ƙirar ƙira, kuma samfuran da aka yi amfani da su nau'in sigar cibiyoyin sadarwa ne da masu canji. Abin takaici, ba kamar sigar baya na upscaler ba, wanda yayi aiki akan kusan kowane kayan aiki, FSR 4 yana goyan bayan RDNA 4 GPUs na gine-gine (saboda fadada umarnin WMMA da aka saita zuwa bayanan FP8). Yanayin duniya tare da sauƙaƙan ƙirar kwamfuta, kamar a cikin XeSS na Intel, ba a bayar da shi ba.
Wani yanki na sake fasalin gine-gine na RDNA shi ne gano hasken haske. Gudun gwaje-gwajen ray na gwaji tare da akwatin BVH kuma tare da alwatika ya ninka kowane Ray Accelerator (saboda haka kowane Unit Compute): daga 4 da 1 a kowace agogo zuwa 8 da 2, bi da bi. Koyaya, toshe RT na kwakwalwan kwamfuta na Intel Battlemage yana aiki a cikin ɗan gajeren lokaci na 18 da 2, kuma a cikin Blackwell GPUs yana samun madaidaicin ray-triangle guda huɗu a kowane zagayen agogo da adadin da ba a buga ba na mahaɗar akwatin ray-BVH. Haka kuma, masu fafatawa suna yin nau'ikan gwaji guda biyu (a kan akwatin BVH da triangle) a lokaci guda, yayin da AMD har yanzu ba ta yi ba. Wasu daidaitawa yana yiwuwa, saboda kowane Ray Accelerator yanzu ya ƙunshi "masu gwadawa" biyu masu zaman kansu, amma ba cikin cikakken sauri ba.
Bugu da kari, Ray Accelerator ya sami ikon sarrafa haskoki daga bi da bi. Rays waɗanda ba sa buƙatar samun dama ga matakan jinkirin ƙwaƙwalwar ajiyar ana sarrafa su da farko, kuma ana adana sakamakon gwajin ta yadda shirin shader ya karɓi su a cikin tsari da ake tsammani.
Ma'anar binciken ray a yanzu yana amfani da tsarin BVH8 mai yawa maimakon BVH4 (takwas maimakon nodes hudu ga kowane reshe na matsayi na girma), wanda ke ƙara saurin wucewa kuma a lokaci guda yana rage matsa lamba akan ƙwaƙwalwar bidiyo. Girman BVH a cikin ƙwaƙwalwar ajiya kuma an rage godiya ga sabon matsi na farko na algorithm. A ƙarshe, AMD ya ba da shawarar ƙaddamar da daidaitawar akwatunan BVH daga cikakkiyar daidaitawar yanayin. Akwatunan jujjuya bisa ga lissafi na abubuwa suna ba da damar rage adadin matakai a cikin bincike.
Abin kunya ne cewa duk da duk canje-canjen da aka jera a sama, RDNA 4 har yanzu tana yin wucewar BVH akan shader ALU maimakon amfani da dabaru na sadaukarwa. Dangane da Imagination Technologies' rarrabuwa na kayan aikin gano radiyo na hardware, mafita na AMD sun makale a matakin 3, GPUs na Intel suna a matakin 4, yayin da kayan aikin NVIDIA tuni suna motsawa daga matakin XNUMX zuwa matakin XNUMX godiya ga fasalin SER - ƙungiyoyin koyarwa mai ƙarfi don haɓaka haɗin kai ga ƙwaƙwalwar ajiya kuma, a sakamakon haka, rage latency.
Babu wani analog na SER a cikin gine-ginen RDNA 4 ko dai, amma aikin Ƙwaƙwalwar Ƙwararren Ƙwararren Ƙwaƙwalwa ya bayyana, wanda ke taimakawa wajen magance wannan matsala ta wata hanya dabam. Siffofin RDNA na baya, suna aiki tare da ƙungiyoyin umarni da yawa (wavefronts a cikin kalmomin AMD), dawo da bayanai daga ƙwaƙwalwar ajiya daidai daidai da tsarin da aka yi buƙatun. Lokacin da bayanan da mai shader ke buƙata yana cikin ma'aji mai sauri, dawowar na iya jinkirta dawowa idan wani shader ya yi buƙatarsa a baya kuma bayanansa suna cikin jinkirin VRAM. A cikin gaskiya, wannan iyakance yana da alama gaba ɗaya mara ma'ana a cikin mahallin kwamfuta mai zare da yawa, amma mai yiwuwa an gabatar da shi don sauƙaƙe ƙungiyar GPU kuma na ɗan lokaci (karanta: a cikin wasannin da aka ɓata) ya kasance daidaitawa mai ma'ana. RDNA 4 yana ba ƙungiyoyin umarni damar dawo da bayanai ba tare da la'akari da tsarin buƙatun ba.
Ƙwaƙwalwar ajiya na oda ɗaya ne kawai daga cikin canje-canje ga tarin ƙwaƙwalwar ajiya a cikin gine-ginen RDNA 4. Rarraba Rukunin Lissafin Lissafi zuwa shirye-shiryen shader da aka yi amfani da su sun kasance a tsaye, dangane da matsakaicin yawan amfani yayin rayuwar shader, wanda ke nufin cewa wani yanki na fayil ɗin rajista ya zama fanko sosai. RDNA 4 yana ba da damar shaders don nema da saki rajistar a hankali. Ta wannan hanyar ana amfani da sararin fayil ɗin rajista fiye da tattalin arziki.
Abin sha'awa shine, Shader Engine, babban tsarin GPU wanda ya haɗu da CU array, rasterizer, primitive block, da rasterization aiki tubalan, ya rasa nasa cache matakin farko a cikin RDNA 4. Amma kowane ɓangaren cache na L2 ya ƙunshi ninki biyu na bankuna. Wannan ya ƙãra ba kawai ƙarar sa ba, har ma da abubuwan da ake amfani da shi, a cikin sassan biyu - zuwa Injin Shader da kuma Infinity Cache (wato, L3). Girman Infinity Cache kanta ya kasance daidai da na Navi 32 - 64 MB.
Navi 48 yana da layin bas na ƙarni na biyar na PCI Express 16, amma a cikin ayyukan caca, ba shakka, ba zai iya mamaye irin wannan saurin dubawa ba.
A ƙarshe, AMD ta ce codec ɗin bidiyo na hardware a cikin GPU na gaba na gaba yana ba da ingantattun ingantattun ƙididdiga a ƙananan bitrates sananne don yawo, amma ba komai ba. Wannan yana nufin cewa goyon bayan YUV 4: 2: 2 samfurin launi na launi, wanda ke samuwa akan na'urori masu sarrafa hoto na NVIDIA da Intel, bai bayyana ba.
Dangane da fitowar hoto, mai kula da nuni na Navi 48 ya dace da sabbin hanyoyin sadarwa na DisplayPort da HDMI: 2.1a da 2.1b, bi da bi.
Halayen fasaha, farashin
Radeon RX 9070 XT ya zo tare da cikakken aiki Navi 48 mutu tare da 64 Compute Units, wanda ke fassara zuwa 4 daidaitattun daidaitattun ALUs. Ba abu mai sauƙi ba ne a sami magabata na sabon samfurin a tsakanin masu haɓaka "ja" na ƙarni na baya, saboda duka grid ɗin ƙirar da ka'idar lambar GPU sun canza (Navi kwakwalwan kwamfuta ba su taɓa kai takwas ba). Radeon RX 096 XT yayi kama da Radeon RX 9070 XT dangane da tsarin ƙididdiga naúrar da halayen tarin ƙwaƙwalwar ajiya, amma AMD tana sanya shi azaman maye gurbin Radeon RX 7800 XTX.
Agogon Wasan da Ƙarfafa Agogo (gudun agogon manufa a cikin wasanni da matsakaicin saurin agogo) na GPU sun ƙaru sosai idan aka kwatanta da halayen ƙirar ƙirar ƙarni na baya kuma suna 2,4 da 2,97 GHz, bi da bi. Radeon RX 7900 XTX a ka'idar shine 26% sauri dangane da matsakaicin saurin agogonsa, amma AMD yana neman daidaita wannan bambanci tare da haɓakawa ga gine-ginen RDNA 4.
Radeon RX 9070 XT ya karɓi 16 GB na ƙwaƙwalwar GDDR6 tare da bandwidth na 20 Gbps kowane fil na bas 256-bit. An yayatawa AMD cewa yayi la'akari da haɓaka VRAM zuwa 32GB, amma a ƙarshe an ɗauka cewa daidaitawar ta wuce kima ga hanyoyin masu amfani. Abin kunya ne, saboda sha'awar wasanni suna ci gaba da girma, kuma adadin ƙwaƙwalwar bidiyo ya kasance muhimmiyar fa'ida ga na'urorin AMD da yawa da suka gabata.
Radeon RX 9070 XT yana da ikon tunani na 304W da farashin dillalan da aka ba da shawarar na $599 ($ 100 fiye da Radeon RX 7800 XT).
| Manufacturer | AMD | NVDIA | ||||||
|---|---|---|---|---|---|---|---|---|
| Samfurin | Radeon RX 7900 XT | Radeon RX 7900 XTX | Radeon rx 9070 | Radeon RX 9070 XT | GeForce RTX 4070 Ti SUPER | GeForce RTX 4080 SUPER | GeForce RTX 5070 Ti | GeForce RTX 5080 |
| GPU | ||||||||
| Title | Farashin 31XT | Navi 31 XTX | Farashin 48XT | Navi 48 XTX | AD103 | AD103 | GB203 | GB203 |
| gine | RDNA3 | RDNA3 | RDNA4 | RDNA4 | Ada Soyayya | Ada Soyayya | Blackwell | Blackwell |
| Tsarin fasaha | TSMC N5/N6 | TSMC N5/N6 | Farashin TSMC N4C | Farashin TSMC N4C | Saukewa: TSMC 4N | Saukewa: TSMC 4N | Farashin TSMC4NP | Farashin TSMC4NP |
| Yawan transistor, biliyan | 57,7 | 57,7 | 53,9 | 53,9 | 45,9 | 45,9 | 45,6 | 45,6 |
| Mitar agogo (Agogon Tushe/Agogon Wasa/Agogon Ƙarfafa), MHz | 1 500 / 2 025/2 394 | 1 855 / 2 269/2 499 | 1 330 / 2 070/2 520 | 1 660 / 2 400/2 970 | 2 310 / 2 610 | 2 205 / 2 550 | 2 295 / 2 452 | 2 295 / 2 617 |
| Shader ALUs (FP32) | 5 (376) | 6 (144) | 3 (584) | 4 (096) | 8 448 | 10 240 | 8 960 | 10 752 |
| Rukunin Taswirar Rubutu (TMUs) | 336 | 384 | 224 | 256 | 264 | 320 | 280 | 336 |
| Raster Operation Blocks (ROPs) | 192 | 192 | 128 | 128 | 112 | 112 | 96 | 112 |
| Tensor kernels | Babu | Babu | Babu | Babu | 264 | 320 | 280 | 336 |
| RT kwatance | 84 | 96 | 112 | 128 | 66 | 80 | 70 | 84 |
| Girman cache matakin ƙarshe, MB | 80 | 96 | 64 | 64 | 48 | 64 | 64 | 64 |
| RAM | ||||||||
| Faɗin bas, rago | 320 | 384 | 256 | 256 | 256 | 256 | 256 | 256 |
| Nau'in guntu | Saukewa: GDDR6 | Saukewa: GDDR6 | Saukewa: GDDR6 | Saukewa: GDDR6 | GDDR6X SRAM | GDDR6X SRAM | Saukewa: GDDR7 | Saukewa: GDDR7 |
| Abubuwan da aka samar a kowane lamba, Gbps | 20 | 20 | 20 | 20 | 21 | 23 | 28 | 30 |
| Jimlar abubuwan da aka fitar, Gbps | 800 | 960 | 640 | 640 | 672 | 736 | 896 | 960 |
| girma, GB | 20 | 24 | 16 | 16 | 16 | 16 | 16 | 16 |
| Yawan aiki | ||||||||
| Mafi girman aikin FP32, TFLOPS | 25,7 (51,5) | 30,7 (61,1) | 18,1 (36,1) | 24,3 (48,7) | 44,1 | 52,2 | 43,9 | 56,3 |
| Ayyukan FP64/FP32 | 1/32 (1/64) | 1/32 (1/64) | 1/32 (1/64) | 1/32 (1/64) | 1/64 | 1/64 | 1/64 | 1/64 |
| Ayyukan FP16/FP32 | 2/1 | 2/1 | 2/1 | 2/1 | 1/1 | 1/1 | 1/1 | 1/1 |
| Wasu | ||||||||
| PCI Express bas | PCI Express 4.0 x16 | PCI Express 4.0 x16 | PCI Express 5.0 x16 | PCI Express 5.0 x16 | PCI Express 4.0 x16 | PCI Express 4.0 x16 | 5.0 x16 | 5.0 x16 |
| Hanyoyin fitarwa na hoto | DisplayPort 2.1, HDMI 2.1a | DisplayPort 2.1, HDMI 2.1a | DisplayPort 2.1a, HDMI 2.1b | DisplayPort 2.1a, HDMI 2.1b | DisplayPort 1.4a, HDMI 2.1 | DisplayPort 1.4a, HDMI 2.1 | DisplayPort 2.1b, HDMI 2.1b | DisplayPort 2.1b, HDMI 2.1b |
| TDP/TBP, W | 315 | 335 | 220 | 304 | 285 | 320 | 300 | 360 |
| Farashin sayarwa (Amurka), $ | 899 (an bada shawarar a ranar fitarwa) | 999 (an bada shawarar a ranar fitarwa) | 549 (an bada shawarar a ranar fitarwa) | 599 (an bada shawarar a ranar fitarwa) | 799 (an bada shawarar a ranar fitarwa) | 999 (an bada shawarar a ranar fitarwa) | 749 (an bada shawarar a ranar fitarwa) | 999 (an bada shawarar a ranar fitarwa) |
Hakanan, GeForce RTX 5070 Ti yana dogara ne akan GB203 GPU iri ɗaya da aka yi amfani da shi a cikin GeForce RTX 5080, amma ƙaramin ƙirar ya rasa 14 masu gudana multiprocessors, wato, 1 FP792 shader ALUs, kuma an rage saurin agogon wasan. A sakamakon haka, GeForce RTX 32 ya rabu da RTX 5080 Ti da 5070% a cikin aikin ka'idar, kuma GeForce RTX 27 Ti kanta ba ta da bambanci a cikin wannan siga daga RTX 5070 Ti SUPER. Kamar yadda yake tare da Radeon RX 4070 XT, rashin ƙarancin ikon sarrafa kwamfuta ana nufin a biya shi ta sabon gine-ginen GPU, kuma da farko ta aikin samar da firam ɗin da yawa ta amfani da DLSS (har zuwa firam ɗin da aka ƙirƙira guda uku a cikin tazara tsakanin “na gaske” biyu).
An ƙididdige GeForce RTX 5070 Ti don 300W na amfani da wutar lantarki kuma, tare da MSRP na $749, yana da $50 mai rahusa fiye da GeForce RTX 4070 Ti SUPER.
Bambanci na $ 150 a cikin ƙimar kuɗin da aka ba da shawarar tsakanin Radeon RX 9070 XT da GeForce RTX 5070 Ti ba a sani ba a farashin Rasha na yanzu: mun sami samfuran abokin tarayya mafi araha a cikin DNS don 99 da 999 rubles. bi da bi. Waɗannan su ne waɗanda za mu gwada a cikin ma'auni. Radeon RX 104 XT yana wakilta ta katin zane na GIGABYTE GAMING OC tare da ma'aikata GPU overclock zuwa 999/9070 GHz, kuma GeForce RTX 2,52 Ti mai haɓaka ce daga Palit GamingPro tare da ƙayyadaddun bayanai.
GIGABYTE Radeon RX 9070 XT GAMING OC: ƙira
Radeon RX 9070 XT daga GIGABYTE yana da madaidaicin ma'auni don katin bidiyo tare da ikon amfani da ƙasa da (kuma, kamar yadda gwaje-gwaje za su nuna, fiye da) 300 W: 288 mm tsayi da ramuka uku a faɗi. Iyakar abin da ke haskaka hasken baya na LED shine ɗan gajeren tsiri a gefe. Farantin tare da tambarin masana'anta abu ne mai motsi: idan kun matsar da shi a kan tsiri na LED, tambarin ya fara haskakawa.
Ana amfani da tsarin sanyaya ta magoya baya uku tare da diamita na 87 mm. Wani ƙaramin sashe na radiator yana ba da damar iska ta wuce ta godiya ga yankewa a cikin farantin baya.
Na'ura mai sarrafa hoto da kwakwalwan ƙwaƙwalwar bidiyo suna ba da zafi zuwa ɗakin ƙaura, amma na farko kai tsaye, da na ƙarshe ta hanyar sararin ƙarfe mai siffar U. Haɗe da fins ɗin radiyo akwai faranti guda ɗaya waɗanda ke aiki azaman magudanar zafi don abubuwan VRM. Bututun zafi guda bakwai ne kawai, diamitansu shine mm 6. Lura cewa GIGABYTE ya zaɓi pads thermal pads akan na yau da kullun. Wannan na iya zama dalilin da ya sa kwakwalwan kwamfuta VRAM ke yin zafi a ƙarƙashin kaya (duba sashin juzu'i na bita).
Plate ɗin baya yana shiga cikin ɓarkewar zafi daga allon da'irar da aka buga ta amfani da pad ɗin thermal da yawa (a nan suna na yau da kullun, ba ruwa ba).
GIGABYTE Radeon RX 9070 XT GAMING OC: PCB
Wasu bambance-bambancen Radeon RX 9070 XT suna zana ƙarfin su daga mai haɗin 12V-2 × 6, amma GIGABYTE ya zaɓi masu haɗin fil takwas na duniya guda uku. Dangane da babban ƙarfin amfani da katin bidiyo da gunaguni da ke gudana game da sabon mai haɗawa, wannan shine mafi aminci, kodayake ba kamar yadda ya dace da ingantaccen bayani ba. Daga cikin matakai 17 na tsarin samar da wutar lantarki, 14, wanda Monolithic Power Systems MP2868A PWM mai kula da shi ke sarrafawa, ana kebe su zuwa na'urar sarrafa hoto, kuma an ware wasu uku zuwa ƙwaƙwalwar bidiyo. Dukansu VRMs an sanye su da matakan wutar lantarki na MP87993 tare da kimanta ƙimar 90A na yanzu.
VRAM ya ƙunshi kwakwalwan kwamfuta na SK hynix, alamar su H56G42AS8DX-014 tana ɓoye bandwidth na 20 Gbps.
Baya ga kwafin BIOS wanda ke aiki ta tsohuwa, katin bidiyo yana da firmware na “silent” madadin.
Palit GeForce RTX 5070 Ti GamingPro: Zane
Palit GeForce RTX 5070 Ti GamingPro yana auna tsayin 331,9mm amma yana da ramukan fadada uku kawai. Ƙarƙashin tambarin GamingPro da ɓangaren gaba na gaban panel sune RGB LEDs. Ana iya daidaita aikin LED tare da motherboard ta hanyar daidaitaccen haɗin ARGB, wanda ke kusa da shigarwar wutar lantarki 12V-2 × 6.
Fuskokin gefen shari'ar an kafa su ta hanyar simintin alluminium da simintin gyare-gyare tare da ramukan samun iska a kan dogayen ɓangarorin. Farantin kariyar da ke bayan PCB yana da grille da aka saba don hura radiyo. Matsakaicin fan suna da diamita na mm 94.
A gindin mai sanyaya akwai ɗakin tururi mai girma wanda zai iya rufe mutuwar GPU da guntuwar ƙwaƙwalwar bidiyo da ke kewaye, yayin da aka samar da abubuwan VRM tare da raƙuman zafi daban-daban. An yi amfani da lamelas na radiator akan bututun zafi guda takwas tare da diamita na 6 mm.
Plate ɗin bayan karfe yana aiki ne kawai na kariya da kayan ado, tunda babu kushin zafi guda ɗaya tsakaninsa da PCB.
Katin bidiyo ya zo tare da adaftar daga masu haɗin wutar lantarki guda takwas zuwa 12V-2 × 6 da ƙaramin kushin linzamin kwamfuta.
Palit GeForce RTX 5070 Ti GamingPro PCB
Sabanin katin bidiyo , wanda muka duba a cikin sake dubawa na GeForce RTX 5080 na baya-bayan nan, samfurin na gaba mafi tsufa a cikin jerin an haɗa shi akan PCB mai fa'ida, amma wutar lantarki ta fi sauƙi. GPU yana da VRM mai kashi 15 mai sarrafawa ta Monolithic Power Systems MP29816 PWM mai sarrafa. Samar da wutar lantarki don kwakwalwan ƙwaƙwalwar ajiya na GDDR7 mataki ne guda uku, dangane da mai sarrafa MP2988. Dukan hukumomin biyu suna sanye take da MP87993 transistor majalisai tare da ƙididdige ƙimar yanzu na 90 A.
Alamar guntuwar ƙwaƙwalwar bidiyo ta Samsung K4VAF325ZC-SC28 tana nuna kayan aiki na 28 Gbps.
Kamar Radeon RX 9070 XT daga GIGABYTE, Palit accelerator yana da zaɓuɓɓukan BIOS guda biyu - "aiki" da "shuru".
Matsayin gwaji, hanyoyin gwaji
| Gwajin tsayawa | |
|---|---|
| CPU | AMD Ryzen 9 7950X3D (PBO +150 MHz, CU -20) |
| Bangon uwa | ASUS ROG Crosshair X670E Jarumi |
| RAM | G.Skill Trident Z5 Neo RGB (F5-6000J3040G32GX2-TZ5NR), 2 x 32 GB (6200 MT/s, CL30) |
| ROM | Soldigm P44 Pro, 2 TB |
| Wurin lantarki | Corsair AX1600i, 1600 W |
| CPU sanyaya tsarin | Tsarin Sanyaya Ruwa na Al'ada (Gurin EK-Quantum² DDC 4.2 PWM D-RGB + EK-Quantum Surface X280M) |
| Gidaje | Buɗe tsayawa |
| tsarin aiki | Windows 11 Pro |
| Software don AMD GPUs | |
| Duk katunan bidiyo | AMD Software Adrenalin Edition 25.2.1/25.3.1 |
| NVIDIA GPU software | |
| Duk katunan bidiyo | NVIDIA GeForce Game Shirye Direba 532.60/532.70 |
| Wasanni ba tare da gano hasken haske ba | |||
|---|---|---|---|
| Game | API | Hanyar gwaji | Saitunan zane |
| Alan wake 2 | DirectX 12 | OCAT, wurin Bright Falls | Max. ingancin graphics |
| Labarin Baƙar fata: Wukong | DirectX 12 | Alamar da aka gina a ciki | Max. ingancin graphics |
| Cyberpunk 2077 | DirectX 12 | Alamar da aka gina a ciki | Max. ingancin graphics |
| F1 24 | DirectX 12 | Gina-ginin ma'auni, Waƙar Monaco (ruwan sama) | Max. ingancin graphics |
| Hogwarts asalin | DirectX 12 | OCAT, Trolley Ride a Hanyar zuwa Hogwarts | Max. ingancin graphics |
| Horizon Zero Dawn An Sake Matsala | DirectX 12 | Alamar da aka gina a ciki | Max. ingancin graphics |
| Metro Fitowa | DirectX 12 | Alamar da aka gina a ciki | Max. ingancin graphics; Yawan Shading: 100% |
| Red Matattu Kubuta 2 | aman wuta | Alamar da aka gina a ciki | Max. ingancin graphics |
| Sake Takawa | DirectX 12 | Alamar da aka gina a ciki | Max. ingancin graphics |
| Jimlar Yaƙi: WARHAMMER III | DirectX 11 | Alamar da aka gina a ciki (Mirrors of Madness Benchmark) | Max. ingancin graphics |
| Ray Tracing Games | ||||||
|---|---|---|---|---|---|---|
| Game | API | Hanyar gwaji | Saitunan zane | Ƙimar ƙira | ||
| AMD | Intel | NVDIA | ||||
| Alan wake 2 | DirectX 12 | OCAT, wurin Bright Falls | Max. ingancin graphics, high quality ray binciken | Ma'aunin FSR | Ma'aunin FSR | Ma'auni na DLSS + Sake Gina Ray (+ Tsarin Tsara) |
| Labarin Baƙar fata: Wukong | Alamar da aka gina a ciki | Max. Ingantattun zane-zane da gano hasken haske (Binciken Hanya) | Ma'auni na FSR (+ Tsarin Tsara) | Ma'auni na XeSS/FSR Ma'auni + Tsarin Tsarin | Ma'auni na DLSS (+ Tsarin Tsara) | |
| Cyberpunk 2077 | Alamar da aka gina a ciki (OCAT don ƙirar ƙira) | Max. Ingantattun zane-zane da gano hasken haske (Binciken Hanya) | Ma'auni na FSR (+ Tsarin Tsara) | Ma'auni na XeSS/FSR Ma'auni + Tsarin Tsarin | Ma'auni na DLSS (Transformer Model) + Sake Gina Ray (+ Firam Generation) | |
| F1 24 | Gina-ginin ma'auni, Waƙar Monaco (ruwan sama) | Max. ingancin graphics da kuma gano ray | Ma'auni na FSR (+ Tsarin Tsara) | Ma'auni na XeSS/FSR Ma'auni + Tsarin Tsarin | Ma'auni na DLSS (+ Tsarin Tsara) | |
| Hogwarts asalin | OCAT, Trolley Ride a Hanyar zuwa Hogwarts | Max. ingancin graphics da kuma gano ray | Ma'aunin FSR | Ma'auni na XeSS | Ma'auni na DLSS + Sake Gina Ray (+ Tsarin Tsara) | |
| Indiana Jones da kuma Great Circle | OCAT, wurin Sukhothai | Max. Ingantattun zane-zane, babban ingancin hasashe (Binciken Tafarki); Ray Traced Shadows: Hasken Rana kawai | Ma'auni na FSR (+ Tsarin Tsara) | Ma'auni na XeSS/FSR Ma'auni + Tsarin Tsarin | Ma'auni na DLSS (Transformer Model) + Sake Gina Ray (+ Firam Generation) | |
| Buga na Enched na Metro Fitowa | Alamar da aka gina a ciki | Max. ingancin graphics da kuma gano ray | N / A | N / A | Ma'auni na DLSS | |
| Sake Takawa | Alamar da aka gina a ciki (OCAT don ƙirar ƙira) | Max. ingancin graphics da kuma gano ray | Ma'auni na FSR (+ Tsarin Tsara) | Ma'auni na XeSS/FSR Ma'auni + Tsarin Tsarin | Ma'auni na DLSS (+ Tsarin Tsara) | |
A yawancin wasanni, matsakaita da mafi ƙaranci (muna nuna kashi 1st na rarrabuwa) ƙimar firam ɗin ana samun su ne daga tsararrun lokutan yin firam guda ɗaya ko ƙimar firam ɗin nan take da aka samu ta amfani da ginanniyar ma'auni. Waɗanda ke keɓance wasannin ne waɗanda ba su da ginanniyar ma'auni da gwaje-gwajen da ke amfani da ƙirar ƙira: a waɗannan lokuta, muna amfani da shirin OCAT don ɗaukar tazarar tsaka-tsakin.
| Aikace-aikace na aiki | ||
|---|---|---|
| Aikace-aikacen | Alamar alama | Saituna |
| Adobe Premiere Pro 25.x | Daidaitaccen (4K) | |
| Blender 4.x | Wakilin 327 Barbershop demo daga gidan yanar gizon Blender | Mai yin cycles |
| Blackmagic Design DaVinci Resolve Studio 19.x | Standard (4K); H.264/HEVC Yanayin rufewa: atomatik | |
| CAD aikace-aikace | SPECviewperf 2020 v3.1 | ƙudurin allo: 3840 × 2160 |
| Gyaran Bidiyo (ffmpeg 5.x) | |||
|---|---|---|---|
| Tsarin | yarda | Sigar ƙididdiga | API |
| H.264 (YUV 4:2:0, 8 ragowa/tashar) | 1920 × 1080 | Babban Bayani, L4.1 | Saukewa: D3D11VA |
| 3840 × 2160 | Babban Bayani, L5.1 | ||
| HEVC (YUV 4: 2: 0, 8 bits/tashar) | 1920 × 1080 | Babban Bayanan martaba, L4.0 | |
| 3840 × 2160 | Babban Bayanan martaba, L5.0 | ||
| 7680 × 4320 | Babban Bayanan martaba, L6.0 | ||
| VP9 (YUV 4: 2: 0, 8 ragowa/tashar) | 1920 × 1080 | N / A | |
| 3840 × 2160 | |||
| 7680 × 4320 | |||
| AV1 (YUV 4: 2: 0, 8 bits/tashar) | 1920 × 1080 | Babban Bayanan martaba, L4.0 | |
| 3840 × 2160 | Babban Bayanan martaba, L5.0 | ||
| 7680 × 4320 | Babban Bayanan martaba, L6.0 | ||
| Rufin bidiyo (ffmpeg 5.x) | |||||||
|---|---|---|---|---|---|---|---|
| Tsarin | yarda | Sigar ƙididdiga | API | ||||
| AMD | Intel | NVDIA | AMD | Intel | NVDIA | ||
| H.264 (YUV 4:2:0, 8 ragowa/tashar) | 1920 × 1080 | -c: v h264_amf - ingancin gudun - coder cabac -refs 1 -b: v 3M | -c:v h264_qsv -saitaccen tsari da sauri -profile:v babban matakin 4.1 -b:v 3M | -c: v h264_nvenc -saitaccen sauri -coder cabac -refs 1 -b:v 3M | AMF | wani VPL | NVENC |
| 3840 × 2160 | -c: v h264_amf - ingancin gudun - coder cabac -refs 1 -b: v 7.5M | -c:v h264_qsv -saitaccen tsari da sauri -profile:v babban matakin 5.1 -b:v 7.5M | -c: v h264_nvenc -saitaccen sauri -coder cabac -refs 1 -b:v 7.5M | ||||
| HEVC (YUV 4: 2: 0, 8 bits/tashar) | 1920 × 1080 | -c:v hevc_amf -sauri mai inganci -b:v 3M | -c:v hevc_qsv -saitaccen tsari mai sauri -tier main -b:v 3M | -c:v hevc_nvenc -saitaccen tsari da sauri -b:v 3M | |||
| 3840 × 2160 | -c:v hevc_amf -sauri mai inganci -b:v 7.5M | -c:v hevc_qsv -saitaccen tsari mai sauri -tier main -b:v 7.5M | -c:v hevc_nvenc -saitaccen tsari da sauri -b:v 7.5M | ||||
| 7680 × 4320 | -c:v hevc_amf -sauri mai inganci -b:v 20M | -c:v hevc_qsv -saitaccen tsari mai sauri -tier main -b:v 20M | -c:v hevc_nvenc -saitaccen tsari da sauri -b:v 20M | ||||
| AV1 (YUV 4: 2: 0, 8 bits/tashar) | 1920 × 1080 | -c:v hevc_amf -sauri mai inganci -b:v 3M | -c:v av1_qsv -saitaccen tsari da sauri -profile main -b:v 3M | -c:v hevc_nvenc -saitaccen tsari da sauri -b:v 3M | |||
| 3840 × 2160 | -c:v hevc_amf -sauri mai inganci -b:v 7.5M | -c:v av1_qsv -saitaccen tsari da sauri -profile main -b:v 7.5M | -c:v hevc_nvenc -saitaccen tsari da sauri -b:v 7.5M | ||||
| 7680 × 4320 | -c:v hevc_amf -sauri mai inganci -b:v 20M | -c:v av1_qsv -saitaccen tsari da sauri -profile main -b:v 20M | -c:v hevc_nvenc -saitaccen tsari da sauri -b:v 20M | ||||
Ana yin rikodin ƙarfin katunan bidiyo daban daga CPU da sauran abubuwan PC ta amfani da na'urar NVIDIA PCAT. Nauyin don gwajin ƙarfi da amo shine wasan Cyberpunk 2077 a ƙuduri na 3840 × 2160 da matsakaicin saitunan ingancin hoto (ba tare da binciken ray ba), kazalika da gwajin damuwa na FurMark tare da mafi girman saiti (shafi 3840 × 2160, MSAA 8x). Ana yin ma'auni na duk sigogi bayan katin bidiyo ya dumama, lokacin da zafin GPU da mitocin agogo suka daidaita.
Gwada mahalarta
Katunan bidiyo masu zuwa sun shiga cikin gwajin aiki:
- AMD Radeon RX 9070 XT (1820/3060 MHz, 20 Gbps, 16 GB);
- NVIDIA GeForce RTX 5070 Ti (2295/2452 MHz, 28 Gbps, 16 GB);
- ;
- ;
- ;
- ;
- ;
- .
Kusan Ana nuna tushe da mitoci masu haɓakawa na GPU a maɓalli.
Gudun agogo, amfani da wutar lantarki, zazzabi, matakin ƙara da overclocking
Navi 48 GPU ya rasa rabe-raben yanki na agogo a cikin gaba da tsararrun CU wanda AMD ya gabatar a cikin ƙarni na baya na kwakwalwan Navi. Koyaya, komai daga cikin mitoci biyu na flagship GPU na layin mai fita (Navi 31) kuka ɗauka azaman farawa, Navi 48 ya sami babban ci gaba - har zuwa 2,9 GHz. GeForce RTX 5070 Ti, a gefe guda, yana da abun ciki tare da ƙimar agogon 2,8 GHz ƙarƙashin nauyin wasa.
| Ayyuka a ƙarƙashin kaya (Cyberpunk 2077) | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| Katin bidiyo | Saituna | Gudun agogon GPU, MHz (yankin shader) | Gudun agogon GPU, MHz (ƙarshen gaba) | Wutar lantarki ta GPU, V | Gudun fan, rpm (% na max.) | Gudun fan 2, rpm (% na max.) | |||
| Matsakaici | Max. | Matsakaici | Max. | Matsakaici | Max. | Matsakaici | Matsakaici | ||
| GIGABYTE Radeon RX 9070 XT GAMING OC (1820/3060 MHz, 20 Gbps, 16 GB) | Silent BIOS | 2900 | 2956 | N / A | N / A | 1,01 | 1,01 | 1479 (30%) | N / A |
| GIGABYTE Radeon RX 9070 XT GAMING OC (1820/3060 MHz, 20 Gbps, 16 GB) | Ayyukan BIOS | 2915 | 2941 | N / A | N / A | 1,01 | 1,01 | 1696 (34%) | N / A |
| GIGABYTE Radeon RX 9070 XT GAMING OC (+300 MHz, 22 Gbps, 16 GB) | Ayyukan BIOS, + 10% TBP | 2980 | 3001 | N / A | N / A | 1,03 | 1,04 | 1880 (38%) | N / A |
| SAPPHIRE NITRO+ Radeon RX 7900 XTX (1720/2499 MHz, 20 Gbps, 24 GB) | BIOS na biyu | 2545 | 2585 | 2753 | 2785 | 0,91 | 0,93 | 1412 (34%) | N / A |
| Palit GeForce RTX 5070 Ti GamingPro (2295/2452 MHz, 28 Gbps, 16 GB) | Silent BIOS | 2816 | 2820 | N / A | N / A | 1,06 | 1,06 | 1357 (36%) | 1357 (36%) |
| Palit GeForce RTX 5070 Ti GamingPro (2295/2452 MHz, 28 Gbps, 16 GB) | Ayyukan BIOS | 2827 | 2835 | N / A | N / A | 1,06 | 1,06 | 1859 (50%) | 1859 (50%) |
| Palit GeForce RTX 5070 Ti GamingPro (+400 MHz, 32 Gbps, 16 GB) | Ayyukan BIOS | 3202 | 3210 | N / A | N / A | 1,05 | 1,05 | 2010 (55%) | 2010 (55%) |
| Palit GeForce RTX 4070 Ti GameRock OC Classic (2310/2610 MHz, 21 Gbps, 12 GB) | Silent BIOS | 2805 | 2805 | N / A | N / A | 1,10 | 1,10 | 1516 (39%) | 1516 (39%) |
| Palit GeForce RTX 4070 Ti SUPER JetStream OC (2340/2640 MHz, 21 Gbps, 16 GB) | 2673 | 2700 | N / A | N / A | 1,04 | 1,05 | 1417 (38%) | 1417 (38%) | |
| NVIDIA GeForce RTX 4080 FE (2205/2505 MHz, 22,4 Gbps, 16 GB) | 2775 | 2775 | N / A | N / A | 1,08 | 1,08 | 1383 (43%) | 1299 (39%) | |
| Palit GeForce RTX 4080 SUPER JetStream OC (2295/2580 MHz, 23 Gbps, 16 GB) | 2722 | 2745 | N / A | N / A | 1,04 | 1,07 | 1473 (39%) | 1473 (39%) | |
| Palit GeForce RTX 5080 GameRock (2295/2617 MHz, 30 Gbps, 16 GB) | Silent BIOS | 2790 | 2790 | N / A | N / A | 1,04 | 1,04 | 1490 (40%) | 1490 (40%) |
Kodayake Radeon RX 9070 XT ba shine samfurin flagship na tsararrakin sa ba (wanda ba zai zama wannan lokacin ba), ikon amfani da sabon samfurin a cikin sigar GIGABYTE ya kai 374 W - kusan iri ɗaya da Radeon RX 7900 XTX. Wannan ya bambanta da ƙarancin ƙarfin abin mamaki na katin bidiyo na "ja" a cikin yanayin aiki - kawai 7 W. GeForce RTX 5070 Ti daga Palit yana cinyewa fiye da 318 W - a matakin 80 model na 40 jerin.
Zaɓin firmware - "shuru" ko "aiki" - yana da tasiri mai tasiri akan aikin GIGABYTE Radeon RX 9070 XT GAMING OC sanyaya tsarin. A cikin yanayin "mai samarwa", zafin GPU baya wuce 62 ° C (da 84 a wuri mai zafi), amma kwakwalwan ƙwaƙwalwar ajiya suna zafi har zuwa 94 ° C. A cikin yanayin “shuru”, karatun zafin jiki yana canzawa zuwa 67, 90 da 98 ° C, bi da bi.
Bi da bi, tsarin sanyaya na Palit GeForce RTX 5070 Ti GamingPro yana kiyaye zafin jiki na kayan aikin hoto a cikin 68 ko 76 ° C (sake, ya danganta da sigar BIOS mai aiki), kuma zazzabi na kwakwalwan kwamfuta na VRAM ya tsaya a matakin yarda da yawa na 70 ko 78 ° C. Direban katin bidiyo na GeForce 50 baya bayar da rahoton zafin zafi na GPU zuwa abubuwan amfani.
Matsayin amo na tsarin sanyaya sabbin samfuran duka yana canzawa sosai, suna yin biyayya ga sauya BIOS. A cikin yanayin "shuru", Palit GeForce RTX 5070 Ti GamingPro ya zama mafi shuru na na'urorin da aka kwatanta, amma GIGABYTE Radeon RX 9070 XT GAMING kuma ya bambanta kanta tare da matsakaita sautin sauti, wanda aka daidaita don babban amfani da wutar lantarki. Akasin haka, a ƙarƙashin ikon BIOS "mai samarwa", katin bidiyo na GIGABYTE yana kiyaye matakin amo mai karɓuwa, yayin da Palit GamingPro ya kai 45 dBA - kuma wannan ba tare da overclocking ba.
Sabbin GPUs na AMD an rufe su daidai da kwakwalwan NVIDIA: yanzu, maimakon babban iyaka akan saurin agogo, mai amfani yana ƙayyade haɓakar da ake so. Ana sarrafa ƙarfin wutar lantarki na GPU zuwa ƙasa kawai. Katin samfurin mu ya tsaya tsayin daka tare da karuwar agogon manufa na 300 MHz, amma tunda ikon ajiyar wutar lantarki kawai zai iya ƙaruwa da kashi 10%, ainihin agogon ainihin ya ƙaru da alamar 65 MHz. Bi da bi, ƙwaƙwalwar bidiyo ta kai bandwidth na 22 maimakon 20 Gbps na asali. Amfanin wutar lantarki na GIGABYTE Radeon RX 9070 XT GAMING lokacin da aka rufe ya wuce alamar 400W, wanda ya haifar da zazzabi na abubuwan da ke haɓaka da digiri biyu, kuma matakin ƙara ya kai 41 dBA a nesa na 30 cm daga magoya baya.
Akasin haka, Palit GeForce RTX 5070 Ti, kamar GeForce RTX 5080, ya mamaye sosai. Duk da yake nau'in naúrar ba ta OC ba ta ba da izinin wuce gona da iri fiye da TBP na masana'anta kwata-kwata, mun sami damar ƙara 375 MHz zuwa ainihin agogon GPU a cikin gwajin wasan kuma ƙara bandwidth ƙwaƙwalwar GDDR7 daga 28 zuwa 32 Gbps. Overclocking bai yi wani babban tasiri akan yawan wutar lantarki da yanayin yanayin yanayin ba, kuma matakin amo ya kasance iri ɗaya ne kawai saboda yana da girma sosai tare da firmware "mai samarwa".
Gwajin wasa (1920×1080)
A 1080p, duka GeForce RTX 5070 Ti da Radeon RX 9070 XT suna ba da ƙimar firam fiye da 60 FPS a yawancin taken gwajin mu, kuma aƙalla XNUMX FPS a cikin Black Myth: Wukong.
| 1920 × 1080 | ||||||||
|---|---|---|---|---|---|---|---|---|
| AMD Radeon RX 9070 XT | AMD Radeon RX 7900 XTX | NVIDIA GeForce RTX 4070 Ti | NVIDIA GeForce RTX 4070 Ti SUPER | NVIDIA GeForce RTX 4080 | NVIDIA GeForce RTX 4080 SUPER | NVIDIA GeForce RTX 5070 Ti | NVIDIA GeForce RTX 5080 | |
| Alan wake 2 | 135 / 142 | 143 / 148 | 101 / 106 | 106 / 112 | 122 / 128 | 123 / 129 | 119 / 126 | 135 / 146 |
| Labarin Baƙar fata: Wukong | 58 / 66 | 53 / 64 | 49 / 56 | 51 / 59 | 58 / 68 | 60 / 69 | 59 / 68 | 67 / 76 |
| Cyberpunk 2077 | 124 / 149 | 138 / 166 | 93 / 108 | 99 / 117 | 118 / 139 | 115 / 138 | 122 / 146 | 128 / 167 |
| F1 24 | 182 / 267 | 167 / 263 | 154 / 223 | 156 / 229 | 171 / 251 | 175 / 253 | 173 / 244 | 187 / 275 |
| Hogwarts asalin | 188 / 213 | 196 / 216 | 145 / 160 | 154 / 170 | 173 / 193 | 181 / 197 | 185 / 202 | 193 / 218 |
| Horizon Zero Dawn An Sake Matsala | 144 / 184 | 146 / 187 | 132 / 164 | 130 / 167 | 142 / 182 | 145 / 185 | 137 / 176 | 141 / 186 |
| Metro Fitowa | 79 / 130 | 81 / 140 | 68 / 118 | 73 / 127 | 79 / 146 | 81 / 148 | 80 / 144 | 88 / 167 |
| Red Matattu Kubuta 2 | 128 / 139 | 122 / 128 | 96 / 103 | 104 / 109 | 119 / 126 | 119 / 127 | 119 / 128 | 125 / 132 |
| Sake Takawa | 97 / 177 | 134 / 211 | 75 / 150 | 101 / 155 | 100 / 179 | 91 / 178 | 112 / 176 | 105 / 199 |
| Jimlar Yaƙi: WARHAMMER III | 85 / 105 | 85 / 107 | 76 / 94 | 81 / 97 | 86 / 103 | 88 / 105 | 85 / 103 | 83 / 105 |
| Max. | + 19% | % 9% | % 2% | + 12% | + 14% | + 11% | + 28% | |
| Matsakaici | + 4% | % 18% | % 14% | % 3% | % 2% | % 3% | + 7% | |
| Min | % 8% | % 28% | % 22% | % 10% | % 9% | % 11% | % 5% | |
Kodayake wasu wasannin sun fi son gine-ginen AMD ko NVIDIA, daga mahangar aiki, sabbin samfuran suna da aiki daidai: GeForce RTX 5070 Ti yana ƙasa da Radeon RX 9070 XT da 3% FPS kawai. Idan aka kwatanta da na'urorin ƙarni na baya, Radeon RX 9070 XT shine 3% a bayan Radeon RX 7900 XTX. GeForce RTX 5070 Ti yayi daidai da GeForce RTX 4080 da RTX 4080 SUPER, kuma bambanci tsakanin RTX 4070 Ti SUPER da RTX 5070 Ti ya sauko zuwa ƙimar firam 13%.
Gwajin wasa (2560×1440)
A mafi yawan ma'auni a 1440p, sabbin samfuran AMD da NVIDIA sun cimma ƙimar firam na aƙalla 80 FPS, amma matsakaicin ƙimar firam a cikin Black Myth: Wukong ya ragu zuwa 51-52 FPS.
| 2560 × 1440 | ||||||||
|---|---|---|---|---|---|---|---|---|
| AMD Radeon RX 9070 XT | AMD Radeon RX 7900 XTX | NVIDIA GeForce RTX 4070 Ti | NVIDIA GeForce RTX 4070 Ti SUPER | NVIDIA GeForce RTX 4080 | NVIDIA GeForce RTX 4080 SUPER | NVIDIA GeForce RTX 5070 Ti | NVIDIA GeForce RTX 5080 | |
| Alan wake 2 | 99 / 104 | 103 / 107 | 73 / 77 | 77 / 82 | 90 / 95 | 92 / 97 | 88 / 93 | 104 / 109 |
| Labarin Baƙar fata: Wukong | 45 / 51 | 43 / 50 | 38 / 42 | 39 / 45 | 46 / 52 | 47 / 54 | 46 / 52 | 54 / 61 |
| Cyberpunk 2077 | 80 / 92 | 90 / 103 | 53 / 63 | 54 / 65 | 67 / 80 | 67 / 80 | 75 / 87 | 89 / 102 |
| F1 24 | 159 / 219 | 162 / 228 | 125 / 176 | 139 / 194 | 148 / 211 | 154 / 218 | 155 / 216 | 161 / 231 |
| Hogwarts asalin | 134 / 154 | 141 / 160 | 102 / 114 | 115 / 128 | 124 / 139 | 125 / 142 | 128 / 144 | 141 / 165 |
| Horizon Zero Dawn An Sake Matsala | 133 / 161 | 130 / 160 | 111 / 132 | 114 / 140 | 128 / 156 | 127 / 158 | 122 / 148 | 124 / 163 |
| Metro Fitowa | 69 / 112 | 72 / 120 | 60 / 98 | 64 / 106 | 69 / 124 | 75 / 126 | 70 / 122 | 85 / 145 |
| Red Matattu Kubuta 2 | 114 / 119 | 106 / 111 | 83 / 88 | 89 / 94 | 104 / 109 | 106 / 111 | 106 / 111 | 109 / 115 |
| Sake Takawa | 89 / 140 | 110 / 162 | 75 / 115 | 81 / 121 | 92 / 138 | 85 / 139 | 92 / 137 | 82 / 154 |
| Jimlar Yaƙi: WARHAMMER III | 80 / 97 | 79 / 97 | 53 / 69 | 59 / 76 | 70 / 88 | 70 / 90 | 72 / 89 | 83 / 97 |
| Max. | + 16% | % 13% | % 5% | + 11% | + 13% | + 9% | + 29% | |
| Matsakaici | + 4% | % 22% | % 17% | % 4% | % 3% | % 4% | + 9% | |
| Min | % 7% | % 32% | % 29% | % 13% | % 13% | % 11% | % 3% | |
Radeon RX 9070 XT da GeForce RTX 5070 Ti an sake raba su ta hanyar ƙimar firam ɗin kusan 4%. Katin bidiyo na "ja" na sabon tsara kuma shine kawai 4% kasa da tsohon flagship - Radeon RX 7900 XTX, kuma GeForce RTX 5070 Ti ya kasance mai maye gurbin samfurin 80th na ƙarni na baya. Idan aka kwatanta da GeForce RTX 4070 Ti SUPER, sabon samfurin NVIDIA yana da sauri 15%.
Gwajin wasa (3840×2160)
Radeon RX 9070 XT da GeForce RTX 5070 Ti gabaɗaya sun dace da wasa akan nunin 4K a matsakaicin ingancin zane. Sai kawai a cikin taken manyan albarkatu kamar Black Myth: Wukong da Cyberpunk 2077 ne mai amfani zai iya jure ƙimar firam ɗin ƙasa da 60 FPS ko amfani da haɓakawa.
| 3840 × 2160 | ||||||||
|---|---|---|---|---|---|---|---|---|
| AMD Radeon RX 9070 XT | AMD Radeon RX 7900 XTX | NVIDIA GeForce RTX 4070 Ti | NVIDIA GeForce RTX 4070 Ti SUPER | NVIDIA GeForce RTX 4080 | NVIDIA GeForce RTX 4080 SUPER | NVIDIA GeForce RTX 5070 Ti | NVIDIA GeForce RTX 5080 | |
| Alan wake 2 | 55 / 58 | 57 / 59 | 40 / 43 | 43 / 46 | 51 / 54 | 50 / 54 | 49 / 52 | 58 / 62 |
| Labarin Baƙar fata: Wukong | 27 / 31 | 27 / 31 | 22 / 25 | 24 / 27 | 28 / 31 | 28 / 32 | 28 / 31 | 33 / 36 |
| Cyberpunk 2077 | 34 / 41 | 38 / 44 | 22 / 27 | 23 / 28 | 29 / 35 | 28 / 36 | 32 / 38 | 39 / 46 |
| F1 24 | 110 / 140 | 118 / 155 | 91 / 116 | 99 / 126 | 111 / 145 | 114 / 146 | 112 / 148 | 127 / 167 |
| Hogwarts asalin | 79 / 91 | 81 / 95 | 58 / 65 | 61 / 68 | 70 / 81 | 75 / 83 | 73 / 84 | 90 / 102 |
| Horizon Zero Dawn An Sake Matsala | 89 / 103 | 87 / 102 | 65 / 77 | 74 / 87 | 83 / 98 | 85 / 100 | 75 / 89 | 86 / 103 |
| Metro Fitowa | 52 / 80 | 52 / 86 | 44 / 66 | 47 / 72 | 55 / 85 | 56 / 86 | 54 / 86 | 67 / 103 |
| Red Matattu Kubuta 2 | 79 / 83 | 78 / 81 | 54 / 59 | 60 / 64 | 70 / 76 | 68 / 76 | 69 / 77 | 78 / 83 |
| Sake Takawa | 55 / 84 | 68 / 98 | 49 / 69 | 44 / 72 | 62 / 86 | 57 / 86 | 57 / 86 | 68 / 99 |
| Jimlar Yaƙi: WARHAMMER III | 40 / 57 | 42 / 59 | 29 / 39 | 33 / 44 | 39 / 52 | 40 / 53 | 39 / 53 | 47 / 63 |
| Max. | + 17% | % 17% | % 10% | + 6% | + 8% | + 8% | + 29% | |
| Matsakaici | + 5% | % 25% | % 19% | % 4% | % 3% | % 4% | + 12% | |
| Min | % 2% | % 34% | % 32% | % 15% | % 12% | % 14% | 0% | |
Babban ƙuduri bai canza matsayi na manyan masu fafatawa ba: Radeon RX 9070 XT har yanzu ya fi GeForce RTX 5070 Ti da 4% a matsakaicin ƙimar firam, amma a lokaci guda yana 5% a bayan Radeon RX 7900 XTX. GeForce RTX 5070 Ti kanta har yanzu tana maye gurbin GeForce RTX 4080 da RTX 4080 SUPER, amma fa'idar akan RTX 4070 Ti SUPER ya girma zuwa 18% FPS.
Ray Tracing Gaming Tests
Wasannin binciken Ray sune mafi mahimmancin gwajin da sabbin masu haɓakawa na AMD za su wuce bayan shekaru na baya bayan hanyoyin NVIDIA a wannan yanki. Abin farin ciki, akwai manyan canje-canje don mafi kyau. Kodayake a cikin wasanni masu ratsa jiki Radeon RX 9070 XT yayi kama da Radeon RX 7900 XTX kuma a matsakaici har ma yana bayan tsohon flagship, a cikin binciken ray Radeon RX 9070 XT ya riga ya riga ya ci gaba tare da rata na 20-21% FPS.
Gaskiya ne, akwai wasanni inda bambanci ya kasance ƙananan (Metro Fitowa da Komawa), kuma a Indiana Jones da Babban Da'irar ƙimar firam akan Radeon RX 9070 XT ya ragu zuwa 1 FPS koda lokacin amfani da FSR, wanda a bayyane yake saboda rashin VRAM. Koyaya, wannan wasan ya buɗe hanya ne kawai ga masu “ja” GPUs a cikin sabon ƙari. Abubuwan ingantawar direban da ake buƙata suna iya ɓacewa, don haka ba mu haɗa da Indiana Jones da sakamakon Babban Circle ba a cikin ƙididdige adadin tsakanin mahalarta gwajin. Af, a cikin duk mahalarta gwajin, Radeon RX 7900 XTX kawai yana da adadin adadin ƙwaƙwalwar bidiyo don gudanar da Indiana Jones da Babban Circle akan allon 4K tare da mafi girman ingancin hoto. Wasu na'urori ba za su iya yin hakan ba ko da tare da mafi girman saitunan sikelin firam ɗin.
| 1920 × 1080 | ||||||||
|---|---|---|---|---|---|---|---|---|
| AMD Radeon RX 9070 XT | AMD Radeon RX 7900 XTX | NVIDIA GeForce RTX 4070 Ti | NVIDIA GeForce RTX 4070 Ti SUPER | NVIDIA GeForce RTX 4080 | NVIDIA GeForce RTX 4080 SUPER | NVIDIA GeForce RTX 5070 Ti | NVIDIA GeForce RTX 5080 | |
| Alan wake 2 | 59 / 62 | 47 / 50 | 52 / 56 | 58 / 61 | 67 / 71 | 66 / 70 | 67 / 71 | 76 / 82 |
| Labarin Baƙar fata: Wukong | 19 / 24 | 11 / 14 | 31 / 38 | 34 / 42 | 40 / 48 | 41 / 50 | 41 / 48 | 49 / 57 |
| Cyberpunk 2077 | 34 / 40 | 28 / 33 | 36 / 46 | 38 / 49 | 47 / 59 | 47 / 59 | 42 / 56 | 50 / 65 |
| F1 24 | 121 / 157 | 70 / 130 | 97 / 128 | 104 / 137 | 104 / 154 | 110 / 157 | 102 / 152 | 110 / 169 |
| Hogwarts asalin | 110 / 128 | 94 / 109 | 95 / 115 | 104 / 123 | 117 / 140 | 120 / 142 | 119 / 142 | 134 / 160 |
| Indiana Jones da kuma Great Circle | N / A | 16 / 17 | N / A | 41 / 43 | 46 / 48 | 46 / 49 | 45 / 47 | 51 / 54 |
| Buga na Enched na Metro Fitowa | 73 / 105 | 68 / 104 | 60 / 94 | 62 / 99 | 71 / 114 | 69 / 117 | 70 / 109 | 71 / 125 |
| Sake Takawa | 75 / 137 | 89 / 132 | 91 / 130 | 92 / 134 | 91 / 154 | 92 / 151 | 90 / 151 | 101 / 169 |
| Max. | % 1% | + 58% | + 75% | + 100% | + 108% | + 100% | + 138% | |
| Matsakaici | % 16% | + 3% | + 10% | + 27% | + 29% | + 25% | + 44% | |
| Min | % 42% | % 18% | % 13% | % 2% | 0% | % 3% | + 8% | |
| 2560 × 1440 | ||||||||
|---|---|---|---|---|---|---|---|---|
| AMD Radeon RX 9070 XT | AMD Radeon RX 7900 XTX | NVIDIA GeForce RTX 4070 Ti | NVIDIA GeForce RTX 4070 Ti SUPER | NVIDIA GeForce RTX 4080 | NVIDIA GeForce RTX 4080 SUPER | NVIDIA GeForce RTX 5070 Ti | NVIDIA GeForce RTX 5080 | |
| Alan wake 2 | 38 / 41 | 30 / 33 | 34 / 37 | 38 / 40 | 46 / 48 | 45 / 48 | 45 / 48 | 54 / 56 |
| Labarin Baƙar fata: Wukong | 12 / 16 | 6 / 9 | 20 / 24 | 22 / 27 | 26 / 32 | 27 / 33 | 26 / 31 | 32 / 37 |
| Cyberpunk 2077 | 21 / 24 | 17 / 20 | 23 / 27 | 24 / 29 | 30 / 35 | 31 / 36 | 28 / 34 | 35 / 42 |
| F1 24 | 92 / 111 | 61 / 91 | 77 / 90 | 78 / 93 | 90 / 112 | 92 / 114 | 89 / 111 | 99 / 127 |
| Hogwarts asalin | 74 / 90 | 63 / 78 | 67 / 83 | 72 / 87 | 82 / 100 | 83 / 101 | 82 / 100 | 97 / 116 |
| Indiana Jones da kuma Great Circle | N / A | 10 / 11 | N / A | 30 / 32 | 30 / 33 | 34 / 36 | 31 / 32 | 38 / 40 |
| Buga na Enched na Metro Fitowa | 58 / 83 | 57 / 80 | 50 / 70 | 54 / 76 | 61 / 89 | 62 / 91 | 56 / 83 | 69 / 99 |
| Sake Takawa | 61 / 103 | 74 / 100 | 66 / 94 | 67 / 98 | 78 / 116 | 67 / 115 | 78 / 113 | 87 / 131 |
| Max. | % 3% | + 50% | + 69% | + 100% | + 106% | + 94% | + 131% | |
| Matsakaici | % 17% | 0% | + 8% | + 28% | + 30% | + 25% | + 48% | |
| Min | % 44% | % 19% | % 16% | + 1% | + 3% | 0% | + 14% | |
| 3840 × 2160 | ||||||||
|---|---|---|---|---|---|---|---|---|
| AMD Radeon RX 9070 XT | AMD Radeon RX 7900 XTX | NVIDIA GeForce RTX 4070 Ti | NVIDIA GeForce RTX 4070 Ti SUPER | NVIDIA GeForce RTX 4080 | NVIDIA GeForce RTX 4080 SUPER | NVIDIA GeForce RTX 5070 Ti | NVIDIA GeForce RTX 5080 | |
| Alan wake 2 | 17 / 20 | 14 / 16 | 4 / 5 | 19 / 20 | 22 / 24 | 22 / 24 | 22 / 24 | 27 / 29 |
| Labarin Baƙar fata: Wukong | 5 / 7 | 3 / 4 | 5 / 10 | 11 / 13 | 13 / 16 | 13 / 16 | 13 / 16 | 16 / 19 |
| Cyberpunk 2077 | 10 / 11 | 8 / 9 | 9 / 12 | 11 / 13 | 14 / 16 | 14 / 17 | 13 / 16 | 16 / 20 |
| F1 24 | 51 / 59 | 35 / 48 | 43 / 48 | 46 / 52 | 55 / 61 | 56 / 62 | 55 / 61 | 63 / 72 |
| Hogwarts asalin | 40 / 51 | 34 / 44 | 34 / 45 | 39 / 50 | 45 / 58 | 46 / 59 | 44 / 58 | 53 / 67 |
| Indiana Jones da kuma Great Circle | N / A | 5 / 6 | N / A | N / A | N / A | N / A | N / A | N / A |
| Buga na Enched na Metro Fitowa | 38 / 50 | 35 / 46 | 32 / 41 | 35 / 45 | 41 / 53 | 34 / 55 | 38 / 50 | 45 / 60 |
| Sake Takawa | 38 / 59 | 44 / 59 | 38 / 52 | 39 / 54 | 48 / 66 | 47 / 67 | 47 / 65 | 51 / 76 |
| Max. | 0% | + 43% | + 86% | + 129% | + 129% | + 129% | + 171% | |
| Matsakaici | % 17% | % 12% | + 10% | + 33% | + 35% | + 32% | + 57% | |
| Min | % 43% | % 75% | % 12% | + 3% | + 5% | 0% | + 20% | |
A cikin cikakkun sharuddan ƙimar firam, Radeon RX 9070 XT yana da aikin da ake buƙata don samar da lakabi a ƙuduri har zuwa 1440p kuma, a tsayin, 4K.
Hakanan yana tafiya don GeForce RTX 5070 Ti, kuma duk da haka matsakaicin ma'aunin yana nuna na'urar NVIDIA tana jagorantar 25-32%, galibi saboda cikakkun wasannin da aka gano. A cikin nasa sansanin, GeForce RTX 5070 Ti ya tabbatar da matsayinsa a matsayin magajin GeForce RTX 4080 da RTX 4080 SUPER: fa'idar wannan ta ta'allaka ne a cikin kewayon 1 zuwa 5%. Idan aka kwatanta da GeForce RTX 4070 Ti SUPER, aikin ya karu da 13 zuwa 19%.
Gwaje-gwajen wasa tare da gano ray da sikelin firam
Godiya ga Madaidaicin haɓakawa, na'urori masu tasowa na gaba daga AMD da NVIDIA suna ba da garantin ƙimar firam ɗin jin daɗi a cikin wasannin "matasan" a ƙuduri har zuwa 4K. Koyaya, tare da fa'idar 20-26%, GeForce RTX 5070 Ti ya fi kyau a sarrafa cikakken lakabi.
| 1920 × 1080 | ||||||||
|---|---|---|---|---|---|---|---|---|
| AMD Radeon RX 9070 XT | AMD Radeon RX 7900 XTX | NVIDIA GeForce RTX 4070 Ti | NVIDIA GeForce RTX 4070 Ti SUPER | NVIDIA GeForce RTX 4080 | NVIDIA GeForce RTX 4080 SUPER | NVIDIA GeForce RTX 5070 Ti | NVIDIA GeForce RTX 5080 | |
| Alan wake 2 | 99 / 105 | 81 / 85 | 97 / 104 | 101 / 109 | 114 / 122 | 114 / 123 | 116 / 123 | 134 / 141 |
| Labarin Baƙar fata: Wukong | 40 / 49 | 24 / 31 | 61 / 73 | 64 / 79 | 73 / 90 | 74 / 90 | 73 / 87 | 83 / 98 |
| Cyberpunk 2077 | 73 / 84 | 61 / 70 | 82 / 96 | 87 / 100 | 104 / 116 | 105 / 117 | 94 / 110 | 110 / 126 |
| F1 24 | 165 / 220 | 92 / 190 | 113 / 174 | 115 / 177 | 114 / 200 | 116 / 203 | 116 / 201 | 121 / 218 |
| Hogwarts asalin | 177 / 203 | 161 / 189 | 140 / 160 | 158 / 176 | 171 / 183 | 171 / 182 | 172 / 183 | 173 / 183 |
| Indiana Jones da kuma Great Circle | N / A | 27 / 29 | N / A | 55 / 59 | 62 / 67 | 65 / 68 | 62 / 65 | 73 / 75 |
| Buga na Enched na Metro Fitowa | N / A | N / A | 69 / 121 | 75 / 125 | 79 / 139 | 81 / 142 | 75 / 134 | 80 / 148 |
| Sake Takawa | 42 / 170 | 123 / 175 | 96 / 165 | 96 / 173 | 110 / 197 | 108 / 194 | 104 / 193 | 97 / 205 |
| Max. | + 3% | + 49% | + 61% | + 84% | + 84% | + 78% | + 100% | |
| Matsakaici | % 15% | + 3% | + 9% | + 22% | + 23% | + 20% | + 32% | |
| Min | % 37% | % 21% | % 20% | % 10% | % 10% | % 10% | % 10% | |
| 2560 × 1440 | ||||||||
|---|---|---|---|---|---|---|---|---|
| AMD Radeon RX 9070 XT | AMD Radeon RX 7900 XTX | NVIDIA GeForce RTX 4070 Ti | NVIDIA GeForce RTX 4070 Ti SUPER | NVIDIA GeForce RTX 4080 | NVIDIA GeForce RTX 4080 SUPER | NVIDIA GeForce RTX 5070 Ti | NVIDIA GeForce RTX 5080 | |
| Alan wake 2 | 75 / 79 | 61 / 65 | 74 / 79 | 78 / 83 | 89 / 95 | 90 / 96 | 90 / 95 | 104 / 109 |
| Labarin Baƙar fata: Wukong | 28 / 34 | 17 / 22 | 43 / 53 | 47 / 58 | 56 / 67 | 56 / 68 | 55 / 65 | 63 / 75 |
| Cyberpunk 2077 | 49 / 56 | 40 / 47 | 56 / 64 | 60 / 68 | 70 / 79 | 71 / 80 | 66 / 76 | 76 / 87 |
| F1 24 | 139 / 178 | 82 / 151 | 102 / 145 | 110 / 155 | 100 / 162 | 105 / 171 | 111 / 167 | 111 / 177 |
| Hogwarts asalin | 154 / 180 | 116 / 142 | 98 / 115 | 107 / 125 | 125 / 144 | 126 / 146 | 122 / 137 | 144 / 163 |
| Indiana Jones da kuma Great Circle | N / A | 19 / 22 | N / A | 45 / 48 | 54 / 56 | 53 / 56 | 49 / 51 | 58 / 60 |
| Buga na Enched na Metro Fitowa | N / A | N / A | 65 / 104 | 70 / 110 | 77 / 124 | 79 / 128 | 77 / 119 | 75 / 134 |
| Sake Takawa | 29 / 142 | 93 / 151 | 93 / 140 | 74 / 145 | 110 / 165 | 109 / 164 | 102 / 164 | 96 / 182 |
| Max. | + 6% | + 56% | + 71% | + 97% | + 100% | + 91% | + 121% | |
| Matsakaici | % 17% | + 2% | + 9% | + 24% | + 26% | + 22% | + 39% | |
| Min | % 35% | % 36% | % 31% | % 20% | % 19% | % 24% | % 9% | |
| 3840 × 2160 | ||||||||
|---|---|---|---|---|---|---|---|---|
| AMD Radeon RX 9070 XT | AMD Radeon RX 7900 XTX | NVIDIA GeForce RTX 4070 Ti | NVIDIA GeForce RTX 4070 Ti SUPER | NVIDIA GeForce RTX 4080 | NVIDIA GeForce RTX 4080 SUPER | NVIDIA GeForce RTX 5070 Ti | NVIDIA GeForce RTX 5080 | |
| Alan wake 2 | 42 / 46 | 34 / 37 | 43 / 46 | 47 / 50 | 54 / 58 | 54 / 59 | 54 / 57 | 62 / 66 |
| Labarin Baƙar fata: Wukong | 15 / 19 | 8 / 11 | 25 / 30 | 28 / 33 | 33 / 39 | 33 / 40 | 32 / 38 | 39 / 45 |
| Cyberpunk 2077 | 25 / 28 | 20 / 24 | 29 / 33 | 31 / 35 | 37 / 43 | 37 / 43 | 35 / 41 | 41 / 48 |
| F1 24 | 91 / 112 | 63 / 95 | 76 / 91 | 80 / 98 | 90 / 113 | 91 / 114 | 90 / 112 | 99 / 128 |
| Hogwarts asalin | 85 / 101 | 72 / 88 | 56 / 67 | 60 / 70 | 70 / 82 | 71 / 83 | 67 / 78 | 82 / 93 |
| Indiana Jones da kuma Great Circle | N / A | 12 / 13 | N / A | N / A | N / A | N / A | N / A | N / A |
| Buga na Enched na Metro Fitowa | N / A | N / A | 51 / 72 | 55 / 78 | 62 / 91 | 63 / 92 | 61 / 87 | 72 / 102 |
| Sake Takawa | 20 / 102 | 81 / 108 | 67 / 91 | 64 / 96 | 78 / 112 | 76 / 112 | 83 / 113 | 93 / 127 |
| Max. | + 6% | + 58% | + 74% | + 105% | + 111% | + 100% | + 137% | |
| Matsakaici | % 16% | + 2% | + 10% | + 29% | + 31% | + 26% | + 47% | |
| Min | % 42% | % 34% | % 31% | % 19% | % 18% | % 23% | % 8% | |
Idan aka kwatanta da Radeon RX 7900 XTX, ingantaccen gine-gine na kwakwalwan kwamfuta na AMD ya samar da Radeon RX 9070 XT tare da haɓaka aikin 18-20% FPS. GeForce RTX 5070 Ti, bi da bi, yana bayan GeForce RTX 4080 da RTX 4080 SUPER har zuwa 4% FPS, amma ya fi GeForce RTX 4070 Ti SUPER da 10-14%.
Gwajin caca a cikin overclocking
Kamar yadda kuke tsammani idan aka ba da ƙarancin saurin agogon GPU, wuce gona da iri na Radeon RX 9070 TX (aƙalla a cikin sigar GIGABYTE GAMING OC) motsa jiki ne mara fa'ida, kawai yana samun ribar ƙimar firam 4%. Palit GeForce RTX 5070 Ti GamingPro, a gefe guda, ya zama 9% cikin sauri sakamakon overclocking.
Gwaje-gwaje a cikin aikace-aikacen samarwa
A cikin ma'auni na Blender, sabon katin zane na AMD ya sami damar fitar da fa'ida mafi girma daga ingantattun hasken rana fiye da Radeon RX 7900 XT. Koyaya, tsohuwar alamar ta gama yin sauri da sauri saboda fa'idar aikinta. Sakamakon haka, ba za a iya yin magana game da gasa daidai ba tsakanin Radeon RX 9070 XT da GeForce RTX 5070 Ti. GeForce RTX 5070 Ti kanta yana ɗaukar matsakaicin matsayi tsakanin RTX 4070 Ti SUPER da RTX 4080.
A cikin ayyukan transcoding ta amfani da Premiere Pro, mai haɓaka "ja" na sabbin tsararraki yana daidai da Radeon RX 7900 XTX, amma yana bayan masu fafatawa "kore" a cikin ƙima gabaɗaya saboda ƙarancin saurin aiki tare da fayilolin RAW. GeForce RTX 5070 Ti, kamar RTX 5080, yana amfani da sabunta H.264 da HEVC decoder, don haka yana da tabbaci ya fi karfin katunan bidiyo na 40-jerin.
Radeon RX 9070 XT yana ɗaukar wuri na ƙarshe a cikin tasirin tasirin GPU na saurin ma'ana, yayin da GeForce RTX 5070 Ti ya bi RTX 5080 da nau'ikan RTX 4080 guda biyu a cikin wannan gwajin.
DaVinci Resolve gwaje-gwajen transcoding ya sanya Radeon RX 9070 XT daidai da Radeon RX 7900 XTX, a bayan GeForce RTX 5080. GeForce RTX 5070 Ti ya fi duk tsofaffin hanyoyin NVIDIA kuma suna bin Radeon RX 9070 XT.
A cikin ma'aunin tasirin tasirin GPU, Radeon RX 9070 XT ya sake faɗi zuwa kasan ginshiƙi. GeForce RTX 5070 Ti, a gefe guda, yana gaba da Radeon RX 7900 XTX kuma shine na biyu kawai ga GeForce RTX 5080.
GeForce RTX 5070 Ti bai dace da yawancin aikace-aikacen CAD ba saboda rashin ingantaccen ingantaccen direba. Anan Radeon RX 9070 XT yana ɗaukar fansa.
Rubutun bidiyo/dekodi
Mai gyara kayan masarufi akan guntuwar Navi 48 ya sami ƙaruwa mara ƙarancin aiki yayin aiki tare da AV1, amma “kore” GPUs har yanzu suna jure wa sauran tsarin bidiyo da sauri.
Hakanan, mai rikodin bidiyo shima ya sami saurin gudu kuma yanzu yana iya yin gogayya da QuickSync na Intel, kodayake yana bayan “kore” NVENC.
Ayyukan kowace watt
Tunda samfuranmu na Radeon RX 9070 XT da Radeon RX 7900 XTX suna da kusan ma'aunin wutar lantarki iri ɗaya, bambancin ingancin wutar lantarki a tsakanin su gabaɗaya an ƙaddara ta hanyar wasan kwaikwayo. Tsohuwar alamar tana kaiwa tare da fa'idar 5% FPS a takamaiman ƙimar firam a cikin rasterization, amma a cikin wasannin da aka gano Radeon RX 9070 XT ya yi tsalle 20%.
| Manufacturer | AMD | NVDIA | ||||||
|---|---|---|---|---|---|---|---|---|
| Samfurin | Radeon RX 9070 XT | Radeon RX 7900 XTX | GeForce RTX 4070 Ti | GeForce RTX 4070 Ti SUPER | GeForce RTX 4080 | GeForce RTX 4080 SUPER | GeForce RTX 5070 Ti | GeForce RTX 5080 |
| GPU | Navi 48 XTX | Navi 31 XTX | AD104 | AD103 | AD103 | AD103 | GB203 | GB203 |
| micro Architecture | RDNA4 | RDNA3 | Ada Soyayya | Ada Soyayya | Ada Soyayya | Ada Soyayya | Blackwell | Blackwell |
| Fasahar aiwatarwa, nm | Farashin TSMC N4C | TSMC N5/N6 | Saukewa: TSMC 4N | Saukewa: TSMC 4N | Saukewa: TSMC 4N | Saukewa: TSMC 4N | Farashin TSMC4NP | Farashin TSMC4NP |
| Matsakaicin amfani da wutar lantarki (FurMark), W | 373 | 372 | 285 | 285 | 332 | 317 | 318 | 397 |
| Aiki/W (ba tare da gano ray ba) | 100% | + 5% | % 2% | + 6% | + 8% | + 14% | + 13% | + 5% |
| Aiki/W (tare da binciken ray) | 100% | % 17% | + 15% | + 44% | + 50% | + 59% | + 55% | + 48% |
Bi da bi, GeForce RTX 5070 Ti a nan shi ma ya zama analogue na GeForce RTX 4080 SUPER, amma wannan katin bidiyo yana da ƙananan TBP idan aka kwatanta da Radeon RX 9070 XT kuma ya zarce dan takarar "ja" a cikin binciken ray. Sakamakon haka, GeForce RTX 5070 Ti shine 13% ko 55% mafi ƙarfin kuzari dangane da hanyar ma'ana.
Takaitacciyar sakamakon alamar wasa ba tare da gano hasashe ba
Ray Tracing Gaming Benchmarks Summary
Takaitaccen ma'auni na wasan kwaikwayo tare da gano hasken haske da sikelin firam
binciken
Daga cikin na'urori guda biyu masu fafatawa, Radeon RX 9070 XT da GeForce RTX 5070 Ti, na ƙarshe shine mafi sauƙin sifa. Katin zane wani bambance-bambancen na GeForce RTX 5080, yanke shi kawai don sanya shi kwatankwacin GeForce RTX 4080 ko RTX 4080 SUPER, kawai tare da ƙarancin shawarar da aka ba da shawarar. Wannan a cikin kanta zai zama abin gafartawa idan ba don gaskiyar cewa GeForce RTX 5070 Ti ya fi RTX 4070 Ti SUPER da 19% FPS a mafi kyau a ƙarƙashin mafi kyawun yanayi, kuma kawai a hukumance yana kashe $ 50 ƙasa, ba a ma maganar ainihin farashin ba.
Radeon RX 9070 XT yana ɗaukar matsayin Radeon RX 7900 XTX a cikin wasannin da aka ba da izini, duk da babban koma baya a cikin ƙarfin ƙididdigewa, kuma mafi mahimmanci, saboda ingantaccen gine-gine, ya ci gaba da 21% FPS a cikin binciken ray. Yana da kyau a ga cewa aƙalla GPU na tsakiyar tsakiyar AMD ya kama kuma ya zarce babban GPU ɗinsa na ƙarshe. Amma kash, ko da wannan karuwa bai isa ya kawo ƙarshen dogon lokaci a bayan NVIDIA ba. Radeon RX 9070 XT ya sami nasara ta alama akan GeForce RTX 5070 Ti a cikin rasterization, amma a cikin wasanni tare da binciken ray, matsakaicin fa'idar GeForce RTX 5070 Ti ya kai 32%, kuma a cikin cikakkun wasannin da aka gano ya ma fi haka. Farashin da aka ba da shawarar Radeon RX 9070 XT ya yi ƙasa da na RTX 5070 Ti, wanda ke daidaita raunin dabaru na RDNA 4. Koyaya, a cikin Rasha, mafi ƙarancin farashin masu fafatawa kusan daidai yake, don haka duk da ci gaban NVIDIA da ci gaban AMD, zaɓi tsakanin Radeon RX 9070 XT da GeForce RTX 5070 Ti yana da sauƙi - aƙalla a yanzu.
Daga cikin ayyukan wasan kwaikwayo na sababbin GPUs, ana ɗaukar wuri na farko ta hanyar haɓakawa - DLSS ko FSR sigar 4. NVIDIA kwakwalwan kwamfuta sun sami ikon samar da firam ɗin da yawa da samfuran canzawa, kuma AMD ta fara amfani da hanyoyin sadarwa na jijiyoyi don wannan dalili. Amma fasahar FSR XNUMX, kamar DLSS, an haɗa shi da kayan aikin ɗan ƙasa, don haka an rarraba kayan aikin ƙirar firam ɗin da ke akwai ga masu “kore” da “ja” accelerators, kuma ba zai yiwu a faɗi cewa katunan bidiyo na GeForce suna goyan bayan DLSS da FSR ba, da Radeon - FSR kawai. Za mu yi la'akari da batun haɓaka inganci da jinkiri mai alaƙa daban a cikin wani bincike mai zuwa.
Radeon RX 9070 XT, kamar GeForce RTX 5070 Ti, yana da 16 GB na ƙwaƙwalwar bidiyo. Wannan yana nufin cewa a cikin ƙarni na yanzu, kawai GeForce RTX 5090 yana ba da ƙarin, yayin da sauran samfuran ke haɗarin shiga cikin yanayin da GPU zai iya ɗaukar matsakaicin ingancin hoto a cikin wasa (har ma tare da taimakon haɓakawa), amma an iyakance shi da adadin VRAM. Ya zuwa yanzu akwai irin wannan wasa guda ɗaya kawai - Indiana Jones da Great Circle, amma wannan shine kawai kararrawa ta farko.
Radeon RX 9070 XT samfurin wasa ne mai tsabta, wanda bai dace da yawancin ayyukan da muka gwada ba (ban da CAD). A gefe guda kuma, GeForce RTX 5070 Ti, ya yi kyau a cikin ayyukan sarrafa bidiyo godiya ga sabunta codec na hardware, amma kuma yana da wuya ya kasance da sha'awar masu sana'a saboda rashin ƙwaƙwalwar gida.
A ƙarshe, bari mu faɗi 'yan kalmomi game da katunan bidiyo waɗanda ke wakiltar Radeon RX 9070 XT da GeForce RTX 5070 Ti a cikin bita - kuma yanzu waɗannan sune mafi arha nau'ikan samfuran biyu akan kasuwar Rasha. The "ja" GIGABYTE accelerator da "kore" Palit suna da mafi girma ikon amfani fiye da wajabta ta da bayani dalla-dalla (a cikin yanayin "Radeon" ya wuce 360 W) da kuma aiki a natse idan BIOS daidai aiki. A lokaci guda, Palit GeForce RTX 5070 Ti GamingPro ya mamaye mafi kyau, kodayake baya ba ku damar haɓaka TBP ba, amma GIGABYTE Radeon RX 9070 XT GAMING OC ba shi da daɗi tare da zazzabi na VRAM sama da 90 ° C da fakitin thermal na ruwa wanda ke wahalar da sabis na katin bidiyo.
source: 3dnews.ru
