Kupa GTX 970

[Leon]

TL;DR,.. nechceš sa rozpísať troška/v skratke o čom sa tam píše ?

[Zoltan Balaton]

O tomto tu pisal asnod,ze nv na dx12 naprd
/

So.. basically, AMD could outperform nVidia with DX12?

Basically yes, R290X ~= 980Ti in DX12 is a scary thought considering how cheap the R290X is ($250-300 vs $650-700 980Ti).

a ja budem musiet kupit nvy zdroj

[scream]

zatial je ten rozdiel jedine v benchmarkoch a alfa verzii jednej hry. Vo vydanych hrach je situacia taka, ze OC 970ka je o chlp rychlejsia ako OC 390tka v 1080p ... Graficke karty sa kupuju vzhladom na aktualnu situaciu, nie na scenare co bude ak bude lebo kym bude, tak dana karta bude zastarala.

[icefire]

Nvidia sa nechce zasa k nicomu vyjadrit, cize ako spotrebitel prihliadam aj na tento fakt. Nezavazi, ze su o chlp vykonnejsi (alebo nie, pisem len ako priklad), ked uz maju za sebou 2 fiaska uz iba pocas 970 modelu (+masivne rozsireny coil whine ako polovicne fiasko on top oproti ostatnym high end modelom).

[Asnod]

Trepes bullshity Scream. Ziadna karta nieje zastarala a dodnes sa pouzivaju aj 10 rokov stare karty, niekomu to totiz staci a to neospravedlnuje to ze kupis tohoto roku kartu ktora je pocas doby kedy ju kupis vykonovo inde ako po vydani novej architektury, kedy zacne byt niekomu nepohodlna a skonci minimalne o uroven nizsie

Po uvedení testovacího dema začala Nvidia rozesílat tiskové zprávy, ve kterých upozorňuje na chybu s vyhlazováním MSAA, která se nachází v demu a proto jsou výsledky takové, jaké jsou.

Tvůrci hry se k tomu ale poměrně rychle vyjádřili. Kód testovacího dema byl pravidelně ověřován s AMD, Intelem i Nvidií a navíc prošel certifikačním systémem Microsoftu pro D3D12. Zdrojový kód dema je už přes rok k dispozici všem partnerům z hardwarové oblasti a inženýři z Nvidie i AMD prý kompilovali kódy téměř každý den. Podle informací se snažili zmiňovanou chybu najít a problém nakonec našli v ovladačích karet od Nvidie.

http://www.dsogaming.com/news/oxide-dev ... benchmark/ hahaha

Oxide’s developer also revealed that NVIDIA’s Maxwell does not support natively Async Compute, and that NVIDIA asked Oxide to disable it for its graphics cards.

Tolko toho podporuje na ukor performance dropu

[Zoltan Balaton]

Ja neviem co to tam pisu inak,teraz som to zacal studovat a Maxwell 2nd gen ma Aasynchronne shadre
takze ukratene su akurat gtx 7xx a nizsie

[scream]

Ja uz sa nebudem vyjadrovat k tomu co som sa vyjadroval... ale ked dame cervenozelene emocie bokom, treba sa zamysliet aj nad tym kolko hier ma zelene logo a ako v tychto hrach behaju cervene karty, s akym oneskorenim su pre tieto hry vydane ovladace, ake efekty a fyziku je mozne v hrach vidiet naviac pri jednej, ci druhej znacke...

Je to sice nefer boj ale podla mojho SUBJEKTIVNEHO pohladu vedie tabor zelenych. Mna ako koncoveho spotrebila ani nezaujima ci to je fer alebo nie.

EDIT: Prave mi blikol update driveru:

What’s New in Version 355.82 WHQL
The latest GeForce Game Ready driver, release 355.82 WHQL, ensures you'll have the
best possible gaming experience for Me
tal Gear Solid V: The Phantom Pain and
Mad Max.

Viete kedy vychadza mad max? ...asi tak.

[Zoltan Balaton]

cela veda je imo v tom, ze nove maxwelly vedia ratat 32 datovych fron..nove radeony 65..tj..nv vie naraz spracovat 1 graficku a 34 vypocetnych front,amd 1 + 64 a tam je ten naskok AMD

maxwell 1 a nizsie vedeli spracovat bud len graficku frontu,alebo len vypocetnu (compute), cize asynchronny shader nichts u nich..a taka 780ti si povedie v hre kde budu pouzite AS velmi biedne

gtx970 by som ale nezatracoval,kedze AS vie

ashes of singularity je ukazka kde budu hodne vyuzite AS..staci sa na to pozret..hrozne graficky nahustene,pocitanie fyziky..

speje mi to ale dost k tomu,ze graficka karta prebera stale vacsiu pracu hlavneho procesora..a nakoniec ten bude neviem naco..asi aby vyratal 15 polygonov

takto som to skratene pochopil ja

edit:
a este ohladom ashes of singularity..v dx12 tam amd len dorovnava nv..v dx 11 ma naopak takmer polovicny vykon..

[Asnod]

Je to jedno, drviva vacsina hier su aj tak konzolovky a tam ti ziadne logo nvidia a ani svatena voda nepomoze. Ak jedna znacka v spolupraci s Ubishitom s prepacenim kur*i cely herny priemysel a zastavuje inovaciu/pokrok tak aj keby mi niekto trepol bejzbalkou po hlave tak si to nekupim.

Ziadny maxwell nepodporuje asynchronne shadery, je uplne jedno co dokaze spracovat bez nich. To vidis v dx11. Nasledok force software emulacie mal za nasledok toho slovicka "podporuje" ten prepad vykonu. Ak clovek pochopi ako funguje pri cpu Hyper transport tak je schopny pochopit ako na rovnakom principe funguju asynchronne shadery. Ziadna veda

[Zoltan Balaton]

hm..tak toto sa velmi mylis a vidiet ze si zrejme zaslepeny
http://diit.cz/clanek/asynchronni-shade ... icke-karty

za dalsie..dnesne radeony nevedia dx 12.1..maxwll 2 ano

[Asnod]

Once again.

Spoiler: ukázať

I see what you guys are saying and I think I've figured out something new that didn't come to me before...

nVIDIA mentioned that Maxwell 2 is capable of queuing 32 Compute or 1 Graphics and 31 Compute. Everyone agrees with this. Now the question becomes... what are these queues tied too?

I assumed that Maxwell 2 had no Asynchronous Compute Engines because nVIDIA has been quite quiet about this. I assumed that they scheduled these queues from within their SMM Units (Compute Units). This would account for the performance deficit.

Now I'm thinking that something else could account for this performance deficit. That is if Maxwell 2 has only a single Asynchronous Compute Engine tied to 32 Queues.

So I began to read up on Kepler/Maxwell/2 CUDA documentation and I found what I was looking for. Basically Maxwell 2 makes use of a single ACE-like unit. nVIDIA name this unit the Grid Management Unit. The CPUs various Cores send Parallel streams to the Stream Queue Management. The Stream Queue Management sends streams to the Grid Management Unit (Parallel to Serial thus far). The Grid Management unit can then create multiple hardware work queues (1 Graphics and 31 Compute or 32 Compute) which are then sent in a Serial fashion to the Work Distributor (one after the other or in Serial based on priority) . The Work Distributor, in a Parallel fashion, assigns the work loads to the various SMMs. The SMMs then assigns the work to a specific array of CUDA cores. nVIDIA call this entire process "HyperQ".
Here's the documentation: http://docs.nvidia.com/cuda/samples/6_A ... HyperQ.pdf

GCN 1.1/1.2, on the other hand, works in a very different manner. The CPUs various Cores send Parallel streams to the Asynchronous Compute Engines various Queues (up to 64). The Asynchronous Compute Engines prioritizes the work and then sends it off, directly, to specific Compute Units based on availability. That's it.
HyperQ is thus bottlenecked at the Grid Management and then Work Distributor segments of its pipeline.

nVIDIAs HyperQ can be thus deemed "in order". HyperQ contains only a single pipeline.

AMDs Asynchronous Compute Engine implementation can be thus deemed "out of order". AMDs implementation contains eight pipelines.

AMDs implementation incurs less latency as well as being superior in terms of efficiently utilizing the available Compute resources.

This explains why Maxwell 2 (GTX 980 Ti) performs so poorly under Ashes of the Singularity. Asynchronous Shading kills its performance compared to GCN 1.1/1.2 (in this case an R9 290x) which barely loses any performance.

GCN 1.1/1.2 are clearly being limited elsewhere, and the explanation I gave (regarding the Peak Rasterization Rate or Gtris/s) still stands as the most plausible cause.

I've been away from this stuff for a few years so I'm quite rusty but Direct X 12 is getting me interested once again.

Ty si ocividne nevedomy.

[Zoltan Balaton]

preco citujes..skus to vysvetlit aby to pochopil aj sedliak,ak to teda dokazes
ja sa len snazim prist na to,kde je pravda..ja nie som ani na jednej strane..prave kupujem novy hw a nebyt teba a oxidu uz som rozudnuty nad 970..trochu sa to skomplikovalo..rad si necham poradit..porovnat dokazy a td..asi to chce este cas,pretoze nic konkretne zatial nevidim,nic co by davalo zapravdu jednej strane..

kazdopadne s 390 si musim kupit novy zdroj

vsade sa pise ohladom poor vykonu 980ti v ashes of singularity..vsade len dohady..
co som nasiel jeden test,co vypada nejak hodnotne a len necituje niekoho ineho,tam im to vyslo takto..takze neviem o com sa tu furt toci


zdroj : http://www.extremetech.com/gaming/21231 ... ad-to-head
asi by sa o tomto mala vytvorit nova tema ..neviem ci nie sme OT

[Asnod]

Ako pre sedlaka mozes vdaka asynchronnym shaderom dostat up to 30-40% vykonu navyse oproti stavu aky je teraz.

[loleqboleq]

takže o rok a pol, keď už aj nejaká DX12 hra bude sa konečne vyrovná fury x 980ti...ovšem to už príde pascal a ten nastaví laťku tak vysoko, že amd zase bude dobehovať

[Asnod]

Dobehovat bude tak akurat pascal lebo co v maxwelli nieje to tam do pascalu musia koli dx12 bohuzial natrepat.