AMD HD68XX Serija - oficijelni thread

[TheBear]

Pravi napredak kada je teselacija u pitanju se izgleda ocekuje od 6950 i 6970 modela, pominje se i skalabilna teselacija poput nvidia-ine.

Pa verovatno, mada i ovo u 68xx seriji nije uopste los napredak, stavise nisu karte mngoo gore od GTX460 u teselaciji... Vec 470 i 480 su dosta bolje po tom pitanju....

Moram da instaliram Metro da probam, da vidim dal radi bolje od 5750 Crossfire-a, psoto sam primetio da se dobro pokazala 6870 u metro-u....

 

[monteboy]

vMem mod

http://vr-zone.com/articles/amd-radeon-hd-6870-memory-voltage-modification--with-a-pencil/10160.html?doc=10160

 

[TheBear]

Spremam jedan User Review za PCAxe Testiracu igre u tri rezolucije 1280x1024, 1680x1050 i 1920x1080 sve sa po 4xAA, i 3DMArk06 i Vantage na default setovanjima.

 

[Chabex]

Super. Sto se mene tice mislim da 1280 rezoluciju ni ne moras da uradis, cenim da to slabo koga zanima i tako bi sacuvao sebe od neportebnog cimanja. Ali kako hoces naravno

 

[Diabolique]

Predlazem rezoluciju 1360x768.

 

[Villy]

Predlazem rezoluciju 1360x768.

Ali opet, ko kupuje kartu od 200+kusur evra da se igra na toj rezoluciji. Mislim da je 1600:900, 1650:1080 i Full HD standard. Odavno nisam video test na netu u nekoj manjoj rezoluciji (naravno 1920:1200, ali bearov monitor to ne podržava). 1280 je stvarno bzvz raditi...

 

[animaN]

Predlazem rezoluciju 1360x768.

+1 ako nije problem! Rezolucija je veoma polularna zbog sve uchestalijih HD Ready LCD-ova.

 

[solajam]

ko ima

Spremam jedan User Review za PCAxe Testiracu igre u tri rezolucije 1280x1024, 1680x1050 i 1920x1080 sve sa po 4xAA, i 3DMArk06 i Vantage na default setovanjima.

neka pozajmi msi hawk talon attack Medi pa da se uporede ove 2 karte u dx11.

 

[TheBear]

Aha, posto vidim da Animan i Diabolique sa razlogom traze 1366x768 onda cu igre koje sam vec testirao u 1280x1024 testirati u toj rezoluciji.

Ubacicu i Furmark u review isto kao Mipko i Lukija.... 3DMArk06 necu raditi jer dobijam smesan rezultat zbog procesora...

Bice DX10 i DX11 igara, ukljucujuci i Unigine Bnchmark sa extreme teselacijjom i sve kao sto rekoh radim sa 4xAA.

 

[Diabolique]

Aha, posto vidim da Animan i Diabolique sa razlogom traze 1366x768 onda cu igre koje sam vec testirao u 1280x1024 testirati u toj rezoluciji.

Ne 1366, već 1360. Plazma uglavnom koristi 1366, a LCD 1360.
Animan i ja koristimo 1360.

 

[TheBear]

U review cu ukljuciti sledece igre:
DX10
Crysis Warhead
Call of Juarez
Resident Evil 5

DX11
Lost Planet 2 (Benchmark)
Dirt 2
Unigine heaven
Stalker Call of Pripyat
Metro 2033 (mozda)

i 3DMark Vantage uradicu Performance, High i Extreme (stavicu Extreme podesavanja ali 1080p rezoluciju jer ne mogu 1920x1200).

Rezolucije ce biti 1360x766, 1680x1050 i 1920x1080 sve sa po 4xAA.

Uskoro Voltage Control za Radeon 6xxx karte ! Nemaju Volteru nego neki Chil kontroler CHL8214

Originally Posted by Unwinder
Good news. First, I've almost finished implementation of full AMD RADEON 68xx family support so I hope that we'll release new beta including core voltage control via CHL8214 and effective low-level monitoring for 68xx series really soon, probably even tomorrow. Second, during the last weeks I've invested a lot of time into polishing our new video capture engine. The most important changes/improvements include the following things:

- Now the frames are stored into compression queue in raw 3D application’s pixel format and converted to 24-bit RGB color space by compression threads explicitly before the compression process. Previously color space conversion was performed directly in 3D application’s renderer thread context before sending a frame to the compression queue. New implementation minimizes 3D application’s framerate decrease during capturing video due to moving the conversion tasks from 3D renderer to compression threads. Some other trick (improved memory management, minimized extra RGB<->BGR swapping etc) have been also implemented to improve situation in this area.

- Due to pipelined nature of 3D rendering process, each frame capture iteration results in stalling rendering pipeline and forcing the GPU to finish all queued rendering requests before the frame can be transferred from GPU to CPU and processed by it. This results in rather serious variations in frame rendering times caused by periodic pipeline flushing during capturing a frame, especially considering that graphics driver’s internal pre-rendering techniques are in use. For example, if you’re using default graphics driver settings configured to pre-render 3 frames ahead and recording video at 30 FPS (i.e. capturing a frame each 33ms), then graphics card should flush up to 3 pre-rendered frames from the queue each time new frame for is about to be grabbed (i.e. each 33ms in our case) then re-fill the pipeline. It is not an issue when VSync is disabled and framerate is rather high (so single frame time is much lower than 33ms), your eyes simply won’t notice a delay caused by periodic pipeline flush. However, it can be really noticeable in form of heavy stuttering when VSync is enabled. Some competing commercial video capture tools just leave it as is, providing really high framerate when recording video with VSync disabled but causing bad stutter when recording video with VSync enabled. Other tools use smarter approach and forcibly flush rendering pipeline on each frame, even when frame is not supposed to be captured, to provide smoother framerate. Afterburner used the second approach in the previous version, but now the pipeline flushing can be forcibly disabled via configuration file now allowing you to compare "apples to apples" when analyzing performance of different video capture tools.

- Now video encoder library comes with configuration file allowing you to enable detailed logging, showing how many CPU time is eaten by video capture in host (i.e. 3D application) thread, in assembler thread and in all compressor threads. Such performance profiling tool allows to identify the bottlenecks easily, as well as allows you monitoring the effect of changing different encoding related options.

- I've developed my own RTV1 and RTV2 (RivaTuner Video V1 and RivaTuner Video V2) video codecs in addition to Motion JPEG codec used in the previous version. The codecs are heavily optimized for MMX/SSE2 and provide much better encoding performance (RTV1 is roughly 7 times faster than MJPG, RTV2 is roughly 2.5 times faster) whilst keeping rather good compression ratio (6x for RTV1 and 8x-12x for RTV2). Those codecs would be optimal choice for slow dual-core CPU owners, for example on my old dual-core E6600 I can encode MJPG video at 1680x1050 at 30FPS in half frame size only, whilst RTV1 provides smooth 30FPS video in full frame size. But these new codecs have drawbacks too - they are based on lossy DXTC compression algorithms so image quality artifacts can be noticed on gradient color areas or in too contrast color areas. The codecs are not required to be installed during recording, but they are required to play video encoded in RTV1/RTV2 format. The installer is not installing the codec yet, you have to manually copy its' .DLL from .\Bundle\OSDServer\Codec\ folder to %WINDIR%\System32 folder (or %WINDIR%\SysWOW64 in case of 64-bit OS) and run included .REG file to register it in the system).

 

[maestroBG]

posto sam bio na odmoru desetak dana pa nisam u toku,

pitam se da li se isplati zameniti 5850 i uzeti 6870 ili sacekati da izadje cayman?

 

[mINUS]

Isplati ti se ako imas kesa za bacanje ili ako dobijes odlicnu cenu za svoju 5850, pa doplatis nesto malo za 6870, ja cekam 5950, tj. malo pred izlazak za koju nedelju

 

[maestroBG]

verovatno cu da sacekam caymana

 

[TheBear]

Definitivno se ne isplati, 5850 je losija na defautl nekih 5-10% samo od 6870 (sa trenutnim drajverima).

U teselaciji je 6870 bolja oko 20% od 5850 def. vs. def. jedino se na tome primetno poradilo.

Oni koji imaju 5850 i 5870 najbolej da sacekaju Cayman, mada ko zna kolik oce kostati, ali ako opravda performansama bice dobro...

 

[solajam]

...

odokativno koje bi bile cene za cayman ? 6950 i 6970 (ukoliko budu bas ove oznake)

 

[SORD]

Ako ima neke logike menjace 5870 i 5970 po cenama.

 

[TheBear]

Ako ima neke logike menjace 5870 i 5970 po cenama.

Tako nesto i ja mislim, ali siguran sam da ce 6950 biti dosta bolja od 5870.

 

[monteboy]

Kako Fudzilla javlja izgleda su u okruzenju AMD'a ili ti njegovi partneri mishljenja da bi novi 6970 sa Cayman XT chipom mogao da ponovi uspeh 9700 Pro i naslednika 9800 Pro !
AMD je stalno ponavljao kako je hd6800-serija update dosadasnje 5800 serije ali lukavo izostavio da to nije slucaj kod nove HD6900 serije i da se tu radi o novoj arhitekturi
Ta taktika iznenadjenja je uspela zadnji put kod prelaza sa 4870 na 5870 koje su takodje bile dosta brze nego ocekivano.

Sam procenat razlike u odnosu na 5870 (111%) govori dovoljno o tome da AMD nije lagao kad je rekao da ce Tessellatia u novoj seriji biti znatno poboljsana

 

[SORD]

http://event.msi.com/vga/afterburner/images/MSIAfterburnerSetup210Beta2.zip

Sa ovim bi trebalo da moze da se dize voltaza na Ati 68xx.


http://www.xtremesystems.org/forums/showthread.php?t=261272

 

[TheBear]

http://event.msi.com/vga/afterburner/images/MSIAfterburnerSetup210Beta2.zip

Sa ovim bi trebalo da moze da se dize voltaza na Ati 68xx.


http://www.xtremesystems.org/forums/showthread.php?t=261272

Samo za 6850 nazalost.

 

[SORD]

Moracemo onda sacekati cika Unwindera da zavrsi novi MSI AB.

 

[kolev]

kakvo je stanje sa minimalnim fps-om sa ovom serijom, gtx 460 se odlicno pokazao na tom polju?

 

[monteboy]

The Return of the King



Prve informacije govore da ce novi Chip biti veci od starog , da AMD odstupa od 5D-Shader'a i implementira 4D-shader model zbog bolje uposljenosti i performansi. To je inache i arhitekturski razlika u odnosu na HD68XX modele koji i dalje baziraju na starom 5D-shader dizajnu.

 

[Chabex]

Oh yeah! Dobro izgleda sviđa mi se dizajn. Jedva čekam da vidimo šta ume .

E sad što nema ništa normalno (dovoljno interesantno i vizuelno "zaprepašćujuće" što bi opravdalo kupovinu) da se igra sa ovako jakom kartom to je već druga priča..