[ninjaspa]

yeah, it might just help if we had some small bitmap comparisons. i'm sure the professional analysts will have this, but we're all tired of waiting on this arbitrary nda to lift and having to resort to all manner of enthusiast benching which we appreciate all the same.
anywho, about this nda ending on the 14th, can we expect sites to update at midnight eastern time or what? anyone know which official tech sites stay most current?

[Geeforcer]

Originally Posted by Silent_Buddha

As to the shots, the one on the R600 looks ever so slightly blurrier in some places, yet on the other hand appears to show more detail in other places. The nose of the blimp for example.

I don't know if it's "ever so slightly":



The XT shot is messed up, probably mostly by compression - you can see the jpeg artifacts.

[ChrisRay]

I think people may end up finding that the blur is not just a compression thing.

[Pete]

Is this the tent filter in action? I notice BECCA had it enabled in earlier sshots, but he didn't say one way or the other with the 3DM06 ones.

[tEd]

Looks very blurred , which i'm a little surprised. Though being no expert shouldn't the correct weightening of the samples with a tent filter minimze the blur effect.

[Silent_Buddha]

I would imagine a wide tent would have a blurrier effect than the narrow tent. And thus far I haven't actually seen a shot of the 12x setting with the narrow tent. He supposedly did that with the Doom 3 screens, but honestly I see no difference in those so I'm rather doubting it's actually being applied in Doom 3.

There have been shots of the 16x setting with the wide tent and that does tend to be noticably blurrier. Enough so that if that actually proves to the be case in actual review coming up, it wouldn't be a setting I'd probably use.

Regards,
SB

[Colourless]

Ow, the XT images are really badly jpeg artifacted and the GTX ones aren't. It's not a fair comparison at all.

[ninjaspa]

everyone look at the file sizes. one is over 600kb and the other is over just 100k

[Shtal]

So far HD-2900XT disappointed me, especially 6 months late and no answer to G80-GTX.
R600XT 320 stream processors could not win over 128 stream on G80-GTX.
Hmmm.... probably because 16TMU's and 16ROP's

Last hope driver improvement!

[Sc4freak]

I know JPEG artifacts when I see them. The 2900XT shot, when zoomed in, is far blurrier due to massive JPEG artifacting. The reason you don't notice it that much when zoomed out is because that's what JPEG is designed to do.

[rwolf]

Originally Posted by ChrisRay

I think people may end up finding that the blur is not just a compression thing.

:rolls eyes:

[chavvdarrr]

offtopic, but it would be great if there was a way to see what JPEG quality setting was used when saving a shot
Wild guess: 80 for XT shot and 90-100 for 8800

[vertex_shader]

Originally Posted by chavvdarrr

offtopic, but it would be great if there was a way to see what JPEG quality setting was used when saving a shot
Wild guess: 80 for XT shot and 90-100 for 8800

I try it, original .bmp size 3.932.216 byte, than i convert to .jpg, quality 65% file size 126.013 byte (xt picture size 126.125byte), quality 98% file size 635.662 byte (gtx picture size 643.841byte).

[Dalton Sleeper]

The last pic must be more compressed, you clearly see those squares on XT, not a single one on GTX.

[satein]

I would like to ask about this explanation made by Kinc on Nordichardware.com as he posted here...
http://www.nordichardware.com/forum/-vp21040.html#21040

This is what he said...

Originally Posted by Kinc@Nordichardware.com

Im sorry to say that there wont be any faster R600 card then the HD 2900XT. Since ATI didnt know until to late that their 512bit interface and 1024bit ringbus did not work well with GDDR4. 1100MHz GDDR4 is about the same bandwidth as 830MHz GDDR3. The bandwidth at stock is over 100GB/s wich allready is more then enough so there is no need to worry about the bandwidth

It sounds as if the 512bit interface and 1024bit internal ringbus cannot be ultilized at its potential with the GDDR4. Here it is my assumption, might it be possible that the GDDR4 performance get satuarated?

I am just a noob reading on this board and thus having got no good basic about how memory and bus will be ultilized. But I just would like to ask if this might be a reason happened that even AMD/ATi themself haven't expected before hand (yes, GDDR4 spec would sound awlsome in the paper anyway)...

Regards,

[bigtabs]

So, we may find that the 1GB version that we think will arrive in June/July will have GDDR3? Interesting.

Is there a particular time B3D is aiming for to get their first piece out on the R600?

[BlizzardOne]

1100MHz GDDR4 is about the same bandwidth as 830MHz GDDR3.

since when has the 'generation' of memory made a difference in bandwidth at a given speed..?

[Galduta]

HL2: LOST COAST FAA@4x , FA@16x , HDR@ful l 8,37

2900XT/8800GTS:
1600x1200
med:130/126

1920x1200
med:113/112

[Galduta]

Splinter Cell: Double Agent AA@4x in game , FA@16x in CCC.

And this tweak

X:\Program files\Ubisoft\Tom Clancy's Splinter Cell Double Agent\SCDA-Offline\System

[Graphics]
bEnableHDR=True
bTrilinearFiltering=True
bSpecularLighting=True
bVerticalSync=False
bHighQualityVisualFX=True

[WinDrv.WindowsClient]
WindowedViewportX=1920
WindowedViewportY=1080
FullscreenViewportX=1920
FullscreenViewportY=1080
TextureDetail=UltraHigh
SkinDetail=UltraHigh
Brightness=0.000000
Contrast=0.000000
Gamma=0.000000
MipFactor=1.000000
UseJoystick=True
CaptureMouse=True
StartupFullscreen=True
ScreenFlashes=True
NoLighting=False
DeadZoneXYZ=True
DeadZoneRUV=False
InvertVertical=False
ScaleXYZ=1.000000
ScaleRUV=1.000000
MinDesiredFrameRate=10.000000
Decals=True
NoDynamicLights=False
SkinDetail=UltraHigh
TextureDetail=UltraHigh
NoFractalAnim=False
Decals=True

This route

http://www.mediafire.com/?cnymd5mdihd

Splinter Cell: Double Agent AA@4x in game , FA@16x in CCC.

2900XT/8800GTS:
1280x1024
min:24/--
med:38/--
max:57/--

1600x1200
min:18/--
med:28/--
max:42/

no AA, FA@16x in CCC , all maximun

1600x1200
min:40/--
med:57/--
max:87/--

And the 8800 GTX and 1900 XTX in this route ..

http://i4.photobucket.com/albums/y11...o/sc4i1600.jpg

[IbaneZ]

Nice, keep the gaming benchies coming. Tired of 3dmark by now.

Only 27 hours left.

[Love_In_Rio]

It seems the bus bandwith is 128 bits instead 512...

[Arnold Beckenbauer]

Originally Posted by Galduta

HL2: LOST COAST FAA@4x , FA@16x , HDR@ful l 8,37

2900XT/8800GTS:
1600x1200
med:130/126

1920x1200
med:113/112

HL2: LC? Is it the integrated VST (video stress test) or custom timedemo? If VST, there is (unfortunetly) no HDR lightning a la VALVe.

[Galduta]

Originally Posted by Arnold Beckenbauer

HL2: LC? Is it the integrated VST (video stress test) or custom timedemo? If VST, there is (unfortunetly) no HDR lightning a la VALVe.

Is the HardwareOC Benchtools Bench , the 1.6 version (the 2.0 not have lost coast time demo )

edit / with HardwareOC Benchtools Bench 2.2 version


HL2: LOST COAST & HardwareOC Tool v2.2v NO AA , NO AF HDR@full y all máximun.
2900XT/8800GTS:
1600x1200
med:137/--

1920x1200
med:125/--

HL2: LOST COAST & HardwareOC Tool v2.2 AA@4x y FA@16x, HDR@full y all máximun
2900XT/8800GTS:
1600x1200
med:76/--

1920x1200
med:64/--

-------------------------------------------------------------------------------------------------

Oblivion 8,37


OBLIVION, all maximun , FAA@4x y FA@16x , HDR



-2900XT/8800GTS:

Forest

1920x1200
12/15

1600x1200
16/18

1280x1024
19/23

This route

http://img120.imageshack.us/img120/7...ncustombw9.jpg

http://i4.photobucket.com/albums/y117/jonelo/ov1650.jpg


Forest . night , raining

1920x1200
10/10

1600x1200
13/12

1280x1024
16/15

[Razor1]

Originally Posted by tEd

Looks very blurred , which i'm a little surprised. Though being no expert shouldn't the correct weightening of the samples with a tent filter minimze the blur effect.

its not that easy , wieghting isn't an accurate process unless done by hand, I can't even imagine a smart alogrithm, how complex it would have to be to really work well.

Yes the XT does blur, anything above x8 has a slight to large blur effect.

[trinibwoy]

This is going to be interesting to watch. If I understand the process correctly this blurring should be easily noticeable with lower screen resolutions and/or higher-resolution textures where neighbouring pixels can have sufficiently different colors that any blending could really throw things off. I've never used Quincunx but I vaguely remember there being a camp that stood by it and others who denounced it as crap. Something tells me history will repeat itself

Anybody post CSAA vs CFAA shots yet?