Geforce FX AA screen shots inaccurate? ** updated! **

[Adul]

http://www.beyond3d.com/forum/viewtopic.php?t=4065

discussion here at beyond 3d, also this post by Kyle


I wonder what has anand heard from nvidia?

GeForce FX Reviews Wrong?
While we are still looking into this, it seems that the in-game screen shots posted on the Net yesterday showing off IQ produced by the GeForceFX 5800 Ultra are "wrong".

There is no doubt that we criticized the GFFX for its AntiAliasing, and now it seems that we may have not had the proper evidence to base our conclusions on. To quote ourselves from this page:


With NoAA you can see the aliasing is quite predominant. 2X AA and Quincunx don?t seem to do much on the GeForceFX visually, but the FPS are effected comparing the shots to the original with no AA enabled.

Of course all of this left us a bit puzzled, and wondering about the AA abilities of the drivers, but the "facts" are the fact correct?

We have been working with NVIDIA on this to get an answer and it seems that now we have the preliminary information to give us a bit more insight on the question.

The GeForceFX's technology applies filters that effect AntiAliasing and Anisotropic filtering after the frame has left the frame buffer. In short, this means that all of our screenshots do not accurately represent the true in-game visual quality that the GFFX can and will produce, as the screen shot were pulled from the frame buffer. We have come to conclusions about the GFFX IQ (Image Quality) that are simply wrong.

While we cannot answer for other reviews of the GeForceFX it is very possible this is an issue with those articles as well, if they were in fact thorough enough to cover IQ.

We are currently working on a way to capture the images properly and will be revisiting the GeForceFX 5800 Preview by covering the IQ portion of our preview with proper screen shot comparison or further addressing the truth surrounding this situation.

Certainly this is a huge issue it seems that NVIDIA was not even aware of when they issued us the review units. Having 48 hours to preview the card over Superbowl weekend compounded this, and while that is no excuse for improper evaluation on our part, it did certainly impact our ability to do a better evaluation. We are sorry for any incorrect evaluations we have made and are working now to remedy the situation. Any new information will be posted here on our news page.

update

Going back yesterday and last night and doing in-game comparisons with the ATI 9700 Pro, it looks to us that NVIDIA's AA is still not up to par with the 9700 Pro's level of quality across the board, but certainly not as lacking as we pointed out at lower levels of AA.

 

[rbV5]

Why do the screen shots show different levels of AA (I'm not just talking 2x) Anands showed differences, wouldn't they all look the same in screenies no matter what the setting if that were true?

 

[Adul]

Originally posted by: rbV5
Why do the screen shots show different levels of AA (I'm not just talking 2x) Anands showed differences, wouldn't they all look the same in screenies no matter what the setting if that were true?

i was wondering that to, but i am sure we will get a more in depth explanation later. I just want to know what's up here. They did after all only have 2 days to review this card, so one cn imagine Anand didn't sleep much if at all.

but the question remains, is there an issue with the way screen shots are captured vs final output?

 

[EdipisReks]

Originally posted by: rbV5
Why do the screen shots show different levels of AA (I'm not just talking 2x) Anands showed differences, wouldn't they all look the same in screenies no matter what the setting if that were true?

it's my impression that there are changes pre and post buffer. therefore, you get some difference in the screen shots, but not the whole effect.

 

[rbV5]

Thats going to make it difficult to make objective comparisons, especially using the screenies as a basis to set the levels for benchmarking...yikes

 

[MichaelD]

And the card still sounds like a leaf blower, no matter WHAT the picture looks like.

 

[Adul]

Originally posted by: MichaelD
And the card still sounds like a leaf blower, no matter WHAT the picture looks like.

well all know that already. just look of the pictures where the GF FX owns a tree

now back to the point about IQ. I just want accuracy that is all

 

[magomago]

Hrm....Now I wonder if Anantech's review were like that...

though their screens did show some MINOR changes in the 2x AA

 

[Adul]

Originally posted by: magomago
Hrm....Now I wonder if Anantech's review were like that...

though their screens did show some MINOR changes in the 2x AA

i was thinking about that, now whether that was the effects of JPG compression or not =\

 

[Rand]

It's been well known for over 9mnths now that the GF4 applied post filtering after the rendering had left the frame buffer so I'm at a loss as to why HardOCP was unaware that the final output wouldnt match that in the frame buffer.
At least for the Quincunx implementation everyone knew long ago that screenshots would not display the full blur filter applied on top of nVidia's 2X RGMS AA as it was done in the DAC stage, thereby making it impossible to screen capture it properly.

Given the evidence pointed out by Uttar at Beyond3D and later expanded upon by others, it certainly seems as though when nVidia's 2X RGMS AA is enabled the two subpixel buffers are not merged. Only one subsample is being rendered in the final output, which is off-centered slightly.

It would certainly explain the screenshots we're seeing and the performance hit relative to no FSAA.


Nonetheless, none of this should affect anything beyond Quincunx and their 2X RGMS.
The screenshots for nVidia's 4X OGMS, and OGMS + OGSS AA should not be affected so they should be valid at least.

 

[Rand]

Perhaps the creator of Hypersnap could devise some way around this, as he was the person that managed to get around 3dfx's post filtering the the VSA-100 and enable proper screen caps.

 

[merlocka]

Rand - what you said makes sense, but the comments by HardOCP made me wonder... The GeForceFX's technology applies filters that effect AntiAliasing and Anisotropic filtering after the frame has left the frame buffer.

I hope someone looks into this ASAP to determine if AF is affected. My assumption is NO, and that AT's IQ comparison with AF + AA is accurate, but I'd like AT to confirm that if they can.

 

[Adul]

i wonder if anand is looking into this has well.

 

[chizow]

Originally posted by: Adul
i wonder if anand is looking into this has well.

Yah definitely interesting. Didn't think Kyle would be able to come up with a revelation of this potential magnitude

N/M forgot Nvidia was in on the discussions

Chiz

 

[Rand]

Originally posted by: merlocka
Rand - what you said makes sense, but the comments by HardOCP made me wonder... The GeForceFX's technology applies filters that effect AntiAliasing and Anisotropic filtering after the frame has left the frame buffer.

I hope someone looks into this ASAP to determine if AF is affected. My assumption is NO, and that AT's IQ comparison with AF + AA is accurate, but I'd like AT to confirm that if they can.

Kyle has clearly misinterpreted nVidia's comments about anisotropic filtering, as what Kyle stated doesnt make much sense in relation to anisotrophy.

To quote him..."The GeForceFX's technology applies filters that effect AntiAliasing and Anisotropic filtering after the frame has left the frame buffer. In short, this means that all of our screenshots do not accurately represent the true in-game visual quality that the GFFX can and will produce, as the screen shot were pulled from the frame buffer."

He seems to be under the impression that anisotropic filtering is applied to data in the frame buffer which is quite erronous. You need texture data to apply anisotropic filtering, not the frame buffer. Anisotropic filtering is never applied to data within the frame buffer... in fact it cannot be done.... yet he says screen shots are inaccurate because it's not done within the frame buffer?

I've not seen any reasons as to why anisotropic filtering would not be captured properly in the screen shots, and Kyle's explanation is obviously impossible.


I'm not sure Kyle understands what he was saying.... he doesnt aware that Quincunx is not screen capped properly which has been known for a long time, and is relatively obvious in any case.
He doesnt seem to grasp how anisotropic filtering is applied to the frame at all.

It's excuseable why he doesnt see that nVidia's explanation shouldnt apply to their 4X + FSAA modes, as I wouldnt have noticed that either had I not seen Uttar's thread at Beyond3D.

 

[Adul]

yes it is clear kyle is where its way over his head , but my question is, Anand what are your thoughts?

 

[Adul]

bump for update

 

[WyteWatt]

Adul interesting but like others have said its too loud still

 

[cmdrdredd]

on the whole it doesn't matter...the 9700Pro looks better (even if marginally) and is faster (even if marginally)

 

[KickItTwice]

There's more news on this at hardocp

 

[chizow]

Originally posted by: cmdrdredd
on the whole it doesn't matter...the 9700Pro looks better (even if marginally) and is faster (even if marginally)

I think thats the crux of the issue; reports from AT and other sites listed the FX being considerably faster with more relaxed AA and AF (aggressive vs. quality) settings, but using the higher settings b/c they achieved comparable IQ.

Chiz

 

[Mem]

I think it`s too early to say one way or the other ,the drivers are beta anyway plus we all know how drivers can effect image quality in general,I`m not saying it`ll make it look better then the 9700 by comparison but I`ll rather wait awhile for more mature drivers and more FX cards (from different brands) to become available.We all know how the 9700 drivers improved from its debut.

 

[Chad]

display the full blur filter

Ick! That's the last thing I want, my images blurred. I just switech from a GeForce 3 to the Radeon and man o man, FSAA actually is awesome now (especially 6x in Nascar 2002). Nice and clean and clear with no jaggies.

With my GeForce I absolutely *HATED* FSAA because it just completely and utterly blurred the entire image!

P.S.

http://home.earthlink.net/~sbp777/pics/nvidiaflow.gif

 

[bluemax]

I'll betcha' the graphics really do have anomalies and nVidia is getting muscular and pushing anything "bad" out of the reviews.
It's happened before.

 

[chizow]

Originally posted by: bluemax
I'll betcha' the graphics really do have anomalies and nVidia is getting muscular and pushing anything "bad" out of the reviews.
It's happened before.

I think there were definitely 2 shooters in the JFK assassination. No way it was a Lone Gunman. I also think we are not alone in the universe and that aliens did land in Roswell, NM.

Chiz