Using the 6800 ultra and control panel af does Nvidia still use 2xaf on all but the first texture stage? Of less importance with brilinear optimizations on is it right that all stages use brilinear but stages 1-7 use a more bilinear filter? When brilinear optimizations are disabled the first stage is trilinear, are stages 1-7 still brilinear?
What level of af is Nvidia using on texture stages 1-7?
- Thread starter dan2097
- Start date
Xmas said:
So does that mean that xbitlabs is justified in saying that ATIs af is not really comparable to Nvidias trilinear optimizations disabled af. I was under the impression although effecting different things, bilinear on later texture stages and a limit of 2xaf on later tetxure stages would sort of cancel out in terms of IQ loss (i.e. one approach looking better in one game). I know in some areas ATIs quality af looks dreadful in halo, although I think you can enable af properly in halo by ini editing or using rtool.
However if they've removed the 2xaf limitation on higher texture stages when using control panel af that would rather inbalance things.
Should have mentioned I'm talking about 60.72. I don't know how 61.11 handle it.
51.75 did limit the higher texture stages to brilinear 2xAF (D3D at least).
52.10 changed this behavior again, to brilinear on all stages but no AF limit. AFAIK there is no AF limit in more recent drivers.
edit: Sorry, this is wrong. 51.75 always limited to 2x on higher texture stages, while 52.10 do limit to 2x only when AF is forced by the driver, not when it is application controlled.
What is comparable depends on what you see on screen.dan2097 said:
51.75 did limit the higher texture stages to brilinear 2xAF (D3D at least).
52.10 changed this behavior again, to brilinear on all stages but no AF limit. AFAIK there is no AF limit in more recent drivers.
edit: Sorry, this is wrong. 51.75 always limited to 2x on higher texture stages, while 52.10 do limit to 2x only when AF is forced by the driver, not when it is application controlled.
radar1200gs
Regular
What I'm about to say should apply to everybody who offers anisotropic filtering, not just one company.
When the user chooses Application controlled anisotropic filtering or forces anisotropic in a control panel without optimizations, then ALL STAGES not just the first stage or the first couple of stages should be based on proper trilinear filtering.
If this does not occur, then you are not doing anisotropic filtering. Its that simple. You can argue all you like that AF is not precisely defined, but all chips offering it prior to R200 always used trilinear filtering for all stages.
Do whatever you want with optimized modes, but in quality and application controlled modes do it properly.
When the user chooses Application controlled anisotropic filtering or forces anisotropic in a control panel without optimizations, then ALL STAGES not just the first stage or the first couple of stages should be based on proper trilinear filtering.
If this does not occur, then you are not doing anisotropic filtering. Its that simple. You can argue all you like that AF is not precisely defined, but all chips offering it prior to R200 always used trilinear filtering for all stages.
Do whatever you want with optimized modes, but in quality and application controlled modes do it properly.
Xmas said:
Only with 6800, and only in the 60.71's
DaveBaumann said:
Dave can you elaborate on what exactly changes with the 61.11s? Also by full angle-dependant af on all texture stages, does that mean no af limit.
I presume you havnt heard anything about whether ATI has any plans to add a "disable trilinear af optimization" or something like that?
PatrickL said:
yah, the 61.11 drivers dont let you disable these 'optimizations', according to techreport's review of the 6800 ultra extreme...
i think the 61.11 drivers were released to reviewers just in time to combat the x800's superior performance at better image quality :?
and the 61.11's added some new issues to far cry. some advancement :?
117597 [TWIMTBP]NV40,WinXP: Corruption when trying to take a screenshot with AA enabled in Far Cry.
117561 [TWIMTBP]NV38/40,WinXP: Corruption of some weapons in Far Cry.
117474 [TWIMTBP]NV40,WinXP: Fog broke in Far Cry.
113476 [TWIMTBP]NV38/40-WinXP: Lighting/shadow problem in FarCry.
117575 [TWIMTBP] NV40/38: Banding visible from wall in FarCry.
Snarfy, stop bashing Nvidia before you get banned like over at nvnews... You sound like a broken record.
Actually, I'd like to see a High Quality option added to Performance/Quality options of anisotropic filtering. Similar to nV, High Quality would be the equivalent of forcing trilinear on all texture stages or removing any trilinear optimizations. I'm pretty sure ATi could default it to Quality in the drivers, but put a disclaimer that High Quality could really had a performance impact. More options are always better for the consumer. 
As for actually returning to uniform AF (okay that is a slight oxymoron) or true AF, i.e., 16x at all possible angles as opposed to adaptive AF, the jury is still out on that one IMHO. I think even the X800 and 6800 could get bogged down too hard at 1600x1200 w/ full 8xaf/16xaf even as powerful as they both seem to be. I think the option would be great with the same disclaimer as above. Do I think it is going to happen? Probably not until a card comes out that is so powerful that it can handle true AF as well as adaptive AF without a large increase in transistor count. Then I could see it coming back. I wouldn't hold my breath though.
As for actually returning to uniform AF (okay that is a slight oxymoron) or true AF, i.e., 16x at all possible angles as opposed to adaptive AF, the jury is still out on that one IMHO. I think even the X800 and 6800 could get bogged down too hard at 1600x1200 w/ full 8xaf/16xaf even as powerful as they both seem to be. I think the option would be great with the same disclaimer as above. Do I think it is going to happen? Probably not until a card comes out that is so powerful that it can handle true AF as well as adaptive AF without a large increase in transistor count. Then I could see it coming back. I wouldn't hold my breath though.
radar1200gs
Regular
Anisotropic filtering is filtering at a level at least equal to or greater than Trilinear filtering. It's the next step up from Trilinear. Bilinear filtering has no place whatsoever in a quality anisotropic filtering setup.
I might grudgingly tolerate brilinear filtering (not on the first two texture stages though) for optimized anisotropic, but proper anisotropic filtering has to use Trilinear as its base.
I might grudgingly tolerate brilinear filtering (not on the first two texture stages though) for optimized anisotropic, but proper anisotropic filtering has to use Trilinear as its base.
You don't really know what you are speaking about.
Given your comments you would actually be frightened if you talked to developers about what filtering they request within their application for many, many textures. Even wondered what that "optimal" filtering mode was in 3DMark03? Thats to select "application" defined filtering whereby FM have chosen the types of textures that require Bilinear and the types that require trilinear - you'd probably be surprised of the mix, and this is just like any other application.
Given your comments you would actually be frightened if you talked to developers about what filtering they request within their application for many, many textures. Even wondered what that "optimal" filtering mode was in 3DMark03? Thats to select "application" defined filtering whereby FM have chosen the types of textures that require Bilinear and the types that require trilinear - you'd probably be surprised of the mix, and this is just like any other application.
As TR pointed out in their X800 review, 61.11 disables the "Tri Opts off" checkbox. So you can check it, but the 6800 remains brilinear.dan2097 said:
ATi defaults to tri on all stages if AF is app-controlled. I suppose you want a "High Quality" setting that forces tri on all layers when AF is set via the CP, though.
Deathlike2
Regular
New sig.
You know, there are reasons NOT to force trilinear filtering on ALL stages. Filtering certain textures may not be worthwhile (high performance hit/quality gain is low).
It should be up to the dev(application) to decide for you.
Dave is right though.
You know, there are reasons NOT to force trilinear filtering on ALL stages. Filtering certain textures may not be worthwhile (high performance hit/quality gain is low).
It should be up to the dev(application) to decide for you.
Dave is right though.
StealthHawk
Regular
dan2097 said:
I was under the impression that NVIDIA removed their texture stage optimizations from the NV3x series before NV40 was launched with some driver like 56.xx.
Does anyone know?
