GPU Image Quality

[RavenGuard]

I've been overhearing a few arguments about the new nVidia graphics cards vs the new ATi cards. In terms of image quality, does anyone here have a preference or think one looks better than the other? Is there a definitive answer?

From what I remember, the X1950xt used to be the video quality champion, but I haven't really contemplated it until recently.

Thanks for any feedback!

 

[thilanliyan]

Here's an article comparing ATI and nVidia image quality:
http://alienbabeltech.com/main/?p=3188

I think at near equal settings, you'd be hard-pressed to notice a difference while gaming.

 

[RavenGuard]

Very interesting. It seems ATi cards are best with AA but the nVidia cards are much better with anisotropic filtering.

Thanks for the article thilan29

 

[WaTaGuMp]

Back in the days ATI had the better IQ, but I agree now a days they are equal.

 

[KikassAssassin]

I went from an 8800 GTS to an HD 4870, and I didn't see any noticeable changes in image quality at native resolution.

One thing I did notice is that the 4870 seems to do a better job at scaling the image to non-native resolutions, which makes playing games at lower resolutions look better on my LCD (not that I really ever need to turn down the resolution to get good frame rates with a 4870, except maybe in Crysis).

 

[videogames101]

ATI indeed used to be king of IQ, but lately it's been hard to tell.

 

[josh6079]

As far as image quality goes, speaking only of the stark differences when comparing, I would say it's pretty much like this:

nVidia
[*]Super-Sampling AA modes
[*]CSAA (Coverage Sample AA)
[*]Better AF
[*]Better transparency AA in Direct3D games

ATi
[*]CFAA edge-detect AA modes
[*]Better transparency AA in OpenGL games

Both provide 2x, 4x, and 8x multisampling AA that are barely distinguishable at zoomed screenshots.

In addition to this, both provide higher multisampling AA modes - nVidia with CSAA and ATi with CFAA. CFAA provides the best pure edge multisampling AA.

Although, you can also force super-sampling AA unofficially through third party programs such as nHancer and RivaTuner on nVidia hardware and reach edge AA that is negligible in difference (if not a little better) while simultaneously providing AA on more scene content. Granted, its heavy performance often reserves its utility to older or less demanding games.

Overall, nVidia gets my vote. But to those less observant that are just using 4x AA / 16x AF in all of their gaming, inferior AF may not be very noticeable - making it a non-issue for some.

 

[SlowSpyder]

Originally posted by: RavenGuard
Very interesting. It seems ATi cards are best with AA but the nVidia cards are much better with anisotropic filtering.

Thanks for the article thilan29

I think these days to see a difference in IQ you have to look at still screen shots and really examine things closely. I don't believe you'd be able to see a difference between one brand over the other while in game, personally.

 

[BFG10K]

Originally posted by: SlowSpyder

I don't believe you'd be able to see a difference between one brand over the other while in game, personally.

I can easily see it with AF and full scene super-sampling, especially in titles that exhibit shader aliasing.

 

[MarcVenice]

He probably can, and I probably could too. Thing is, when gaming I'm not looking for it. Up untill recently i wasn't even bothered with jaggies, coz when gaming I just would not notice them. That's because my focus would be on the action, in an RTS or FPS, and not on the details, like shadows and edges being jaggie. The action is way to 'fast-moving', explosive' and chaotic, to really notice small details like AA/AF being applied or not, and then in different modes too. But when looking for it, sure you can see a difference. If one is explicably better then the other, meaning you should pay more for videocard A then for videocard B, don't think so.

 

[cusideabelincoln]

When you game a 1280x1024 (or below), jaggies are extremely noticeable (especially on a 19" monitor) and I find them to be a distraction.

 

[MarcVenice]

I've been gaming on 1680*1050 for a long time now. And yes, jaggies are easily noticable, the difference in which Nvidia and AMD videocard's deal with them isn't very noticable though, unless you really look for it. Which I just explained, I usually don't.

 

[josh6079]

Originally posted by: MarcVenice
The action is way to 'fast-moving', explosive' and chaotic, to really notice small details like AA/AF being applied or not...

...I've been gaming on 1680*1050 for a long time now. And yes, jaggies are easily noticable, the difference in which Nvidia and AMD videocard's deal with them isn't very noticable though, unless you really look for it. Which I just explained, I usually don't.

That's one opinion. I myself notice a difference between no AA and no AF even if I'm in intense gameplay.

But as far as how both vendors apply AA and AF, and if both are set at identical levels (e.g., 4x AA / 16x AF), I agree. I may not be able to tell without concentrating on the differences - when trying to get the head shot first my focus is elsewhere.

Nonetheless, the differences are there. And if you want the best image quality, as the OP asked about, you have to step outside what you might be able to see when casually gaming.

Some, even if they don't know if they would catch the difference(s) (in, say, AF quality) might still like to have ease of mind - knowing that they shouldn't encounter anything but the best.

 

[MarcVenice]

I think I worded it wrongly. When I game, be it with a Nvidia or ATI-card, with AA/AF applied, I do not notice the difference between both brands, during actual gameplay. When looking for it, a difference can be spotted.

 

[josh6079]

Yeah, I agree it depends on the gamer. Some are able to spot the differences easier than others.

But as far as which has better image quality - when we sit down and try to find the differences - I'd have to say that nVidia claims that crown at the moment.

 

[SirPaulie]

Can tell the difference with AF and the hybrid modes and concur wth BFG10K.

 

[konakona]

Originally posted by: MarcVenice
He probably can, and I probably could too. Thing is, when gaming I'm not looking for it. Up untill recently i wasn't even bothered with jaggies, coz when gaming I just would not notice them. That's because my focus would be on the action, in an RTS or FPS, and not on the details, like shadows and edges being jaggie. The action is way to 'fast-moving', explosive' and chaotic, to really notice small details like AA/AF being applied or not, and then in different modes too. But when looking for it, sure you can see a difference. If one is explicably better then the other, meaning you should pay more for videocard A then for videocard B, don't think so.

while I agree it is hard to notice anything during the busy parts of action, even the most action packed games gave moments when I would realize how beautifully the backgrounds were done and observe the scenery, not exactly trying hard to discern differences.

I guess there are certain things that stand out more than the other (i.e. fences)

 

[BFG10K]

Originally posted by: MarcVenice

When I game, be it with a Nvidia or ATI-card, with AA/AF applied, I do not notice the difference between both brands, during actual gameplay.

Again, I?d have to personally disagree. When I first started gaming on my 4850 I could see surfaces shimmering that I never saw shimmering on nVidia hardware (or at least shimmered much less), and I wasn?t even looking for it.

I was like ?that surface never used to move so much, what?s going on??

To me it felt like the GF7 with optimizations enabled going against ATi?s HQ AF, back in the day. AF is a day/night difference to me because I?ve been using it since it came out in November 2006.

I also never expected to find a screenshot to show such stark differences, yet I managed to get one for that review with little effort.

nVidia?s superior AF really does enable texture surfaces to sit still more, especially if you combine it with full scene super-sampling. Old games that can use xS modes basically get perfect image quality because not one single pixel moves or is out of place. It?s like playing an offline render.

 

[josh6079]

Originally posted by: BFG10K
Old games that can use xS modes basically get perfect image quality because not one single pixel moves or is out of place. It?s like playing an offline render.

Yeah, running HL2 with 16xS and 16x AF is pretty much perfect.

Although, there is one area I've seen so far that, even under those settings, still some shimmering. That said, it isn't much at all and you really have to look for it.

 

[BFG10K]

Originally posted by: josh6079

Although, there is one area I've seen so far that, even under those settings, still some shimmering. That said, it isn't much at all and you really have to look for it.

Yes, and I'll bet I know exactly which one: the metallic walls of the red buildings where you first get the buggy? There's some pretty bad shader aliasing there that takes a lot of super-sampling to squash, even with nVidia?s AF.

If you're feeling brave you can try some of the pure super-sampling modes (e.g. 3x3), though I personally avoid them due to their lacking edge strength.

This is why I advocate nVida implementing combined modes that have RGSS/SGSS with more than four samples. 4xOGSS is good, but it?s not good enough sometimes, especially with (near) horizontal & (near) vertical lines.

 

[josh6079]

Originally posted by: BFG10K

Yes, and I'll bet I know exactly which one: the metallic walls of the red buildings where you first get the buggy?

Actually I was talking about the gravel in the trench where the train comes through near the beginning.

It's not much, but I can still spot some.

If you're feeling brave you can try some of the pure super-sampling modes (e.g. 3x3), though I personally avoid them due to their lacking edge strength.

Yeah, I've tried them all and like the combined modes better because of how they address that. Personally, 16xS is about the lowest I'd want to go now that I've played with it on a few titles. It covers all areas minimally, in my opinion.

This is why I advocate nVida implementing combined modes that have RGSS/SGSS with more than four samples. 4xOGSS is good, but it?s not good enough sometimes, especially with (near) horizontal & (near) vertical lines.

Have you tried using 4x4 OGSS on that area in HL2 you mentioned above? I've not tried anything higher than 2x2 in the scene I've seen shimmering still occurring.

 

[BFG10K]

Originally posted by: josh6079

Have you tried using 4x4 OGSS on that area in HL2 you mentioned above? I've not tried anything higher than 2x2 in the scene I've seen shimmering still occurring.

Yes, and the performance hit is unacceptable.

And again, the pure super-sampling modes have pretty poor edge AA relative to their sample size. 4x4 is only about as effective as 4xRGMS at (near) horizontal and (near) vertical edges, despite rendering the scene sixteen times bigger.

If any vendor implemented multi-GPU super-sampling on a single board, that would be awesome.

 

[yacoub]

Back in the day ATi was known for better image quality and color saturation/vibrancy. Enough so that it was worth putting up with their less stable drives. But once NVidia caught up (around the 6000-7000 series), it was no longer worth it to deal with the (at the time) noisier, hotter, and (this part seems to still be true) less stable drivers of ATi.
I ran ATi up through the X800 series and then switched to NVidia and haven't had a reason to look back since.

 

[josh6079]

Originally posted by: BFG10K
If any vendor implemented multi-GPU super-sampling on a single board, that would be awesome.

Couldn't the GTX 295 be capable of this?

 

[BFG10K]

Originally posted by: josh6079

Couldn't the GTX 295 be capable of this?

By single board I mean single GPU.