AnandTech meets image quality

[Tim Murray]

Jesus Tapdancing Christ.

This one in particular.

I call bias. Just the friggin location for the screenshot... no floor, no way to judge filtering quality. No mention of what settings were used. 6x AA/16x AF on ATI cards versus 8xAA and 8xAF on NVIDIA?

Christ. I wonder if Derek Wilson would like to come on over and explain what he was smoking while he was writing this article.

--edit--Upon further review, I find that antialiasing and anisotropic filtering are used only a few times. Or maybe they aren't. IT NEVER REALLY SAYS! (with the exception of JA, where it says 4x/8x versus 4x/8x) It says at a few points, "We turned on antialiasing." Thanks, Chief. That's really enlightening. It's also really overly critical of ATI, ignoring any problem from the NVIDIA side as something that will be fixed in the next driver while ATI's problems are just that, problems, and seems to focus entirely upon ATI. Even I will say that. Definitely biased. If you're going to call shenanigans on one side for driver bugs, you'd best at least have the balls to call shenanigans on the other for driver bugs.

This is the kind of thing that gives people who actually prefer NVIDIA cards over ATI cards a bad name.--/edit--

 

[tazdevl]

I noticed the same thing and put a comment on the article. Hope you do the same Baron.

At this point, I'd bet Brian Burke wrote it for Anandtech. No specifics whatsoever... so much for detailed comparisons and benchmarking.

Ever since the NV30 debacle, Anand has continued to lose credibility, journalistic objectivity and integrity IMO.

 

[StealthHawk]

I don't know how you can take Anandtech seriously. I've given up on them as a source of factual verifiable information as they repeatedly resort to ambiguity and hearsay.

 

[cthellis42]

Using "Anandtech" and "biased" in the same sentence seems to have become rather commonplace for a long time now. Didn't detect anything particularly harsh or "Tom-like" from this piece, though. I'm not cruising through all the minutae, however. Quite possibly not picking the best screens and the best methods of testing, but it's hard to read anything truly overt. For those folks who tend to just skip over to the Conclusion anyway, it seems pretty measured and accurate. (I've been up a while though, so I might just be too tired to notice. )

I give it two shrugs up. Hehe...

 

[Tim Murray]

Using "Anandtech" and "biased" in the same sentence seems to have become rather commonplace for a long time now.

I'm sorry, I had forgotten the review a few months back that caused an AT guy to come to the NV News forums. Wowie Zowie, that was funny. Stealth, you should get some links from that (since you'll be up for the next seven hours anyway ).

 

[digitalwanderer]

Using "Anandtech" and "biased" in the same sentence seems to have become rather commonplace for a long time now.

I'm sorry, I had forgotten the review a few months back that caused an AT guy to come to the NV News forums. Wowie Zowie, that was funny. Stealth, you should get some links from that (since you'll be up for the next seven hours anyway ).

Oh, that was a fun one! Please, please, please find a way to lure him back for some more fun!

Did anyone notice on this page the last paragraph?

This has been dubbed “brilinearâ€￾ filtering by some, as the method blends bilinear filtering more heavily than usual into higher order filtering techniques. We have yet to see any perceptible image quality issues arise from this, and as long as banding is not evident and the highest possible resolution textures are used, this filtering method serves its purpose. NVIDIA assures us that if we can point out a loss in image quality in any game, they will correct it.

Oh man, what I wouldn't give to have an FX in one of my rigs right now so I could show nVidia an image quality difference in a game!

Also notice this on the next page:

In general, these images look very similar. The biggest difference we noted was that ATI seems to have a higher threshold for how opaque something needs to be in order to be drawn. Originally, we had noticed this issue in frame 4000, and NVIDIA pointed out an instance of the problem in frame 5100.

(bolding mine)

Nice of nVidia to point out problems with ATi for 'em.

I also liked this bit from page 10:

Also, ATI has helped us track down the motion issue that we've been seeing on their cards. ATI supports using a separate filtering scheme for texture magnification (filtering when the screen pixels are smaller than texels) and minification (filtering when the screen pixels are larger than texels). NVIDIA hardware requires that magnification be done using the same filtering method as minification. Since TRAOD only requests anisotropic filtering be done on texture minification, NVIDIA does anisotropic filtering just fine while ATI doesn't do anisotropic filtering on magnification. The difference in filtering methods causes the flickering effect by which we have been bothered, and could be fixed via a patch from EIDOS; though, ATI reports that EIDOS is unwilling to do so. ATI could fix the problem themselves by doing application detection (determining that TRAOD is running and then adjusting settings specific to that game), but ATI is unwilling to take this step (even though it would be valid and helpful) in order to avoid the controversy.

Now why wouldn't Eidos want to fix it?

And from his conclusion:

The decision on what is acceptable is out of our hands, and we can't really declare a clear winner in the area of image quality. We can say that it appears from the tests we've done that, generally, NVIDIA hardware does more work than ATI. Honestly, it is up to the reader to determine what aspects of image quality are important, and how much of what we covered is relevant.

What does how much work the card does have to do with image quality?

 

[Deathlike2]

The decision on what is acceptable is out of our hands, and we can't really declare a clear winner in the area of image quality. We can say that it appears from the tests we've done that, generally, NVIDIA hardware does more work than ATI. Honestly, it is up to the reader to determine what aspects of image quality are important, and how much of what we covered is relevant.

That wasn't true since Det 44.03->52.16 on UT2003... but that's not covered obviously...

Noone forced NVidia to do more filtering. The driver should be doing what the APPLICATION requests unless the user WANTS to override the application's preference.

 

[FUDie]

Also, ATI has helped us track down the motion issue that we've been seeing on their cards. ATI supports using a separate filtering scheme for texture magnification (filtering when the screen pixels are smaller than texels) and minification (filtering when the screen pixels are larger than texels). NVIDIA hardware requires that magnification be done using the same filtering method as minification. Since TRAOD only requests anisotropic filtering be done on texture minification, NVIDIA does anisotropic filtering just fine while ATI doesn't do anisotropic filtering on magnification. The difference in filtering methods causes the flickering effect by which we have been bothered, and could be fixed via a patch from EIDOS; though, ATI reports that EIDOS is unwilling to do so. ATI could fix the problem themselves by doing application detection (determining that TRAOD is running and then adjusting settings specific to that game), but ATI is unwilling to take this step (even though it would be valid and helpful) in order to avoid the controversy.

"NVIDIA does anisotripic filtering just fine"? Sure, if out of spec is just fine. If the application specifies min anisotropic and mag linear, why would it expect to get something other than that? That's completely out of spec.

Funny that EIDOS bent over backwards to remove a patch for TRAOD because NVIDIA didn't like the benchmark results. Now, EIDOS refuses to issue a simple patch to enable mag anisotropic if supported by the card.

Typical.

-FUDie

 

[andypski]

My understanding is that we render the flashlight in Halo correctly, so I'm not sure where they got the idea that the edges should be more blurry. When presented with two images and having taken no reference image for comparison it seems very strange to make a statement like -

The most noticeable thing in this comparison is that the flashlight on the NV card has soft edges, while the ATI has hard edges. Since soft edges on lighting or shadowing effects are generally harder to calculate, it seems like the NVIDIA card is doing a little more work in this scene

A blanket statement like this, taking knowledge from one area of graphics and applying it as a principle in another seems wholly unwarranted. A softer edge could be the result of any number of things, rather than some mythically larger amount of calculation. I seem to recall some earlier drivers where it was shown that ATI were rendering the beam of the flashlight as a circle, and on some other cards the beam was more spread out (more blurry), and not circular. It turned out that the effect as seen on the original XBox version was circular, just as we were rendering it.

I am aware of no problems with our rendering in Halo - as far as I know our rendering in this application is completely correct.

As to the rest of it - I'm not sure why they would choose to compare texture filtering quality only using control panel forced settings, which have no formal specification, rather than using application settings, or at least doing some comparisons in this mode.

With ATI, we can see that the card doesn't do either anisotropic or trilinear filtering when looking at any texture stage other than 0.

I have no idea where they can have got the idea that we don't do anisotropic on any texture stage other than 0 when using control panel AF as it's completely wrong, and also doesn't appear to be borne out by any of the screenshots that they show. When you do the mouseover effect to show texture stage 1 then there is bilinear filtering as they state, but the level of anisotropy being used does not change. In light of this their statements are very curious.

 

[THe_KELRaTH]

Maybe I've just read too many biased reviews to consider this one really biased.

TRAOD - ATI Better lighting, better AA
As for the problem with ATI AF, it would have been helpful here to know which is considered the correct format - NV single format or ATI's dual format.

HALO - He mentions that soft shadows aren't working on the ATI card - well if that's what was requested then here the NV card wins. I hadn't realised all the missing lighting noted in previous reviews have now been fixed on the NV card (if it hasn't then it should have been mentioned).

Jedi Academy - ATI Better AA, lightsabre glow effect different to Nvidia but as I've not seen a reference shot I couldn't say which is truely correct (I prefer the ATI version)

X2 - ATI Much better AA - Smoother gameplay motion

Tron2 - ATI Better AA

UT2003 - Mentioned the ATI detail texture problem but also mentioned the simple fix - he could have gone further and mentioned that when fix is used then there's far more detail than with the Nvidia images.
Probably could have used a different image shot so it showed that, yet again, Nvidia does a poor job with AA but then again 70% of the game is like the images he used.

As for NV cards working harder, well that's just a silly amateur reviewer thing to say as he's now making pure personal assumptions.
Besides his AF filtering information errors and more assumptions I'd say the biggest flaw with this review is the lack of games / variety tested.

 

[Thowllly]

HALO - He mentions that soft shadows aren't working on the ATI card - well if that's what was requested then here the NV card wins. I hadn't realised all the missing lighting noted in previous reviews have now been fixed on the NV card (if it hasn't then it should have been mentioned).

No, he said that the edge of the light from the flashlight was soft on the NV card and hard on the ATI card (neither has soft shadows). That is simply not true. If you look at the pictures you'll see that the ATI card has a soft edge to the light from the flashlight, while the NV card hardly has any edge at all to the light. While you could say that the edge on the ATI card is harder, it's wrong to say that it is hard. A hard edge would be an aliased edge where each pixel where either lit or unlit.

 

[Enbar]

I've seen it in many web reviews where nvidia's hardware is called "more flexible" like this one in the conclusion. The reason why Nvidia's hardware is always considered more flexible is because they can have longer shaders. I'm pretty sure no game uses shaders longer than ATI's and I doubt many will for some time so the usefullness of this flexibity is debateable. It can also easliy be worked around with multipass.

ATI has more flexible floating point textures and render targets. These features have been used in real games in ways Nvidia can't support. However I've never heard about people calling ATI's hardware being more flexible...

My point here is that they are both more flexible in certain areas but to call one more flexible overall is probably a mistake.

 

[ben6]

Maybe I need coffee in the morning, but my nitpick in this article is this page: http://www.anandtech.com/video/showdoc.html?i=1931&p=4

That's far too simplistic a explanation of how Nvidia AND ATI does FSAA

 

[CorwinB]

Maybe I need coffee in the morning, but my nitpick in this article is this page: http://www.anandtech.com/video/showdoc.html?i=1931&p=4

That's far too simplistic a explanation of how Nvidia AND ATI does FSAA

Err... Is it me, or did they copy-paste an explanation of how Voodoo5 and GF2 were doing AA, with some MS/SS explanation thrown in for good measure ? Did those guys ever heard of gamma-corrected AA, for example ?

 

[nelg]

I've seen it in many web reviews where nvidia's hardware is called "more flexible" like this one in the conclusion. The reason why Nvidia's hardware is always considered more flexible is because they can have longer shaders.

I think it's because the PCB is 1 meter in length.

 

[Bambers]

No, he said that the edge of the light from the flashlight was soft on the NV card and hard on the ATI card (neither has soft shadows). That is simply not true. If you look at the pictures you'll see that the ATI card has a soft edge to the light from the flashlight, while the NV card hardly has any edge at all to the light. While you could say that the edge on the ATI card is harder, it's wrong to say that it is hard. A hard edge would be an aliased edge where each pixel where either lit or unlit.

I would agree with this. It appears to be a bit of an optical illusion of sorts (non identical pictures doens't help). If you look at each picture closely you'll notice that the areas out of the light are much darker on the shot from the 9800 than the 5950, it's this that gives the light on the 5950 shot a softer look imo.

 

[THe_KELRaTH]

Maybe the ATI torch is using Lumiled 5 Watt Luxeonstar LED's

 

[WaltC]

...
As for NV cards working harder, well that's just a silly amateur reviewer thing to say as he's now making pure personal assumptions.
Besides his AF filtering information errors and more assumptions I'd say the biggest flaw with this review is the lack of games / variety tested.

Heh...Heh..."working harder"....I like that....*chuckle*....

I suppose that might be true, if you think about it like comparing a 4-cyclinder engine to an 8-cyclinder. Both engines can get you to 60 mph, but the 4-cyclinder "works harder" while doing it (runs at a higher rpm, makes a lot more noise, runs hotter, and vibrates a lot more, etc.) I'm not sure, though, that "working harder" is where I'd want to be though, in such a case... An old boss of mine instilled in me a concept which has served me well: "Why work harder if you can work smarter, instead, and do a better job?"

People have funny ideas..."works harder"...Heh... Puts me in mind of a horse chasing a carrot, or something.

 

[Brent]

HALO - He mentions that soft shadows aren't working on the ATI card - well if that's what was requested then here the NV card wins. I hadn't realised all the missing lighting noted in previous reviews have now been fixed on the NV card (if it hasn't then it should have been mentioned).

No, he said that the edge of the light from the flashlight was soft on the NV card and hard on the ATI card (neither has soft shadows). That is simply not true. If you look at the pictures you'll see that the ATI card has a soft edge to the light from the flashlight, while the NV card hardly has any edge at all to the light. While you could say that the edge on the ATI card is harder, it's wrong to say that it is hard. A hard edge would be an aliased edge where each pixel where either lit or unlit.

Here is a good comparison of the flashlight thing I did a long while ago: http://www.hardocp.com/article.html?art=NTM3LDQ= look at the bottom of the page under "Screenshots". Notice the 4th picture, from the Truth and Reconciliation mission. That shows the difference very well.

What I haven't heard yet is which way the Xbox renders it.

 

[cthellis42]

Indeed, Brent, I remember your look at the game then was a good comparison--and even in those, there's not a notable "hard edge/soft edge" anything to read into, but a difference in the lighting equation from the flashlight.

What I haven't heard yet is which way the Xbox renders it.

Was definitely answered somewhere, but I can't remember offhand. (Of course anyone with the game on Xbox can boot up and check manually themselves.) The circular method the Radeons adopt are the proper method. (Or at least "closer"--I don't know how exact.) I'll root around, but I figure someone else will remember and post the reference before I can find it anyway. Hehe...