R420 IQ Comparison

[Crusher]

Don't assume anything with me, I merely pointed out what I recognize as a noticible difference. If you can't see the difference in the screen shots, you're either blind, tired or lying to yourself.

Is it possible you're mistaking a polygon edge for a mip-map boundry? The only obvious difference I can see in the red square on the image you posted is that the ground polygons closer to the wall have a blurrier looking texture, but that's more to do with the angle difference between the the polygons (and lack of anisotropic filtering) than any mip-mapping or bi-/tri-linear filtering.

 

[Malfunction]

Explaining trilinear filtering to you would be a waste of time. Assuming you think the NV17 image is correct and the X800 is wrong, I want to know how you came to that determination. It should be simple for someone that already has all the answers.

Don't assume anything with me, I merely pointed out what I recognize as a noticible difference. If you can't see the difference in the screen shots, you're either blind, tired or lying to yourself.

I assumed you pointed this out as evidence that ATI's trilinear filtering method is lowering image quality somehow. If that was not your intention then I apologize.

Thank you, I accept your apology.

I never said their (ATi's) filtering method was bad, I only stated I could see a difference between the images. I have stated that not telling people more about it, covering it in the launch of a product or allowing the feature to be enabled or disabled is disappointing and sinister.

I mean, did they honestly believe they would not be put on a firing line for not being forthcoming in the first place?

 

[Ruined]

and whats witht he "mathematical thing" comment? surely even you can comprehend that the easiest way to see image differences is to compare them mathematically, not with your eye. If there is no mathematical difference, and you insist you "see" one, then you are wrong, period.

Honestly, it doesn't really matter to me if the program you use tells you that the images were both taken from the same source, because its even more obvious from the big zoomed in pics below that there is a difference, regardless of whether your program found it or not. The differences can be seen by eye if you look at the screen in the circles I marked, there is no trickery, the r420 appears subtly sharper while the nv17 appears subtly smoother. If you look back through the past few pages you will see about 10 people have already confirmed the differences. They are there. I provided more proof below.

Full picture:
http://www.reflectonreality.com/nv17r420/

Red circled areas blown up A LOT w/ mouseover
http://www.reflectonreality.com/nv17r420/cita.html
http://www.reflectonreality.com/nv17r420/citb.html

After you've seen it zoomed in, go back and look at the original as you will see the same effect in the big picture as well. Both areas look like they may be near mip transitions, and I don't see the same effect anywhere else on the ground except in those two areas.

Just to reiterate, I really have no problem with ATI's optimization. But I'm not making stuff up

 

[croc_mak]

Quote: "The DX9 reference rasterizer does not produce an ideal result for level-of-detail computation for isotropic filtering, the algorithm used in NV40 produces a higher quality result. Remember, our API is constantly evolving as is graphics hardware, we will continue to improve all aspects of our API including the reference rasterizer."

Interesting question(but a very dumb and stupid question) under current circumstances...

How come Nvidia did not talk about their high quality level of detail calculation at NV40 launch? Why didn't they market it as some "UltraLOD"? They must be trying to mis-lead everyone..They had something to hide...definetly..

Ok...for people who did not catch the sarcasm above...you probably are among the crowd asking the dumb stupid questions to begin with

 

[Bjorn]

How come Nvidia did not talk about their high quality level of detail calculation at NV40 launch? Why didn't they market it as some "UltraLOD"? They must be trying to mis-lead everyone..They had something to hide...definetly..

Yes, i know that it's sarcasm. Though i would understand if Nvidia doesn't want to talk about their exceptional AF quality since it's lower then on their 3-4 year old products.

 

[SiliconAbyss]

I am 1000% astounded at some of you. Honestly, I really am. If you can actually convince yourself that ATi's AF routine is showing mipmap boundries in these screenshots posted, then you are either smoking crack or you need to step away from the computer and never come back. The images ARE IDENTICAL save a slight sharpening on the ATi shots and the odd pixel being different here and there.

*shakes head*

 

[SmuvMoney]

and whats witht he "mathematical thing" comment? surely even you can comprehend that the easiest way to see image differences is to compare them mathematically, not with your eye. If there is no mathematical difference, and you insist you "see" one, then you are wrong, period.

Honestly, it doesn't really matter to me if the program you use tells you that the images were both taken from the same source, because its even more obvious from the big zoomed in pics below that there is a difference, regardless of whether your program found it or not. The differences can be seen by eye if you look at the screen in the circles I marked, there is no trickery, the r420 appears subtly sharper while the nv17 appears subtly smoother. If you look back through the past few pages you will see about 10 people have already confirmed the differences. They are there. I provided more proof below.

Full picture:
http://www.reflectonreality.com/nv17r420/

Red circled areas blown up A LOT w/ mouseover
http://www.reflectonreality.com/nv17r420/cita.html
http://www.reflectonreality.com/nv17r420/citb.html

After you've seen it zoomed in, go back and look at the original as you will see the same effect in the big picture as well. Both areas look like they may be near mip transitions, and I don't see the same effect anywhere else on the ground except in those two areas.

Just to reiterate, I really have no problem with ATI's optimization. But I'm not making stuff up

I actually took Althornin's suggestion and downloaded the Compressonator. It did not find any differences where you had circled in the regular *.png. It also didn't any difference in the blown up pics either - no dots at all. Granted, I do see the slight image differences from going from one to the other wiht mouseover. Howveer, this could be attributable to several factors outside of mipmap transitions or any attempt to cheat or lessen IQ.

 

[bloodbob]

You know what would have made a far better comparision a long corridor flat corridor such which would be used to examine AF.

 

[noko]

This is what I am talking about in Mafia:

Mafia 2xAF Bi

Mafia 2xAF Tri

Bilinear sticks out like a sore thumb in Mafia and isn't even a choice in the game due to this reason I believe, so I had to force it using rTool on my G9700Pro and AF at 2x. Playing it with just bilinear with 2xAF the mipmaps are so obvious that it is unplayable unless you turn on Trilinear filtering. Even with Tri you can still make out the mipmaps fairly easily. I think this would be one of the best conditions to put ATI's optimization method for trilinear to the test. Especially in motion, anything less then Tri is very obvious in this game!

 

[mikechai]

This is what I am talking about in Mafia:

Mafia 2xAF Bi

Mafia 2xAF Tri

Bilinear sticks out like a sore thumb in Mafia and isn't even a choice in the game due to this reason I believe, so I had to force it using rTool on my G9700Pro and AF at 2x. Playing it with just bilinear with 2xAF the mipmaps are so obvious that it is unplayable unless you turn on Trilinear filtering. Even with Tri you can still make out the mipmaps fairly easily. I think this would be one of the best conditions to put ATI's optimization method for trilinear to the test. Especially in motion, anything less then Tri is very obvious in this game!

Thanks noko. Its obvious. Will be a good game to test.
________
Ford Motor Company Of New Zealand Picture

 

[Quitch]

I couldn't see a difference between the shots. Frankly, if ATI can reduce filtering (thus sharpening textures) without reintroducing texture crawl (which would only be possible with an adaptive feature like this) then they are to be praised for finding another way to boost IQ.

However, I do think it needs to be seen in motion, because it's not all about spotting the differences. If anything seems sharper then the danger of texture crawl raises its ugly head. This needs testing.

As I said, if none is seen then this is a *good* feature.

 

[Bjorn]

However, I do think it needs to be seen in motion, because it's not all about spotting the differences. If anything seems sharper then the danger of texture crawl raises its ugly head. This needs testing.

We definitely need to see the differences in motion. Just take the what seems to be bilinear CoD shots on the R420. There were a lot of people that thought it looked better then the trilinear ones just because it was sharper further away.

 

[tEd]

However, I do think it needs to be seen in motion, because it's not all about spotting the differences. If anything seems sharper then the danger of texture crawl raises its ugly head. This needs testing.

We definitely need to see the differences in motion. Just take the what seems to be bilinear CoD shots on the R420. There were a lot of people that thought it looked better then the trilinear ones just because it was sharper further away.

yeah but those people would also think that point sampling looks better or a -15 LOD so they are not a reference

 

[Quitch]

However, I do think it needs to be seen in motion, because it's not all about spotting the differences. If anything seems sharper then the danger of texture crawl raises its ugly head. This needs testing.

We definitely need to see the differences in motion. Just take the what seems to be bilinear CoD shots on the R420. There were a lot of people that thought it looked better then the trilinear ones just because it was sharper further away.

Though it should be noticed that this "bilinear" filtering appears to be due to reviewer error, and not because of the cards adaptive filtering.

 

[Bjorn]

Though it should be noticed that this "bilinear" filtering appears to be due to reviewer error, and not because of the cards adaptive filtering.

Yes, that seems to be the case. Although i'd prefer to get a screenshot with the correct settings so that i can judge for myself.

 

[Quitch]

I took it more as a sign of just how desperate people are to portray this in a negative light, and it has taken all this time for all these threads to finally start reaching the conclusion that, in actual fact, this method may well look better than the naive trilinear method.

There's more testing to be done, but I hope we're past the full on knee jerk phase now, that was just irritating.

 

[Bjorn]

I took it more as a sign of just how desperate people are to portray this in a negative light, and it has taken all this time for all these threads to finally start reaching the conclusion that, in actual fact, this method may well look better than the naive trilinear method.

There's more testing to be done, but I hope we're past the full on knee jerk phase now, that was just irritating.

It's not that difficult to understand that people are upset in the light of Ati's PR documents talking about optimizations and the same workload when doing filtering + that you should use color mipmapping tools to check to check what the card is actually doing.

And claiming that they've passed the WHQL tests as a proof of IQ is hardly making things better.

 

[blackfish]

16AF in Halo with normal Ati AF only applying trilinear to the top texture:
http://www.gaeugf.ch/ted/halo2.jpg

16AF in Halo with trilinear forced through all stages with a 3rd party application:
http://www.gaeugf.ch/ted/halo1.jpg

Note the wall to the left, draw your own conclusions.

Is this representing quality degradation of ati's new filtering feature or another "mistakenly" grabbed shot?
This is from
http://www.nvnews.net/vbulletin/showthread.php?t=28942&page=5&pp=15

By the way, is it normal that first ss showing zig-zag type mipmap boundaries if it was bilinear?
Not enough informations were given in original forum. No detailed info about system, driver setting, etc.

 

[Xmas]

Interesting question(but a very dumb and stupid question) under current circumstances...

How come Nvidia did not talk about their high quality level of detail calculation at NV40 launch? Why didn't they market it as some "UltraLOD"? They must be trying to mis-lead everyone..They had something to hide...definetly..

Ok...for people who did not catch the sarcasm above...you probably are among the crowd asking the dumb stupid questions to begin with

Well, you're right, they had something to hide - from people who don't understand a bit about LOD selection and aliasing, and consider the algorithm used in NV40 as inferior and cheap approximation/optimization, while in reality it's exactly the other way round. Ironically, ATI uses this now to claim "Better isotropic filtering". Ridiculous. And it might have made reviewers more aware of the reduced quality of AF.

The same could be said about an "exact" "fast trilinear" algorithm that takes the "right" 16 texels from one mip level. For box-filtered mipmaps this would indeed be better than trilinear. Maybe ATI is using something similar...

btw, I see the differences in Ruined's magnified images. And they are measurable.

 

[Skinner]

I'm detecting NV40 brilinear in 10 seconds, while it took me 5 minutes to choose R420 @ 16xAF as the sharpest shot, yam!

Defenitly not in my eyes, the NV40 is a little sharper, look at the brickwall and the textures at the doorway on the floor.