R420 IQ Comparison

[crushinator]

lay off the drugs man. there is NO difference in those red circles.

There are differences. The txtures are sharper on R420, what should normaly cause less smoother transitions, which actually is not detectable by my eyes, and I guess, that's because the next level is sharper than on NV17 too. Some people might call it oversharpened. In that case, we should judge both filtering types seeing them in motion. Then, if you can't detect more aliasing than on NV17 yet, go and thank ATi for their optimisation.

 

[Sworkhard]

BTW, since this is a pattern/synthetic test and not from an actual game, how do we know for sure this is not one of those "special cases" where the ATi optimization is disabled to get the "correct output?"

We don't, but looking at the app name I would guess that this is not the case :|

 

[Sandman]

First off, thank you Dave for going out of your way to help us out.

Second, I have to ask, given what we've seen so far, does this really matter? First, very few people are saying that they can even see a difference. Also, if the only way you can tell a difference is by switching back and forth between screenshots for a while and really concentrating on specific areas, then do you really think you'll ever notice anything in a game?

Of course, the above statement assumes that no new screenshots or videos show real-life (in game) cases which show a noticable difference in IQ.

 

[Dave Baumann]

BTW, since this is a pattern/synthetic test and not from an actual game, how do we know for sure this is not one of those "special cases" where the ATi optimization is disabled to get the "correct output?"

Ummm, it does - thats kinda half of the point of the test app!

 

[Ruined]

BTW, since this is a pattern/synthetic test and not from an actual game, how do we know for sure this is not one of those "special cases" where the ATi optimization is disabled to get the "correct output?"

Ummm, it does - thats kinda half of the point of the test app!

Heh, well it will be interesting to see the results. Thanx for putting up all the A/Bs. I hope you enjoyed my fantastic circling job

 

[Rugor]

Thanks Dave.

It looks like ATI is to be commended.

 

[SsP45]

ATI's method does seem to work quite well. I can't tell the difference, and yet it gives a much smaller hit with AF than any other card.
________
Vaporizer Guide

 

[Ostsol]

Anyone else think that a similar test and poll should be done for NVidia's trilinear optimizations?

 

[Bry]

Thanks Dave for all thew work your doing to get to the bottom of this. You should have enough info for a fine article when this is all done. And I look forward to your conclusions

As for that pic with the red circles. The R420 pic is definately shartper in those areas. The NV17 one is more blurry.

 

[karlotta]

Anyone else think that a similar test and poll should be done for NVidia's trilinear optimizations?

good god NO, let the hacks a Xbits do it first, then we can fix it again

 

[Sphinx]

The Problem is wich some Users have you can not EAT it you can not Smell it and hardly not see or noticing it. It is like the TemporalAA wich pretend your eyes a better AA - How will Reviewers Bench it or compare it with the competition...

 

[digitalwanderer]

How will Reviewers Bench it or compare it with the competition...

Subjectively, the way it probably should be done now.

It looks like it's about time for one of them pardigm shift-thingies to hit the graphic review world. 8)

 

[Ruined]

How will Reviewers Bench it or compare it with the competition...

Subjectively, the way it probably should be done now.

It looks like it's about time for one of them pardigm shift-thingies to hit the graphic review world. 8)

Nice to see I won't have to argue with you any more on that issue, dig

I guess I was "ahead of my time" last year

 

[digitalwanderer]

Nice to see I won't have to argue with you any more on that issue, dig

I guess I was "ahead of my time" last year

Last year I could see a difference without having to look for it, this year I can't tell a visual difference to save me arse.

Last year you were still wrong Ruined.

 

[Ruined]

Last year I could see a difference without having to look for it, this year I can't tell a visual difference to save me arse.

Last year you were still wrong Ruined.

But if you are saying things are now subjective, then I can't be wrong, can I? If you can't see it, I may be able to, or vice versa. If there is a paradigm shift and "apples to apples" and "mathematically equivalent" are no longer the words of the day, then what I said last year about whether or not you can actually personally notice the optimization in game or not is very much in accordance with what you are promoting now

 

[blackfish]

I don't see any obvious mipmap transition here.
Both pictures looks just fine.

So the ss discussed here

http://www.beyond3d.com/forum/viewtopic.php?t=12578

was concluded as misgrabed image?

 

[Ruined]

I don't see any obvious mipmap transition here.
Both pictures looks just fine.

So the ss discussed here

http://www.beyond3d.com/forum/viewtopic.php?t=12578

was concluded as misgrabed image?

Well I don't think anyone has heard back from xbit, but people seem to be assuming the x800 grab was bilinear. Ironically, several people here and from nvnews stated they liked what we can assume to be the straight bilinear shot better than the brilinear and trilinear shots. So much for filtering

 

[crushinator]

But if you are saying things are now subjective, then I can't be wrong, can I? If you can't see it, I may be able to, or vice versa. If there is a paradigm shift and "apples to apples" and "mathematically equivalent" are no longer the words of the day, then what I said last year about whether or not you can actually personally notice the optimization in game or not is very much in accordance with what you are promoting now

I agree with both of your opinions. One of the hands says "If it's not noticable in games or in a realworld apps, it should be legit" and the other hand says "I personally notice that in X Y and Z" what IMHO means, that such optimisations should be adjustable by users themselves. It's all about subjective feeling/seeing what ever, isn't it?

 

[mozmo]

funny enough i couldn't find many in game comparison shots of nvidia's trilinear optimisations on a lot of sites, but i did find some from the hothardware review. http://www.hothardware.com/viewarticle.cfm?articleid=517&page=7

Here they are.

nv40 no AF

nv40 16xAF trilinear ops off

nv40 16xAF trilinear ops on

R420 no AF

R420 16 AF

slight gamma differences aside between nv and ati, both of their optimised trilinear methods are to my eyes very good, ie i can't see any obvious mip transistions. True that nvidias method in colour mips falls down in quality and ATI's goes into full trilinear which we know isn't the case in game, so i guess in game both approaches are damn close in quality and probably should be benched and compared too. ie if ATI enable full trilinear, then that can be benched against nvidias full trilinear and then they can go head to head on their trilinear optimisations. I think that's fair.

 

[Althornin]

TO ALL THOSE WHO MADE NICE PICTURES WITH RED CIRCLES:

YOU ARE BLIND.

Load up the crompresonator and actually COMPARE the images.
your circled areas are most assuredly not relevant. You are decieving yourselves.