IQ Filtering Test: Which is your Preference

[bloodbob]

God the results have changed a bit their has been a reversal of the leaders ( the worest filter is still the worest) anyway I was want to know when the poll would close so we can talk about the results.

 

[ZoinKs!]

I, also, voted wrong. At first I thought they were pretty much the same but now notice some differences on closer inspection.

Does this poll have a hanging chad clause?

 

[bloodbob]

Relase the results already no one has voted for hours :/ and their are fools out their on other forums who say I dunno what A,B and C are and I wanna rip them apart . Alternately dave should submit the IQ test to /. ( I'm not gonna do it without his blessing he might appericate the server getting /.ed ).

 

[mjtdevries]

Actually, it might be interesting not to tell which card is which just yet, but to start discussing first what differences people saw in the images, and what vote they did based on that.

I have the impression that people might have seen the same differences, but then have a difference of opinions on whether that equals higher or lower IQ.

 

[bloodbob]

OOOh 2 more votes but yeah can we call this poll at its end already so we can atleast start discussing visual problems?

 

[Quitch]

I thought B looked the best and A the worst.

 

[Xmas]

I thought B looked the best and A the worst.

So did I. I decided mostly on brightness distribution. The A AF screenshot shows a /`-´\-shaped area that appears brighter than the rest (on my TFT with my gamma settings), which really can be disturbing.

The problem with this, however, is that when I set gamma to my "3D" setting, the three shots are much closer, and I'd give A a very slight edge, because the one moiré pattern is less visible.

 

[madshi]

Found B to be best looking. A and C were worse, not sure which of them was less worse than the other.

 

[mjtdevries]

I found a large difference between the AF and non-AF images.

In the non-AF the A shot looked worst, with B and C being equal.
In the AF shot A still had worst IQ, but now C looked sharper then B but without showin mipmap bounderies. The result of that was that C had a bit more moire. I think that in a game C would give the best result.

 

[ERK]

c_af=b_af>a_af>b>c>a
Am I blind?

 

[Althornin]

I found a large difference between the AF and non-AF images.

In the non-AF the A shot looked worst, with B and C being equal.
In the AF shot A still had worst IQ, but now C looked sharper then B but without showin mipmap bounderies. The result of that was that C had a bit more moire. I think that in a game C would give the best result.

wow, i must say i complete disagree.

In the AF shots, a>c>b, where a is sharper and has less moire.
in non af, b>a>c

c is the worst filter, period - worst aliasing, most obvious mip boundaries.

 

[Killer-Kris]

A = NV38
B = R360
C = R420

I'm not sure if everyone has seen this in the pick the R420 thread or not.

So I guess I'll go ahead and explain why I picked the way I did. For the non-AF shots they were all pretty close, but on the AF shots it was apparent to me that A had much less moire than the other two selections.

Now of course I do believe someone mentioned that moire is not seen in games and is actually an indication of sharper textures. If that's the case I wonder how many people voted like I did for the same reasons I did, and that maybe this test case is not really a very good indication of real texture quality.

Anyways just my 2cents.

 

[tEd]

b>c>a for both

 

[KimB]

Now of course I do believe someone mentioned that moire is not seen in games and is actually an indication of sharper textures.

Incorrect. Moire is not seen in actual game screenshots. It is, however, an indication of texture aliasing, and this will show up readily in games in motion.

Anyway, I noticed discolored moire in A's screenshot, and so threw it out. I suppose another red flag to all of us should have been that there is a marked difference in the way A looks in the near field vs. the far field (in the near field, the moire appears monochromatic. In the far field it appears discolored). That would be obvious in a game.

I also think that the additional moire in C's screenshot is an indication of extra texture aliasing. This means that when in motion, regular high-contrast textures will look worse with the R420's "brilinear" than they would with traditional trilinear.

 

[FUDie]

I also think that the additional moire in C's screenshot is an indication of extra texture aliasing. This means that when in motion, regular high-contrast textures will look worse with the R420's "brilinear" than they would with traditional trilinear.

Have you used the card? Have you seen the card in use? Didn't think so. Maybe, just maybe, you should reserve judgement until you've actually seen the board.

Comparing results on the 9600 with ixbt's test program shows very little difference between "optimized" and "standart"(sic) trilinear. I'd go so far as to say no visible difference. In motion, I don't see any extra aliasing.

Maybe you should stick to speaking from experience and not from your suppositions and bias.

-FUDie

 

[crushinator]

Dave,

how about telling us which one's whose, please?

 

[AlphaWolf]

Dave,

how about telling us which one's whose, please?

look back about 10 posts.

 

[KimB]

Have you used the card? Have you seen the card in use? Didn't think so. Maybe, just maybe, you should reserve judgement until you've actually seen the board.

Doesn't matter. The things that lead to aliasing are quite simple, and therefore should be independent of the texture used (What I'm trying to say is, if filtering algorithm X produces more aliasing in a specific situation with a specific texture, then it will produce more aliasing in that situation with any texture).

So, it's right there in black and white. The difference isn't huge, of course, but that just reduces the number of potential situations where the difference will be visible in-game. It doesn't mean that the difference won't be visible.

Comparing results on the 9600 with ixbt's test program shows very little difference between "optimized" and "standart"(sic) trilinear. I'd go so far as to say no visible difference. In motion, I don't see any extra aliasing.

Many people don't see texture aliasing anyway. I do, so it matters to me.

 

[bloodbob]

I gotta love how at the start of the POLL brilinear was winning I'll put up some interesting modification to these images later.

 

[gokickrocks]

i noticed the same...