The Best AA

[John Reynolds]

One thing I noticed about the 9700's AA is that it does not do a great job on high contrast edges.

That is exactly what I've noticed, too.

 

[OpenGL guy]

One thing I noticed about the 9700's AA is that it does not do a great job on high contrast edges.

That's funny, I thought the opposite. Look at the screenshots comparing the 9700's gamma corrected AA to the GeForce 4's non-gamma corrected AA.
http://www.beyond3d.com/forum/viewtopic.php?t=2454

Also, since you don't have a comparison shot, how can you blame the 9700 for the results you gave? Maybe there's something else here that you are missing.

 

[Lincoln Bauman]

One thing I noticed about the 9700's AA is that it does not do a great job on high contrast edges.

That's funny, I thought the opposite. Look at the screenshots comparing the 9700's gamma corrected AA to the GeForce 4's non-gamma corrected AA.
http://www.beyond3d.com/forum/viewtopic.php?t=2454

Also, since you don't have a comparison shot, how can you blame the 9700 for the results you gave? Maybe there's something else here that you are missing.

OpenGL guy,

I didn't intend this as a slam against the 9700, overall I think the AA looks great.
It's just an observation I made after I installed the 9700. I always use 4xS mode when running Morrowind with my Ti-4600, and I immediately noticed that even with 6xAA enabled the GF4 does a better job in the situation I illustrated earlier (on near horizontal edges only).
I'll try to provide a comparison for you.

Lincoln

 

[SirPauly]

This is LLB from Rage, right?

 

[OpenGL guy]

I always use 4xS mode when running Morrowind with my Ti-4600,

4xS != MSAA

Please try 4x MSAA on the GeForce 4.

 

[jb]

V5's AA isn't bulletproof, especially at the resolutions you are forced to run it at, you can still notice some shimmer or flicker on edges. The V5 does not have "perfect" edge-AA, no way, no how.

What I ment by that is the V5 AA worked all the time and all the cases. Todays other methods dont work all the time.

Moreover, I'm not really concerned with edge-AA IQ on super old 16-bit games when every thing else is a joke (color artifacts, very low res textures, blurred detail everywhere)

Just because you dont play them does not overcome the fact that it is a short coming of the other methods. Besides I am sure we can fine plenty of DAoC fans that would disagree with you

I can't think of any technical reason for how the V5's AA could possibly be better than what the R300 is capable of.

I can. There are cases were the R300 as good as it is does not work. Thus V5 method was a bit more robust.

The alpha texture issue is somewhat of a red herring. Modern games are replacing alpha texture tricks with real geometry ( the fence or railing tricks) and atleast for me, the popular counterstrike maps that use it enough to notice are few (cs_italy)

Actually there are only a handful of CS maps that dont have any alpha's in them (dust, dust2) as the majority of other maps do.

Basically, the R300 has solved the aliasing problem for me, especially at high resolution. Maybe if you are running below 1280x1024 you'll notice it, I sure don't.

Funny thing it has for me too. Yes I still see one or two issues with it, but the other IQ ehancemetns make up for the minor shrot commings. I am very happy with it.

 

[Lincoln Bauman]

I always use 4xS mode when running Morrowind with my Ti-4600,

4xS != MSAA

Please try 4x MSAA on the GeForce 4.

The topic of this thread is edge AA quality, and in fact Nvidia's 4xS AA mode is listed in the poll.
I am only posting my personal observations based on what I have seen with my own eyes with both cards.
Here is an example from the Ti-4600, it's impossible to reproduce the exact same scene but I tried as best I could.
Although there are only three visible color gradients compared with the 9700's five, the three are more evenly balanced between colors.
This is again only on the near horizontal edges, the 9700 looks better on the near vertical edges.
Since you work for ATI, could the AA sampling algorithm be adjusted to better handle the example I have provided?
I also want to reiterate that I am not bashing the 9700, only nitpicking about the antialiasing, overall it looks great!

Lincoln

 

[OpenGL guy]

The topic of this thread is edge AA quality, and in fact Nvidia's 4xS AA mode is listed in the poll.

That's great, but I was interested in the MSAA quality.

 

[Humus]

One thing I noticed about the 9700's AA is that it does not do a great job on high contrast edges.
This screenshot is with 6xAA, the bottom edge is the most noteworthy.
Looking at where the gradient starts, it goes from almost black to light gray.
The following gradients are very close in color, reducing the effectiveness of the antialiasing.
If it were represented with numbers, it looks more like "10 5 4 3 2" rather than "10 8 6 4 2".
In my opinion the "smoothness" of the gradient could be improved.

Lincoln

I suppose that's the effect of the gamma correct AA.

 

[martrox]

Guys, the real problem here is that there is no ONE best AA for everything. I too still believe that V5 FSAA was the best at almost everything, BUT I could no longer live with the other things that come with a 2 year old unsupported card. (slooow, no AF....etc.) However, while going through Geforce after Geforce, I NEVER felt I could live with the FSAA they gave, even 4Xs! Now that I have a Radeon 9700, I feel good about FSAA again. While not perfect, it does work better than anything else in the consumer sector. Sure, you pick at anything and find faults, BUT, overall, at this point, no videocard currently available can match it's speed and IQ with all the eyecandy turned on, period.

 

[OpenGL guy]

[quote="HumusI suppose that's the effect of the gamma correct AA.[/quote]
I don't think so, or else we would have seen the same thing in the images from the other discussion: http://www.beyond3d.com/forum/viewtopic.php?t=2454

I'm trying to get an idea as to what is up (like look at the GeForce 4's 4x MSAA).

 

[Humus]

We do see the same effect there. Cut'n'pasted from the 4xAA images, top is R9700, bottom is GF4. The R9700 goes to brighter pixels sooner, but it's no doubt that it does look better when not zooming in like this. Zoom pictures is a quite useless way to compare AA IMO. It's normal behaviour for a gamma correct AA to go bright earlier than non-gamma correct.

 

[Bigus Dickus]

Is anyone else wondering why the 9700 has a lot of scattered pixels of the darkest intermediate shading? They're on either side of the line, and hard to see...


p.s., I have a nice software line-AA shot for comparison if someone would like to host it... up to ten intermediate shades AFAICT.

 

[Bigus Dickus]

lol, no, I decreased the line angle closer to horizontal/vertical, and there are over 60 intermediate shades.

Someday hardware will be that robust...

 

[Bambers]

The 9700 has the 'scattered pixels' because its was taken from a jpeg.

Couldn't a possible reason for some people finding non gamma corrected FSAA smoother be that they use different gamma settings in game?

 

[LittlePenny]

You can send the picture my way if you like. (Assuming the file is under 2mb use my email link below)

 

[Bigus Dickus]

The 9700 has the 'scattered pixels' because its was taken from a jpeg.

I had thought of that, but wasn't sure if the original was in .png or just Humus' compilation... I guess I could have checked.

bah, jpeg sucks if it artifacts a simple b&w line in compressing it.

 

[Bigus Dickus]

Thanks to LittlePenny for the hosting.

Some software line AA shots for comparison, reference, or just plain admiration:

 

[antlers]

Yes, difficult to tell what's what when dealing with images zoomed from .jpgs[/png]

Synthetic images for image quality comparison should use .png's, please.

Even Windows Paint will let you save as .png, (provided you've opened a .png file once).

 

[Bigus Dickus]

For reference, my images were zoomed as bitmaps first, then saved as jpg's in the zoomed condition.