Alternative AA methods and their comparison with traditional MSAA*

Becouse it affects all polygons no mather size, can be applied to transparencies and generally the lighting strenght/method or any other effect wont cancel it out or mess with it. End result is one uniform smooth image with no edges/surfaces lacking AA. In games with long draw distances and fine grain distant detail that isn't made out of bitmaps or such it would create even bigger difference.
 
Not sure what the GoW3 devs did differently but they said their technique was slightly different to MLAA (like they further refined the technique). Maybe they did something about the sub pixel problem? Anybody with the game confirm if it's still a problem?
 
Would it be too much to ask why?

All the post filters work after the scene detail has been lost through the rendering process. As others indicated, it is particularly important on highly aliasing elements like small or distant stuff.

These new approaches do their best to reconstruct that information, but can't get as good even with practically unlimited resources...

However, it seems reasonable to me to combine the two approaches, which might be what GOW3 does.
 
Games that lack one or more of the following compared to other games in their genre:
- Polygon detail (too few polygons)
- Pixel shader detail (not as many fancy effects)
- Framerate

Aiight, I will wait for the release build of GoW3 to see the final images. Given the resource constraints, their MLAA implementation may be the best choice for similar games.

For GT5, their AA doesn't seem sufficient even in replay (2x or 4xAA ?). I don't need the distant AA. If there is some way to filter the cars only, I'd be happy. I don't pay attention to background objects, especially far away ones.
 
All the post filters work after the scene detail has been lost through the rendering process. As others indicated, it is particularly important on highly aliasing elements like small or distant stuff.

These new approaches do their best to reconstruct that information, but can't get as good even with practically unlimited resources...

However, it seems reasonable to me to combine the two approaches, which might be what GOW3 does.

I don't think GoW3 does both MLAA and 2xAA.
What I understand from Christer is that they went beyond the proposed MLAA described in the original paper and solved the sub-pixel problem.
 
It could have been worse, patsu. It could have been worse. :p (/alstrong awards no points today, and may Kratos have mercy on souls...)
 
What I understand from Christer is that they went beyond the proposed MLAA described in the original paper and solved the sub-pixel problem.

...

He did not say he solved the sub-pixel problem but he said he went beyond the conventional MLAA. So I guess he has solved it.

Well, no, all you are saying is they used a modified version of MLAA. We don't know anything more.
This sort of speculation, and statement of fact based on speculation is where misinformation starts.

Unless / until we have clear comment from the developers themselves specifically stating they have solved 'the sub pixel problem' then all we can say is they use a modified MLAA implementation.

I saw that, Mr. Graham ! :devilish:

You saw nothing.






:mrgreen:
 
For GT5, their AA doesn't seem sufficient even in replay (2x or 4xAA ?). I don't need the distant AA. If there is some way to filter the cars only, I'd be happy. .

That is shader aliasing. Dunno how much can be done via sahders to fix or the brutal method of SSAA to fix it.
 
Back
Top