Alternative AA methods and their comparison with traditional MSAA*

[djskribbles]

well 4xMSAA looks way much better than GoW MLAA.
i think 2x looks better too. much more consistent, imho.

Ok, that's just wrong.

 

[assurdum]

well 4xMSAA looks way much better than GoW MLAA.
i think 2x looks better too. much more consistent, imho.

wat Based on what? The MLAA has similar edge of 16xAA, sure has its limits but overall is a lot better of simple 2 or 4x MSAA...

 

[Nesh]

wat Based on what? The MLAA has similar edge of 16xAA,

It does???

 

[Alucardx23]

It does???

I think the 16XAA comment came from the Digital Foundry article on the use of MLAA on Saboteur, but it was later found that Saboteur does not use MLAA, but edge-filter/blur, I haven’t heard the equivalent AA that MLAA offers.

"The PS3 rendition of Pandemic's The Saboteur is different though. It's special. It's trying something new that's never been seen before on console, or indeed PC, and its results are terrific. In a best-case scenario you get edge-smoothing that is beyond the effect of 16x multi-sampling anti-aliasing, effectively delivering an effect better than the capabilities of high-end GPUs without crippling performance."

http://www.eurogamer.net/articles/digitalfoundry-saboteur-aa-blog-entry

 

[zaftxx]

You are crazy! Did you only play the GoW demo or something? Because that's 2xMSAA, and the difference between that and the final code's MLAA is huge.

MLAA is really in another league, I think. I'd have to see it in more games to be sure, but in God of War at least it looks phenomenal and the quality is just fantastic. Didn't you see the Digital Foundry feature?

gow 3's art style hides jaggies very well. 4xAA is better then mlaa any day of the week.

 

[Arwin]

gow 3's art style hides jaggies very well. 4xAA is better then mlaa any day of the week.

Your opinion seems to go against general perception, so maybe you can show us?

 

[patsu]

I think the 16XAA comment came from the Digital Foundry article on the use of MLAA on Saboteur, but it was later found that Saboteur does not use MLAA, but edge-filter/blur, I haven’t heard the equivalent AA that MLAA offers.

"The PS3 rendition of Pandemic's The Saboteur is different though. It's special. It's trying something new that's never been seen before on console, or indeed PC, and its results are terrific. In a best-case scenario you get edge-smoothing that is beyond the effect of 16x multi-sampling anti-aliasing, effectively delivering an effect better than the capabilities of high-end GPUs without crippling performance."

http://www.eurogamer.net/articles/digitalfoundry-saboteur-aa-blog-entry

As I understand, Saboteur's AA uses luminance to detect edges, causing artifacts in some cases (e.g., red bordering brown). MLAA uses a more sophisticated edge detection technique and should give more consistent result. Should play GoW3 to appreciate its smoothness. I only hope some day, Gran Turismo can use it.

 

[TheWretched]

I hope so too (GT5 using GoWAA). At least if it does look better than 4xAA at 720P or 2xAA at 1080P upscaled.

GT5 however is mostly "ugly" through the lower res shadows (which have been getting constant upgrades since GT5P, though) and shader aliasing... the latter could look better with GoWAA, no?

 

[nightshade]

MLAA doesn't helps with shader aliasing, you need either super sampling or you need to do something with the shader itself if you want to eliminate shader aliasing.

gow 3's art style hides jaggies very well. 4xAA is better then mlaa any day of the week.

I don't think anyone would agree with that, so can you tell us why you think so ?

 

[N_B]

The Saboteur uses luminance to detect edges...but how does it smooth them? It's not blurring. It's not MLAA. What is it?

 

[Shifty Geezer]

gow 3's art style hides jaggies very well. 4xAA is better then mlaa any day of the week.

That's plainly not true, at least not in terms of long-edge jaggie reduction (the technical reasons for choosing one form over another are very platform-centric). 4xMSAA gives 4 discrete intensity steps along a near horizontal edge, best case eing a good averaing but with limited sampling patterns these averages can be non-ideal. Whereas what's been shown of MLAA so far generates an almost line-drawn algorithm quality AA with smoothly progressive intensity. Look at the very-nearly horizontal wall in the background - you couldn't tell it's not horizontal without zooming in because the AA hides the vertical displacement so well...
http://images.gamersyde.com/image_god_of_war_3-12091-1708_0009.jpg
A case may be made for choosing 4xMSAA over MLAA in games that produce lots of artefacts with current MLAA implementations, but "any day of the week" seems a poorly considered generalisation to me, while 'GOW 3's art style hides jaggies very well" seems to be ignoring the high contrast moments that shining weapons against dark backgrounds produce, all cleanly AA'd. Plus as a post effect working on intensities, MLAA should work on HDR in fancy colour spaces. MLAA applied to HS should have produced better looking results with less AA failures that resulted in zero AA in the shipped game.

 

[Kasersky]

i think one case where 4xmsaa will destroy mlaa is for far objects especially fences and wires (any very thin items) because the sampling will help from things from just simply disappearing into the distance but you can sorta make art assets in a way that favors mlaa like those fences in that pic you just posted.

 

[Shifty Geezer]

i think one case where 4xmsaa will destroy mlaa is for far objects especially fences and wires (any very thin items)

Absolutely. However, if we're going with smart rather than brute-force AA methods, IMO we should also shift rendering of distant objects away from the trinagle rasteriser. Anything less than a pixel in size can be rendered as a dot or thin-line using other drawing methods better suited to the task.

 

[Kasersky]

Absolutely. However, if we're going with smart rather than brute-force AA methods, IMO we should also shift rendering of distant objects away from the trinagle rasteriser. Anything less than a pixel in size can be rendered as a dot or thin-line using other drawing methods better suited to the task.

is something like that complicated to implement

 

[Neb]

i think one case where 4xmsaa will destroy mlaa is for far objects especially fences and wires (any very thin items) because the sampling will help from things from just simply disappearing into the distance but you can sorta make art assets in a way that favors mlaa like those fences in that pic you just posted.

I wonder what would happen with games that use A2C where MSAA "smooths" out alpha. Also MSAA+TSAA will in most games result in the best option if possible becouse you get AA alphas and geometry.

I'll also wonder where ATI AA method 'edge detect' which tripples AA amount would place itself. 4xMSAA + edge detect is "12xAA", 8xMSAA + edge detect is "24xAA".

 

[Kasersky]

I wonder what would happen with games that use A2C where MSAA "smooths" out alpha.

i thought thats why they did 4xmsaa for alan wake

 

[N_B]

The demo of GOW3 had some A2C, but in the final game those areas look significantly better with MLAA (the hair):

http://images.eurogamer.net/assets/articles//a/1/0/0/1/8/6/1/Helios_Demo.jpg.jpg
http://images.eurogamer.net/assets/articles//a/1/0/0/1/8/6/1/Helios_Final.jpg.jpg

The MLAA could be doing wonders on the hair, but they could might just changed the way it's rendered..

 

[Neb]

i thought thats why they did 4xmsaa for alan wake

Yes.

The demo of GOW3 had some A2C, but in the final game those areas look significantly better with MLAA (the hair):

http://images.eurogamer.net/assets/articles//a/1/0/0/1/8/6/1/Helios_Demo.jpg.jpg
http://images.eurogamer.net/assets/articles//a/1/0/0/1/8/6/1/Helios_Final.jpg.jpg

The MLAA could be doing wonders on the hair, but they could might just changed the way it's rendered..

Maybe they skipped A2C as the final version but still seems to have aliased hair. But the thing I was wondering is since textures with alpha have solid texture surface but with alpha to decide which areas are transparent then, Would it even be possible to detect which areas are non transparent/detail pattern within texture?

 

[assurdum]

I think the 16XAA comment came from the Digital Foundry article on the use of MLAA on Saboteur, but it was later found that Saboteur does not use MLAA, but edge-filter/blur, I haven’t heard the equivalent AA that MLAA offers.

"The PS3 rendition of Pandemic's The Saboteur is different though. It's special. It's trying something new that's never been seen before on console, or indeed PC, and its results are terrific. In a best-case scenario you get edge-smoothing that is beyond the effect of 16x multi-sampling anti-aliasing, effectively delivering an effect better than the capabilities of high-end GPUs without crippling performance."

http://www.eurogamer.net/articles/digitalfoundry-saboteur-aa-blog-entry

It's not based only on this statement; the step by step analysis with MLAA indicated similar edge of 16xAA resulted although is a total different AA method with different limits of course. The saboteur technic from what I have understood is 'similar' to MLAA to detected aliasing edge but indeed use morphologic method use a simple blur filter (at least based on the explanation of SMS reading in DF)

 

[Alucardx23]

It's not based only on this statement; the step by step analysis with MLAA indicated similar edge of 16xAA resulted although is a total different AA method with different limits of course. The saboteur technic from what I have understood is 'similar' to MLAA to detected aliasing edge but indeed use morphologic method use a simple blur filter (at least based on the explanation of SMS reading in DF)

From where else are you taking the 16XAA quote from? I understand that MLAA must offer a similar or better technique that the one used on saboteur when it comes to remove aliasing, but I have not seen a comparison to normal MSAA as in Saboteur.