What are the different types of Anti-Aliasing?

[jasper1605]

I've been trying to read on wikipedia what the differences are between anti-aliasing and such like MSAA QSAA or other kinds like what I see in the options on Left For Dead 2. Unfortunately, it gives me a history and calculus lesson before it tells me the answer.

About all I know is that AA is for softening images to reduce jaginess, but that's as far as my comprehension goes; can anyone explain further what it means?

Also, sorry if this is in the wrong section, I wasn't sure where a good place to put this (or even search for it) would be.

 

[Munki]

Im curious as well. I haven't done any research other than to figure out why COD 2 and mu 9800 GT were having a civil war givin' me a gray screen.

 

[jasper1605]

seems like civil wars like to happen often within computers. That tweak itch that we all get just forces us to change things even if we know it's working perfectly just to see if it can push a little bit farther, lol.

 

[pantherx12]

Wikipedia explains it great

http://en.wikipedia.org/wiki/Anti-aliasing

FSAA ( full screen anti aliasing) is what is done over 3d games so scroll to that in particular.

 

[Nailezs]

thanks panther, very intersting

 

[jasper1605]

but what's the difference between things like msaa qsaa etc... ???

 

[Techtu]

but what's the difference between things like msaa qsaa etc... ???

I believe each has it's own method in which way it codes the AA (anti Aliasing).

 

[pantherx12]

Had no idea, so I googled.

This forum post makes sense to me

"Let's put it another way. 16x QSAA would give you the same level of edge anti-aliasing as 16x MSAA, at the same cost as 8x MSAA. The trade-off is that textures will become blurrier, making 16x AF look like 4x AF.

MSAA is the quality option, QSAA is the speed option. In things like racing games, where you are moving too quickly for slight texture blurring to matter, but aliasing is super-apparent, then QSAA is a great choice. In something like Borderlands, where the unique texturing is a huge feature of the game, then QSAA is a crappy choice.

"

 

[Techtu]

Had no idea, so I googled.

This forum post makes sense to me

"Let's put it another way. 16x QSAA would give you the same level of edge anti-aliasing as 16x MSAA, at the same cost as 8x MSAA. The trade-off is that textures will become blurrier, making 16x AF look like 4x AF.

MSAA is the quality option, QSAA is the speed option. In things like racing games, where you are moving too quickly for slight texture blurring to matter, but aliasing is super-apparent, then QSAA is a great choice. In something like Borderlands, where the unique texturing is a huge feature of the game, then QSAA is a crappy choice.

"

I didn't know the actual difference, although I did have something kind of like that in my head, so thank's for the input

 

[douglatins]

My control classes are teaching me AA in common signals and fuctions and the application is the same in games, but rather a dot matrix

 

[Trigger911]

Had no idea, so I googled.

This forum post makes sense to me

"Let's put it another way. 16x QSAA would give you the same level of edge anti-aliasing as 16x MSAA, at the same cost as 8x MSAA. The trade-off is that textures will become blurrier, making 16x AF look like 4x AF.

MSAA is the quality option, QSAA is the speed option. In things like racing games, where you are moving too quickly for slight texture blurring to matter, but aliasing is super-apparent, then QSAA is a great choice. In something like Borderlands, where the unique texturing is a huge feature of the game, then QSAA is a crappy choice.

"

+1 I didnt know myself now I do

 

[TheLaughingMan]

I am not sure but I do know two things that may help you find more info.

MSAA = Multisampling AA
QSAA = Supersampling AA

The primary difference is the algorithm used. As stated, one (Super) is the quick and dirty AA which is why it is much faster and lighter on video card. Multisampling is the "measure twice, cut once" principle. The algorithm takes, as implied (several samples of the image) to ensure accurate location when the output is done. The result is cleaner and less jaggy, but at the cost of additional process time.

 

[jasper1605]

thanks much for that description panther!
Exactly what I was wondering

 

[pantherx12]

We can thank "darklordjames" of the AVS forums for the original post XD

 

[Polarman]

This could be helpfull too.

http://www.xbitlabs.com/articles/video/display/gf8800_14.html

 

[daehxxiD]

QSAA is not supersampling. I think it is Nvidias own technique called "Quincunx" available since the Geforce 3 (?). What this does is give you the double the edge-smoothing at an expense of texture-definition but not of speed. So actually what reads as 16x QSAA is very likely 8x Quincunx AA; like what in geforce 3 times was simply called 2x Quincunx-Antialiasing (which was basically in between 2xMSAA and 4xMSAA).

Supersampling on the other hand is a very high-quality Antialiasing method, that also anti-aliases alpha-blended textures (like fences and stuff).

Very simply put.

 

[qubit]

Because frame rate is important above all else for me, especially in a twitchy fps, I turn AA off and just use AF, usually at x16. I find that the jaggies are not all that objectionable to me most of the time and the frame rate is much better without AA.

I like a solid 85Hz refresh with no dropped frames on my CRT monitor (or 120Hz on Samsung LCD 2233RZ monitor) at a decent resolution and can usually achieve that on modern games on my hardware, which is fairly high end. Heck, I paid enough for it...

 

[daehxxiD]

Because frame rate is important above all else for me, especially in a twitchy fps, I turn AA off and just use AF, usually at x16. I find that the jaggies are not all that objectionable to me most of the time and the frame rate is much better without AA.

I like a solid 85Hz refresh with no dropped frames on my CRT monitor (or 120Hz on Samsung LCD 2233RZ monitor) at a decent resolution and can usually achieve that on modern games on my hardware, which is fairly high end. Heck, I paid enough for it...

Personally I think AA is one of the most Picture-Improving techinques possible. Games nowadays almost look like CG in some cases and it's only the Aliasing that give it away as Realtime. So enableing AA in some Cases really helps; a lot.

With Crysis for example, I also think that the boost in quality is not worth the Performance hit, but in other games AA doesn't cost all that much anymore and at >1Megapixel even 2xAA makes a world of diffrence.

 

[qubit]

Personally I think AA is one of the most Picture-Improving techinques possible...

Absolutely - AA can make a big difference. The type and specific game you play, your hardware and your preferences all come into it. The effect of using or not using it is very subjective to a large extent, so there's no right or wrong to this.

In a fast game like Unreal Tournament 2004, where every millisecond and every frame can mean the difference between fragging or being fragged, picture quality comes second to a rock solid, vsynced frame rate, but the compromise doesn't have to be very much with today's hardware. Also, you get much better lag performance and lack of motion smear on a CRT, so I keep a couple just for games.

 

[Mussels]

new thread title. This one actually asks the same question as your post, not a different one.


IIRC

MSAA blurs the colors between 'pixels', to make the edges softer.

SSAA (super sampling) actually renders the scene at a higher resolution then shrinks it.


Nvidias "Q" AA modes are mostly marketing crap, it tends to mean "we turned the performance cheats off in this mode, so it looks a little better"

 

[jasper1605]

thank you mussels

 

[pantherx12]

Mussels is wrong by the by MSAA is multisampling an optimised version of Super Sampling it AAs the edges of the polygon but won't really bother with anything else so you can get artifacts inside the polygon still.


"The term generally refers to a special case of supersampling. Initial implementations of full-scene anti-aliasing (FSAA) worked conceptually by simply rendering a scene at a higher resolution, and then downsampling to a lower-resolution output. Most modern GPUs are capable of this form of antialiasing, but it greatly taxes resources such as texture bandwidth and fillrate. (If a game is highly TCL-bound or CPU-bound, supersampling can be used without much performance hit.)

According to the OpenGL GL_ARB_multisample specification, "multisampling" refers to a specific optimization of supersampling. The specification dictates that the renderer evaluate the fragment program once per pixel, and only "truly" supersample the depth and stencil values. (This is not the same as supersampling, but by the OpenGL 1.5 specification, the definition had been updated to include fully supersampling implementations as well.)

In graphics literature in general, "multisampling" refers to any special case of supersampling where some components of the final image are not fully supersampled. The lists below refer specifically to the ARB_multisample definition."

 

[Mussels]

hey i said 'iirc' - if i recall correctly

 

[pantherx12]

Ah right, I didn't know that iirc meant that

 

[TheLaughingMan]

AA, turn that $#^( ON as high as your card can handle.