AA names help

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

darckhart

Senior member
Jul 6, 2004
517
2
81
you know, in addition to this definition guide it would be good to mark where you can actually set each one in the respective driver software panels. (i just switched from ati to nvidia and i can't find anywhere that lets me choose the AA type like MSAA, CSAA, SGSSA etc)
 

taltamir

Lifer
Mar 21, 2004
13,576
6
76
you know, in addition to this definition guide it would be good to mark where you can actually set each one in the respective driver software panels. (i just switched from ati to nvidia and i can't find anywhere that lets me choose the AA type like MSAA, CSAA, SGSSA etc)

done, I converted the third post in this thread into a guide for selecting your AA type.

Edit: This does re-enforce your point about AA terms being confusing though.

Heh, it really is terrible. For example, I just found out that when nvidia first introduced the GeForce 3 (their first card to have MSAA), they invented THREE separate names for it (MSAA, HRAA, and Quincunx AA) and used all 3 interchangeably in that very same article!
 
Last edited:

taltamir

Lifer
Mar 21, 2004
13,576
6
76
Added SRAA info. If anyone has links to tests on this new AA form please share them.
 

SirPauly

Diamond Member
Apr 28, 2009
5,187
1
0
One of the most impressive abilities I had the pleasure to game with was on the 1900XTX CrossFire Platform and their Super-AA feature. The key to this feature was it offered a combined x8 sparse grid multi-sampling with a x2 rotated grid Full-scene SSAA, and the kicker was, the x2 full scene FSAA was virtually free performance and helped curb texture aliasing a lot --from aftifacts from shader, transparency alpha test and blend, specular and ordinary alpha blends, mip-map transitions, moire, and when ATI used modest optimizations. So, if one GPU offered 60FPS with x4 AA, two GPU's would offer x8 AA with x2 SSAA and would bench 60 FPS.

Virtually no one talked about it.

And another kicker was, the incredible flexibility with transparency, so if I had 60 fps with x4 MSAA with x2 adaptive on transparency with 1 GPU would have 60 FPS with x8 MSAA, x2 SSAA and x4 adaptive on transparency with 6 distinct samples total on transparency due to the x2 SSAA, too.

Virtually no one talked about.

All many talked about was performance, attacking IHV's, how this one is biased and that one is biased, personal shots, how one product rocked because it wins a bench by 10 percent.
 
Last edited:

thilanliyan

Lifer
Jun 21, 2005
11,823
2,021
126
Thank you for this guide. I think it will be useful for a lot of people.

By the way, what is the "Edge detect" feature on ATI cards?
 

taltamir

Lifer
Mar 21, 2004
13,576
6
76
Thank you for this guide. I think it will be useful for a lot of people.

By the way, what is the "Edge detect" feature on ATI cards?

I am guessing it detects the "edges" of objects (which, is kinda the definition of AA; find jugged edges and smooth them). I haven't found any specifics explaining it
Its the "tier" part that is confusing me, what does it mean?
Note that "tier edge detect" is in the "currently eluding definition" section.

There was Matrox and their FAA -- Fragment AA, too.

thanks, I will search for info on it
 
Last edited:

darckhart

Senior member
Jul 6, 2004
517
2
81
hm why havent we stickied this yet? i had to get creative with search terms to find this thread again.

also, in your third post you can add to the nv control panel AA options:
32x, 32xQ - dont know what the breakdown is ie 8xMSAA + 24xCSAA
and the SLI modes which has the same but splits the load evenly among the cards
8,16,32 and their Q variants
 

taltamir

Lifer
Mar 21, 2004
13,576
6
76
hm why havent we stickied this yet? i had to get creative with search terms to find this thread again.

also, in your third post you can add to the nv control panel AA options:
32x, 32xQ - dont know what the breakdown is ie 8xMSAA + 24xCSAA
and the SLI modes which has the same but splits the load evenly among the cards
8,16,32 and their Q variants

I added them, but they are not showing up in my own nvidia control panel... if you don't mind can you tell me what card you have? maybe they are not available for my card (GTX260)
 

Idontcare

Elite Member
Oct 10, 1999
21,118
58
91
hm why havent we stickied this yet?

Because no one pm'ed me with a request to do so.

If the OP is willing to maintain the OP on a regular basis so it doesn't go stale then I would be willing to sticky it.

Are you game Talta?

Idontcare
Super Mod
 

Cythreill

Member
Apr 6, 2011
31
0
0
It depends on the scene content, but SSAA is almost always better than MSAA.

If I told you I'd like to know in regards to Oblivion, would that help?

And are you saying x2 SSAA is better than x8/x16 MSAA in almost all cases, pending the scene content?
 

BFG10K

Lifer
Aug 14, 2000
22,487
2,424
126
If I told you I'd like to know in regards to Oblivion, would that help?
No because I haven't really played it. Why don't you try it for yourself?

And are you saying x2 SSAA is better than x8/x16 MSAA in almost all cases, pending the scene content?
If the game has a lot of shader/texture aliasing (e.g. Riddick, Crysis) then yes. If not then a higher level of MSAA will be better. Again, it depends on the scene content.

But 4xSSAA is almost always better than 4xMSAA, regardless of the scene.
 

darckhart

Senior member
Jul 6, 2004
517
2
81
I added them, but they are not showing up in my own nvidia control panel... if you don't mind can you tell me what card you have? maybe they are not available for my card (GTX260)

see sig. GTX580 SLI

@Cythreill:
In addition to the jaggies that bother me, the next would be shimmering, then loss of detail.

If the level is set incorrectly based on the scene, then SSAA can sometimes cause detail loss.

In most UE3-based games I play, I feel SSAA wins hands down. (But that's because I can afford to run it at ridiculous levels. Unfortunately, 4xSSAA and higher is VERY taxing on the card.)
 

taltamir

Lifer
Mar 21, 2004
13,576
6
76
Because no one pm'ed me with a request to do so.

If the OP is willing to maintain the OP on a regular basis so it doesn't go stale then I would be willing to sticky it.

Are you game Talta?

Idontcare
Super Mod

I would be glad to.

see sig. GTX580 SLI

thank you. I am guessing the non SLI 32x modes come from having a fermi based card. I will look into it to see if I can find if GTX4xx series has that and what the exact figures are.

Also, do you mind posting a screenshot of the panel like I did in post 3? that way we can reference it for SLI users. thank you
 
Last edited:

Idontcare

Elite Member
Oct 10, 1999
21,118
58
91
Making this a sticky, the OP has agreed to maintain it and keep it from going stale. Thanks Taltamir :thumbsup:

Idontcare
Super Mod
 

LiuKangBakinPie

Diamond Member
Jan 31, 2011
3,910
0
0
This jaggedness in 3D graphics isoften referred to by gamers simply as 'jaggies', but the effect is formally known as Aliasing . The reason aliasing occurs is because the image on your screen is only a pixellated sample of the original 3D information your graphics card has calculated. At increasingly higher resolutions, as more pixels are used, that sample comes closer to the original source and hence the image is clearer and displays less aliasing. However aside from the fact that higher resolutions can degrade performance, many monitors simply cannotdisplay the very high resolutions needed to effectively remove the problem.
This is where a technique originally called Full Scene Anti-Aliasing (FSAA), and now simply Antialiasing (AA), can be used by your graphics card to make the jagged lines in computer graphics appear smoother without having to increase your resolution. It essentially resamples and blends jagged lines with their surroundings, and can be applied to either 2D or 3D graphics.

To see the different levels of AA use this
http://www.nhancer.com/help/AASamplesLow.htm

Here's the rest of the above explanation
http://www.tweakguides.com/Graphics_11.html
 

Aristotelian

Golden Member
Jan 30, 2010
1,246
11
76
In the above link, 4x4 SSAA looks the best to me. This is also the setting I use in games (forced override application settings in NV control panel).
 

sticks435

Senior member
Jun 30, 2008
757
0
0
So if I understand this right, if you have a G80 or Fermi card, there are now 3 different names for MSAA 8X?

Edit: Well, looks like 16X didn't change with G80 or GT200, as seen here.

http://www.nvnews.net/vbulletin/showthread.php?t=79858

Edit 2: After some pretty hard digging, it seems Nvidia might have changed the way they counted Multi/color and coverage samples from G80/Gt200 to Fermi. From what I have read here : http://alienbabeltech.com/abt/viewtopic.php?f=4&t=15337 and a couple of other places http://www.anandtech.com/show/2918/5 before Fermi, they counted multi-samples twice in their papers, once as multi-samples and once as coverage samples, so they list 8x as having eight coverage samples and 16x/16xQ as having sixteen. Which is the number everyone quoted back then, such as the Nhancer doc's. If you read any reviews of Fermi, or in fact read the AA section of the GF100 white paper, they refer to 16XQ as 8 color and 8 coverage samples.

So it seems they only counted the coverage samples once when explaining 32X with Fermi, and retroactively applied this to G80/GT200 was well. Hence why I believe the Guru3d article is inaccurate. If you were comparing a fancy new card with a fancy new high level of AA, wouldn't you want to compare the highest from last gen to highest of new gen? I think they meant to quote the Nvidia numbers for 16XQ, but messed up and only put 16X.

Edit 3: I was thinking about this more while trying to sleep, and I think I have a better explanation. I don't think they counted the samples twice to get the final number for the name, they just added the MSAA and coverage samples together. According to the ABT article I linked to, 8X was really 4+4, 16X was 4+12, 16XQ was 8+8 and now 32X is 8+24. It looks like anything higher than that is SLI only as I have a 570 and it only goes to 32x.
 
Last edited:

sticks435

Senior member
Jun 30, 2008
757
0
0
Also, I think the transparency section should be updated. TrAA doesn't necessarily mean only Super Sampling AA. Technically there are 2 modes with Nvidia cards. TrSS and TrMS. The problem was that on G80 and GT200, you could only use MSAA to test the transparent area's, not CSAA. So the TrMS mode sucked, because you only had a max of 8 samples to use, so everyone used TrSS, which then became de facto TrAA. With Fermi, you can now use CSAA for the transparent area's, with means that the TrMS mode is now a viable option.

This method of AA will super sample the transparent portions of a texture to help decide where to smooth the opaque portions, while leaving the rest of the picture for the much quicker MSAA technique.

I don't think this is quite right. TrSS would sample the whole texture where the opaque section appears, while TrMS would only sample the opaque portion itself, correct?

Also, I believe that there is a new mode called [FONT=verdana,geneva]Accelerated Jittered Sampling, which is only on Fermi and only used on shadows. [/FONT]
 

taltamir

Lifer
Mar 21, 2004
13,576
6
76
thank you, I will look into it in a couple of days (i got a bunch of tests and projects right now)
 

sticks435

Senior member
Jun 30, 2008
757
0
0
No problem. Also, I think 32xQ is probably regular 32X, but split between SLI cards for the same quality as faster performance. I think there is a 64X also, which is probably 32X on each card.