2xAA vs 4xAA

[ArchAngel777]

Originally posted by: ElFenix

Originally posted by: rbV5
In virtually every case I've tried 2xAA is virtually a worthless setting. IMO, a higher resolution is always better than a lower resolution@ 2Xaa.

jaggies are quite noticeable at 1600x1200 ('edge crawling') but turn on 2x AA and it is much much better.


i seriously wonder if any of you have every bothered turning on 2xAA before.

I have used it many times and I didn't find it all that great in the resolutions I used. Since my current monitor is limited to 1280 X 768 (which is fine with me) I have only been able to test it as high as that resolution with my 7800 GTX. Quincunx was a joke in my opinion and always has been. Since I have never really gamed at 1600 X 1200 (due to refresh rate limitation @ 75Hz) I cannot say how 2X AA works... According to many people here, and I believe them, it helps a lot.

Basically, I just stick to 1280 X 768 and run 8XS AA and 16XAF with WoW and many other games, it looks great, runs perfectly smooth and looks pretty good.

I would say the general concensus is that 1280 X 960 and under, 2X has little improvement. But 1600 X 1200 and over, 2X has much improvement.

 

[xtknight]

Are we talking about multisampling or supersampling here? Rotated grid or ordered grid? Multisampling affects only vertex such as lines, whereas supersampling renders at a higher resolution and blends to a lower one so you get some of the effect of the high-res.

I'll post some BF2 2xAA and 4xAA screenshots. Unfortunately I won't be able to get them exact because IIRC I have to reload the map in BF2 for the change to take effect.

 

[ArchAngel777]

Originally posted by: xtknight
Are we talking about multisampling or supersampling here? Rotated grid or ordered grid? Multisampling affects only vertex such as lines, whereas supersampling renders at a higher resolution and blends to a lower one so you get some of the effect of the high-res.

I'll post some BF2 2xAA and 4xAA screenshots. Unfortunately I won't be able to get them exact because IIRC I have to reload the map in BF2 for the change to take effect.

I have noticed that screenshots are quite different then moving game play... Though some people may dissagree, I do not put much faith into a screenshot. Aliasing a moving object seems to be more difficult than aliasing a non moving object. Jaggies will usally be the most noticable while running through a hallway or archway and so on... But standing still? Not so bad.

 

[TGS]

From a competetive gamers stance, jaggies will cost you big time in FPS games. People typically snipe or hide on the edges of hills/buildings where they can be easily spotted. Pending you are using 4xAA or better. I have a hard time determining players against a horizon with 2xaa or less. The jaggies have that nice escalator effect while you pan the screen back and forth, making it nearly impossible to pick out a player.

Now if you aren't concerned about a player popping his head out next to a building, I can see where 0xAA or 2xAA would be fine.

 

[xtknight]

I tested in-game No AA, 2xAA, 4xAA, then I tried the best AA mode available in nHancer (2x2SS+4xRGMS) with AA off in BF2 so they didn't coincide. The scene I captured (with print screen) should be very close in each shot because I suicided until I got the same spawn point as the first shot and made sure I didn't move my mouse at all during the process. The reason for testing the super RGMS AA mode was to demonstrate damn good/ideal AA versus what was available here. It's easy to see BF2 uses multisampling because the trees aren't smoothened out like they are in the 2x2SS mode.

BF2 Daqing Oilfields 64 No AA
BF2 Daqing Oilfields 64 2x AA
BF2 Daqing Oilfields 64 4x AA
BF2 Daqing Oilfields 64 2x2SS+4xRGMS AA

Uncompressed bitmap screen captures

My opinion:

While 2xAA still looks better than NoAA, the 2xAA does a fairly poor job. 2xAA->4xAA is quite a difference. 4xAA and RGMS aren't too far apart in terms of lines.

 

[Lonyo]

Originally posted by: xtknight
I tested in-game No AA, 2xAA, 4xAA, then I tried the best AA mode available in nHancer (2x2SS+4xRGMS) with AA off in BF2 so they didn't coincide. The scene I captured (with print screen) should be very close in each shot because I suicided until I got the same spawn point as the first shot and made sure I didn't move my mouse at all during the process. The reason for testing the super RGMS AA mode was to demonstrate damn good/ideal AA versus what was available here. It's easy to see BF2 uses multisampling because the trees aren't smoothened out like they are in the 2x2SS mode.

BF2 Daqing Oilfields 64 No AA
BF2 Daqing Oilfields 64 2x AA
BF2 Daqing Oilfields 64 4x AA
BF2 Daqing Oilfields 64 2x2SS+4xRGMS AA

Uncompressed bitmap screen captures

My opinion:

While 2xAA still looks better than NoAA, the 2xAA does a fairly poor job. 2xAA->4xAA is quite a difference. 4xAA and RGMS aren't too far apart in terms of lines.

I would agree with your opinion, the roof near the centre top is the most obvious point to look at.

 

[ArchAngel777]

Originally posted by: xtknight
I tested in-game No AA, 2xAA, 4xAA, then I tried the best AA mode available in nHancer (2x2SS+4xRGMS) with AA off in BF2 so they didn't coincide. The scene I captured (with print screen) should be very close in each shot because I suicided until I got the same spawn point as the first shot and made sure I didn't move my mouse at all during the process. The reason for testing the super RGMS AA mode was to demonstrate damn good/ideal AA versus what was available here. It's easy to see BF2 uses multisampling because the trees aren't smoothened out like they are in the 2x2SS mode.

BF2 Daqing Oilfields 64 No AA
BF2 Daqing Oilfields 64 2x AA
BF2 Daqing Oilfields 64 4x AA
BF2 Daqing Oilfields 64 2x2SS+4xRGMS AA

Uncompressed bitmap screen captures

My opinion:

While 2xAA still looks better than NoAA, the 2xAA does a fairly poor job. 2xAA->4xAA is quite a difference. 4xAA and RGMS aren't too far apart in terms of lines.

Before I read your edited statement, I thougt the same thing. The only part I could tell a difference in was the slanted wooden roof. The difference between 2XAA and no AA was extremely minor... 2x2SS+4xRGMS looked the best, IMO.

 

[PrayForDeath]

Personally, I can't stand jaggies, so I need at least 2x AA. I suppose a lower res with AA is faster than a higher res with no AA, but you need to see benchmarks to be sure.
And when you're gaming on a +19" monitor then the jaggies are not minimal on 1600x1200.

 

[DRavisher]

Originally posted by: xtknight
My opinion:

While 2xAA still looks better than NoAA, the 2xAA does a fairly poor job. 2xAA->4xAA is quite a difference. 4xAA and RGMS aren't too far apart in terms of lines.

I actually disagree. I think those screens show that 2xAA is quite good. The difference between 2xAA and 4xAA seems very minor, and at 1920x1200 it is not something i even notice. I am quite aware of jaggies with no AA, but with 2xAA they cease to bother me completely. I do however understand that at 1280x1024 and lower 4xAA may be preferable, but 2xAA definately helps a lot.

 

[TGS]

Originally posted by: xtknight
I tested in-game No AA, 2xAA, 4xAA, then I tried the best AA mode available in nHancer (2x2SS+4xRGMS) with AA off in BF2 so they didn't coincide. The scene I captured (with print screen) should be very close in each shot because I suicided until I got the same spawn point as the first shot and made sure I didn't move my mouse at all during the process. The reason for testing the super RGMS AA mode was to demonstrate damn good/ideal AA versus what was available here. It's easy to see BF2 uses multisampling because the trees aren't smoothened out like they are in the 2x2SS mode.

BF2 Daqing Oilfields 64 No AA
BF2 Daqing Oilfields 64 2x AA
BF2 Daqing Oilfields 64 4x AA
BF2 Daqing Oilfields 64 2x2SS+4xRGMS AA

Uncompressed bitmap screen captures

My opinion:

While 2xAA still looks better than NoAA, the 2xAA does a fairly poor job. 2xAA->4xAA is quite a difference. 4xAA and RGMS aren't too far apart in terms of lines.

What I would be concerned about is the forested area to the right of the aim reticule. Without AA someone could easily be in that area and short of muzzle flash effects you wouldn't see them. Or have great difficulty trying to do so. Especially so when you started to pan across that hill side. The entire horizon would appear to be moving.

 

[ElFenix]

Originally posted by: rbV5
I test all my settings both in-game settings and forced with the drivers with several rigs in all my games a couple times "minimum" every month.

shens

 

[rbV5]

Originally posted by: ElFenix

Originally posted by: rbV5

Originally posted by: ElFenix

Originally posted by: rbV5
In virtually every case I've tried 2xAA is virtually a worthless setting. IMO, a higher resolution is always better than a lower resolution@ 2Xaa.

jaggies are quite noticeable at 1600x1200 ('edge crawling') but turn on 2x AA and it is much much better.


i seriously wonder if any of you have every bothered turning on 2xAA before.

I test all my settings both in-game settings and forced with the drivers with several rigs in all my games a couple times "minimum" every month. My statement is "IMO" and is true for me for the vast majority of cases.

shens

Shens? because prefer higher resolution to 2XAA? FYI, I'm a ATI beta tester for Catalyst Drivers, and have been for some time now. I have no less than 3 rigs running ATI graphics at any time, along with a houseful of PC gamers and PC users, even on months where my job keeps me particularly busy, I manage to get some testing done

 

[ElFenix]

Originally posted by: rbV5
FYI, I'm a ATI beta tester for Catalyst Drivers, and have been for some time now.

nm then

 

[BFG10K]

Like I said in the other thread, 2xAA is practically useless since the sampling pattern is not big enough to impact IQ very much.

What exactly is 2xAA?

Two multi samples, rotated grid.

 

[nRollo]

Originally posted by: BFG10K
Like I said in the other thread, 2xAA is practically useless since the sampling pattern is not big enough to impact IQ very much.

What exactly is 2xAA?

Two multi samples, rotated grid.

It makes a big difference, as evidenced by my link BFG. Not everyone can afford the sort of hardware you buy, and 2X AA may be all they can do.

Somebody with better hardware than you have might find the next level of resolution playable by dropping to 2X.

It's not "useless" and several here have posted that.

 

[nRollo]

Originally posted by: ElFenix

Originally posted by: rbV5
FYI, I'm a ATI beta tester for Catalyst Drivers, and have been for some time now.

nm then

rbV5 is also a beta tester of ATI products AFAIK, he does this sort of testing all the time.

 

[Munky]

From my own experience 2xAA is amost as bad as no AA. Not only are you getting less samples, but the 2 sub-pixel samples always lie on a line, and any polygon edge that has an angle of similar value to the angle of that line will get NO AA. I made some screenshots from the COD2 demo to illustrate my point.

0x AA
2x AA
4x AA

There's a HUGE difference between 2x and 4x, especially looking at the antennas on top of the boat. Also, look at the near side of the bow - it looks like it's getting no benefit at all from 2x AA, probably because the edge lies on a 45 degree angle which happens to be the same angle my video card uses to place the 2 sub-pixel samples.

Bottom line - 2x AA provides only a marginal improvement in IQ at best, and worst case scenario - no benefit.

 

[Pete]

To continue from the last thread, Rollo, QC does indeed blur the entire scene. You can see this by switching b/w the 2x and QC modes in that NVNew applet you linked, and note how the textures get blurred (and therefore somewhat smoothed, as is evident in the eagle texture below the left window) with QC. Personal preference.

Lonyo, it's really user pref and possibly game-dependent whether higher res is preferable to higher AA. If you read HOCP's or AT's reviews, sometimes the reviewer prefers more AA samples to a higher res. Below 10x7, you'd probably prefer higher res simply b/c textures will also benefit, AA or not. Above 10x7 (at what we consider high resolutions), you generally have enough pixels to do textures justice, so edge aliasing may become more prominent.

xtknight, 2xOG is either 2x1 or 1x2 (XxY), as its samples lie on one axis only. 2xRG (rotated--aka jittered--grid) actually offers twice the edge resolution in both axes, as its samples are rotated and so contribute to both axes. We're talking RG here, mainly because every consumer card that did MSAA (starting with GF3, IIRC) used a RG with 2x. 4x was another matter.

munky, your screenshots show 2x doubling edge resolution in both axes (therefore it's jittered/rotated, not ordered, grid). This is exactly the same improvement in edge resolution you get going from 2x to 4x. I'd find 2x worthwhile, definitely preferable to no AA if the performance hit is minimal (which it is with modern cards). I think your 2x shot looks noticably nicer than your AA-less shot.

BTW, who asked for 2x AA performance? (Performance relative to no AA, actually, and only on Trackmania Sunrise at 16x12 with the fastest cards available.) That link also show 4x and 6x, if possible. Note that 6x on ATI cards is a sparse, not rotated/jittered, grid. Also note that MSAA performance hit will vary game to game, depending on how much is being AA'ed.

Edit: Here's an eye-opening post (in an eye-opening thread) showing the difference b/w ATI's gamma-"corrected" AA and nV's non-GC AA. Eventually we discover that ATI's GC AA looks better on CRTs than most LCDs, possibly b/c CRTs tend to be closer to the gamma ATI is aiming for (2.2). It's worth a read if you're curious about AA.

Also worth a read is this paper, which explains AA and also the diff b/w OG and JG/RG AA. Hmmm, the all-important later pages (with the very enlightening screenshots) are missing. You can grab the PDF of this paper (which they prepared for 3dfx but apparently everyone and his brother are hosting) here (more links to the PDF here). The pics show why 4xOG is generally superior at ~45 degree angles, and why 4xRG is superior at ~0 and ~90 degree angles (which, b/c of our monitors' relatively low DPI, tend to stand out more as aliased--see p.14 of the PDF).

 

[BFG10K]

It makes a big difference, as evidenced by my link BFG

No it doesn't, neither in screenshots nor in actual gameplay.

Not everyone can afford the sort of hardware you buy, and 2X AA may be all they can do.

Some people can't afford better than a cardboard box. So what? That has what to do with the issue of 2xAA offering next to no IQ gain?

It's not "useless" and several here have posted that.

Several people also posted the 5800U sucked and that 6600GT/6800 SLI is useless compared to single cards.

 

[DRavisher]

Originally posted by: BFG10K

It makes a big difference, as evidenced by my link BFG

No it doesn't, neither in screenshots nor in actual gameplay.

I can't understand how you can say that. Looking at the screenshots show that 2xAA versus no AA is a big difference. 4xAA is better, but I honestly don't even bother switching on 4xAA even if I can run it with decent FPS. 2xAA is more than enough, unless you game at a low resolution of course.

 

[gsellis]

Originally posted by: Pete

Edit: Here's an eye-opening post (in an eye-opening thread) showing the difference b/w ATI's gamma-"corrected" AA and nV's non-GC AA. Eventually we discover that ATI's GC AA looks better on CRTs than most LCDs, possibly b/c CRTs tend to be closer to the gamma ATI is aiming for (2.2). It's worth a read if you're curious about AA.

2.2? NTSC? Interesting.

And thanks. I just realized why one of my graphics is blown out on a DVD I just finished... I used Gamma of 1 and it should have been 2.2. Re-render, re-recode, re-burn ...

 

[dug777]

i find 0-2xAA to be a more important step than 2-4xAA myself, 2xAA cleans up the worst jaggies very nicely...but that's just my 2c

 

[nRollo]

Originally posted by: BFG10K

It makes a big difference, as evidenced by my link BFG

No it doesn't, neither in screenshots nor in actual gameplay.

You're an interesting ATer BFG.

There is clearly a noticeable diffference in the screenshots I posted a link to, as others in this thread have confirmed, yet you continue to state there isn't. The article I linked to said there was a difference, and XTKnights shots show a difference.

Yet you say only 4X AA makes a difference.

OK you want to talk sampling differences?

I say to you your 4X AA is "useless, not worth bothering with" because my 8X and 16X SLI AA offer "superior" IQ. Yep, I don't see how anyone could even look at such meager AA when 16X AA is available on 7800GTX SLI*, that 4X stuff hurts my eyes.
Only people who haven't properly tweaked their hardware are stuck at 4X I guess.

*don't even try to say you can play 16XAA on a 6800U, Quake1 and Unreal don't count.


Obviously I'm kidding here, but the point is people probably shouldn't:
A. Deem only the settings they run as "useful" or "worthwhile"
B. State things that obviously aren't true. Anyone who can see can note the difference in 2XAA and none, and to a person considering a card that can only do 2X that is an important distinction.

 

[otispunkmeyer]

i find 2xaa at 10x12 to be the best balance of IQ and Performance

most games are absolutely fine at 2xaa, with bf2 as the exception...that still looks jaggy at 4xaa!

 

[ArchAngel777]

Originally posted by: Rollo

Originally posted by: BFG10K

It makes a big difference, as evidenced by my link BFG

No it doesn't, neither in screenshots nor in actual gameplay.

You're an interesting ATer BFG.

There is clearly a noticeable diffference in the screenshots I posted a link to, as others in this thread have confirmed, yet you continue to state there isn't. The article I linked to said there was a difference, and XTKnights shots show a difference.

Yet you say only 4X AA makes a difference.

OK you want to talk sampling differences?

I say to you your 4X AA is "useless, not worth bothering with" because my 8X and 16X SLI AA offer "superior" IQ. Yep, I don't see how anyone could even look at such meager AA when 16X AA is available on 7800GTX SLI*, that 4X stuff hurts my eyes.
Only people who haven't properly tweaked their hardware are stuck at 4X I guess.

*don't even try to say you can play 16XAA on a 6800U, Quake1 and Unreal don't count.


Obviously I'm kidding here, but the point is people probably shouldn't:
A. Deem only the settings they run as "useful" or "worthwhile"
B. State things that obviously aren't true. Anyone who can see can note the difference in 2XAA and none, and to a person considering a card that can only do 2X that is an important distinction.

Rollo is correct, I follow his thinking here spot on. 2X isn't useless. But depending on the person, they may find it either works great, or doesn't do it for them. Either way, it isn't useless, because it has a use, you just are not interested in using it.