How Important Is FSAA

Suma (of BI) has said:

We will try to enable MS AA in some patch. It should be possible, if you will tolerate some minor artifacts on some shadowed edges (I have seen the same artifact in many games with AA enabled, therefore I think it is not that bad).

How high of a priority is having Full Scene Anti Aliasing to you?

Suma (of BI) has said:

How high of a priority is having Full Scene Anti Aliasing to you?

Getting myself a cola is a higher priority.

A fillrate of 200% is enough to satisfy me.

And if you run at Full HD resolutions, AA is redundant.

Getting myself a cola is a higher priority.

A fillrate of 200% is enough to satisfy me.

And if you run at Full HD resolutions, AA is redundant.

On a 37 inch display, AA is anything else than redundant!

And higher fillrate is an absolute performance killer on these resolutions.

FSAA does not nearly degrade the performance as much as fillrate does.

as a 3rd generation game, i think it would be nice to have such to increase realism.

Well not everyone has an enormous display, and/or a GPU that requires a power supply that can double as a welder. I think we'll see a response that reflects the Steam System Survey Statistics, where for most the benefit from FSAA is enormous with regard to both performance and aesthetic realism -- but of course I could be completely be wrong...

Personally, I can't feature a contemporary squad tactical shooter, no less a simulator where long draw distance to realistically support ranged combat is the norm and precise separation of detail is essential not just for aesthetic realism on many systems, but for actually seeing what's going on; I've shot at wiggling alias anomalies far too many times in GRAW, and find the prospect in ArmA II even more unappealing.

:butbut:

Well not everyone has an enormous display, and/or a GPU that requires a power supply that can double as a welder. I think we'll see a response that reflects the Steam System Survey Statistics, where for most the benefit from FSAA is enormous with regard to both performance and aesthetic realism -- but of course I could be completely be wrong...

Personally, I can't feature a contemporary squad tactical shooter, no less a simulator where long draw distance to realistically support ranged combat is the norm and precise separation of detail is essential not just for aesthetic realism on many systems, but for actually seeing what's going on; I've shot at wiggling alias anomalies far too many times in GRAW, and find the prospect in ArmA II even more unappealing.

:butbut:

x2 to all, don't understand the "new way" of no AA in games, and no full hd and similar resolutions, in a monitor, is not as near as a 4xAA smoth quality;

I only could thing of an 5120 x 4096 resolution in an 19 inch monitor to equal 4xAA ,lol :p , not that far, but triying to say that high resolution is not the solution for smooth edges in a TFT, because when the native resolution increase, the inches do it also, so edges still visible, so AA IS needed, maybe not for some people but NOTHING equals the smoothness of AA nowadays

Add to that ATi and NVIDIA have spent millions on hardware optimization of Anti-Alias technology, it's sort of the gold standard of hardware accelerated render candy, that offers a high performance and broad range of settings to scale sharp alias artifact free 3D across a wide range of systems...

:butbut:

FSAA is extremly useful for targeting enemies far away. Without FSAA you cant differ an enemy from a moving tree. I don´t understand why this feature wasn´t implemented from the beginning...

Pls BI make it available in 1.02! :(

It's funny, though.

language="English";

adapter=-1;

3D_Performance=100000;

Resolution_Bpp=32;

Resolution_W=1024;

Resolution_H=768;

refresh=75;

Render_W=1280;

Render_H=800;

FSAA=0;

postFX=0;

HDRPrecision=8;

lastDeviceId="";

localVRAM=498637057;

nonlocalVRAM=258998271;

winX=0;

winY=0;

winW=1024;

winH=768;

winDefW=1024;

winDefH=768;

And once again my vote goes in favor of giving the end-user the option of doing what the user wants to do with the graphic-settings of the game.

Without FSAA you cant differ an enemy from a moving tree.

How often do you observe mobile trees? I once heard of a helicopter being shot down by a tree... dubbed that one the Rocket Propelled Pinetree...

I've shot at wiggling alias anomalies far too many times in GRAW, and find the prospect in ArmA II even more unappealing.

I seriously wonder how you managed that, considering GRAW 2 doesn't draw enemies beyond 149 meters, and afaik GRAW is no better at this. But yes, it becomes an issue in a game where long ranges exist. I've done the math and to properly ID targets (that is, have him take up at least 10 pixles in width, assuming 45cm as average human width) at 300m while keeping a 60 degree FOV, you would need almost 8000 pixels for your monitor's width. 4000 for him to be sized as 5 pixels. Just as a side note, notice how many pixels your aimpoint's dot takes up...

AA doesn't make you see a whole lot better, but it does help things.

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

The checkerboard example shows nicely what AA is about.

Anyone checked the results of setting AA in the card settings and/or the CFG?

I seriously wonder how you managed that, considering GRAW 2 doesn't draw enemies beyond 149 meters, and afaik GRAW is no better at this.

Here's an illustration:

At the end of the block, just above and slightly to the left of the reticule is a car less than 80m away, standing just to the left of the car is a soldier, you can't tell it's a soldier, or a bush, or just alias artifact crawl in-game until he shoots and you see the muzzle flash (and are likely injured or dead if you play with remotely challenging AI)...

But yes, it becomes an issue in a game where long ranges exist. I've done the math and to properly ID targets (that is, have him take up at least 10 pixles in width, assuming 45cm as average human width) at 300m while keeping a 60 degree FOV, you would need almost 8000 pixels for your monitor's width. 4000 for him to be sized as 5 pixels. Just as a side note, notice how many pixels your aimpoint's dot takes up...

Your math is obtuse, and not full-context unless it includes: resolution, how LOD is calculated and drawn, and FOV... In fact at the mean game resolution of 1280*1024 (Valve's Statistics), you can't typically id a target to with in half the distance you can in the real world, with FSAA off at typical FOV... The significance of the improvement FSAA offers depends on your criterion of 'significant', any move toward more realistic range of engagement and IFF id is worth while IMHO, and obviously the preference of the majority...

AA doesn't make you see a whole lot better, but it does help things?

That depends on what you mean by 'a whole lot better'; in versus multi-player if you're on the receiving end and have a lower resolution it can mean the difference between beign able to see who's about to kill you and not being able to.

Aesthetically it's pretty obvious some in the poll (the minority) have very different standards of what's acceptable, I for example find alias artifacts a ridiculous (and I use that word literally) anomaly considering the render path, and hardware acceleration (that many of us paid a premium price for) available to remove it.

Probably more helpful and germane would have been a poll category of 'Will not buy the game unless it offers FSAA', as many that saw GRAW PC did not buy the game on that account, and even more that bought GRAW did not buy GRAW 2 for the same reason... All perfectly silly in the case of ArmA II with its DirectX 9 engine that should be easily able to support every AA mode...

:(

Here's an illustration:

At the end of the block, just above and slightly to the left of the reticule is a car less than 80m away, standing just to the left of the car is a soldier, you can't tell it's a soldier, or a bush, or just alias artifact crawl in-game until he shoots and you see the muzzle flash (and are likely injured or dead if you play with remotely challenging AI)...

Your math is obtuse, and not full-context unless it includes: resolution, how LOD is calculated and drawn, and FOV... In fact at the mean game resolution of 1280*1024 (Valve's Statistics), you can't typically id a target to with in half the distance you can in the real world, with FSAA off at typical FOV... The significance of the improvement FSAA offers depends on your criterion of 'significant', any move toward more realistic range of engagement and IFF id is worth while IMHO, and obviously the preference of the majority...

That depends on what you mean by 'a whole lot better'; in versus multi-player if you're on the receiving end and have a lower resolution it can mean the difference between beign able to see who's about to kill you and not being able to.

Aesthetically it's pretty obvious some in the poll (the minority) have very different standards of what's acceptable, I for example find alias artifacts a ridiculous (and I use that word literally) anomaly considering the render path, and hardware acceleration (that many of us paid a premium price for) available to remove it.

Probably more helpful and germane would have been a poll category of 'Will not buy the game unless it offers FSAA', as many that saw GRAW PC did not buy the game on that account, and even more that bought GRAW did not buy GRAW 2 for the same reason... All perfectly silly in the case of ArmA II with its DirectX 9 engine that should be easily able to support every AA mode...

:(

Hi, registered only to say x2 to all of this, can't stand alias in games, real life do not have it, please let the game be as realistic as posible (specially if it's technically possible like this)

Yes please, if possible already for 1.02. Right now, I actually don't want to dive fully into the game since I'd rather experience the campaign with PERFECT graphics, which includes FSAA for me.

To the guy who said "Use Fillrate" - it's a totally uneconomic and not especially good looking cheap substitute. And no, with HD resolutions comes screen size, so FSAA is never redundant, and this is where fillrate gives the most problems too since the cards suddenly have to render a real ton of pixels, whereas 2x FSAA costs about 2 frames on 1920x1200 with current hardware.

...

A fillrate of 200% is enough to satisfy me.

And if you run at Full HD resolutions, AA is redundant.

Err HD isnt much.. 1900/1200.... You can always get crawlies and jaggies at any rez... it worse on 37in screens...

yes, please add conventional AA... Fillrate-setting costs far to much performance...

FSAA? Couldn't care less. I want every CPU resource i can get to be used for more FPS or other details.

Resolution doesn't matter, as the calculation was "how much resolution do I need to see X?". FOV is listed, FOV in meters is calculated by the FOR angel, range and basic trigonometry. Of course, in your (great) example, the target is both partially concealed, and therefore has less than 45cm visible, plus you run at a rather low resolution which I'm surprised to see is standard, considering even rather old monitors can run 1600X1200, which is almost as detailed (pixel-wise, not by other qualities such as color) as top-notch monitors.

My main question is how much AA would help you in identifying the target in the above picture. Similar examples from games with AA would be appreciated. It's interesting whether something like higher terrain/object detail (ignoring "very low" for obvious reasons) would help you see him better than increasing AA, though the more I look at it the more it seems like low details + high AA is the superior solution for long range IDing/shooting.

Note that my vote is definitely for having AA, after all, anything that can improve things in any way (and is a basic graphic card function) needs to be possible.

)rStrangelove;1303671']FSAA? Couldn't care less. I want every CPU resource i can get to be used for more FPS or other details.

I'm sure no one would care 2-4 fps less with 2xAA (optional of course), if you only want fps, run at 640x480 in super low quality, and you have a nice tactical shooter of the mid 90's.. :p

Resolution doesn't matter, as the calculation was "how much resolution do I need to see X?".

Incorrect... Do the math: if at 100m on one screen the player model occupies only six pixels and at twice the resolution twelve pixels the invisible becomes visible -- you can demonstrate this empirically on any PC but turning off scaling in your driver and switching resolution.

I do understand what you're saying i.e. how much object resolution in pixels is required to see a man-target, but you can't take this in isolation of dpi/pixel pitch and resolution and really have a meaningful number.

My main question is how much AA would help you in identifying the target in the above picture. Similar examples from games with AA would be appreciated. It's interesting whether something like higher terrain/object detail (ignoring "very low" for obvious reasons) would help you see him better than increasing AA, though the more I look at it the more it seems like low details + high AA is the superior solution for long range IDing/shooting.

This is very engine dependent, as a lot of variables effect how much FSAA improves distance vision: color depth, resolution, type and amount of AA employed, LOD, light model, texture resolutions, etc....

For this discussion and for the sake of making the best evaluation about FSAA in ArmA II, using ArmA or VBS2 with FSAA on and off to make comparisons would probably be the most constructive.

The summery answer though is the distance differences are small, but significant in the life and death context of seeing and being seen in MP gaming.

I will conceed; for SP FSAA is more of an aesthetic consideration, but as I offered elsewhere -- I find alias artifacts an intolerable and ridiculous anomaly considering the render path, and hardware acceleration available to remove it that we all paid for.

Note that my vote is definitely for having AA, after all, anything that can improve things in any way (and is a basic graphic card function) needs to be possible.

Yes, I suspected as much, and I think the technical discussion is good.

:)

Right now there are many more important modifications ArmA2 needs.

FSAA is something I would like to see eventually, but there are countless things I would like to see before it.

Incorrect... Do the math: if at 100m on one screen the player model occupies only six pixels and at twice the resolution twelve pixels the invisible becomes visible -- you can demonstrate this empirically on any PC but turning off scaling in your driver and switching resolution.

The math I was doing was meant to answer the question "If I need someone to take up >=10 pixels of my screen in order for me to identify him, using 60 degrees of FOV, what resolution would I need to have on my monitor to be able to ID him at 300m?" That means if your resolution is lower than the result, which it definitely is at this range/required size and even with 1/2 the required size (5 pixels), then you cannot identify the target. The idea was to show how hard it is to identify distant targets even with the highest resolutions out there. The zoom feature helps a lot, but it really needs to be tweaked so that my visibility range at max zoom is the same as it is IRL, while my FOV at minimal zoom is the same as IRL. Currently, even with max zoom your FOV is too wide to be able to ID targets at anywhere near the distance you should be able to IRL. All using the naked eye, of course, though optics should get similar treatment, tuning them using their RL values.

FSAA important here. Dont play a game without it. Dont have ARMA2 yet but friends with hefty machines reports heavy FPS loss with Fillrate so yeah - FSAA please if you can BIS.

Alex

Currently 71.28% vote for FSAA :D

Comparing ArmA1 with 4xAA and ArmA2 with no AA is a big difference, fillrate is not a good option when running 1920x1200 (it's only good for screenshots)

Currently 71.28% vote for FSAA :D

Comparing ArmA1 with 4xAA and ArmA2 with no AA is a big difference, fillrate is not a good option when running 1920x1200 (it's only good for screenshots)

In fact 'Fillrate' at it's maximum setting is not even technically as good as 2xAA. I think a lot of people if they understood what 'Fillrate' was would find they like it even less... Esentially Fillrate is using your video card (in the case of 200% setting as an example) to double the resolution your system has to calculate and render, but then descales that to your screen resolution. Very few people have the render overhead to essentially double the resolution output of their GPU on an already resource heavy game, and even worse; the output at this maximum setting will still have alias artifacts, and descale artifacts that are similar to alias artifacts.

The math I was doing was meant to answer the question "If I need someone to take up >=10 pixels of my screen in order for me to identify him, using 60 degrees of FOV, what resolution would I need to have on my monitor to be able to ID him at 300m?" That means if your resolution is lower than the result, which it definitely is at this range/required size and even with 1/2 the required size (5 pixels), then you cannot identify the target.

Yes, I understand, it's just that the way you've gone about calculating and discussing this is exclusive to your system rather then a specific dot pitch at a specific resolution to allow comparisons as we're all on very different systems (see below).

The idea was to show how hard it is to identify distant targets even with the highest resolutions out there. The zoom feature helps a lot, but it really needs to be tweaked so that my visibility range at max zoom is the same as it is IRL, while my FOV at minimal zoom is the same as IRL. Currently, even with max zoom your FOV is too wide to be able to ID targets at anywhere near the distance you should be able to IRL. All using the naked eye, of course, though optics should get similar treatment, tuning them using their RL values.

What you describe is a serious limitation of all FPS games that the original Ghost Recon attempted to scale and correct by giving up the realistic aesthetic of a first person perspective for an aesthetically less realistic aim-zoom capacity to see much closer to realistic ranges of engagement. The idea was a game that played more realistically then how its player perspective looked -- and to an extent it was successful in that regard.

The amount of zoom required to get to realistic distance vision in games will as RSE found obviate having a first person perspective as your weapon gets huge, or it stays the same and you just have a massive folding of FOV where the amount of zoom needed you get a rather sickening perspective slew were you appear to move every time you zoom your FOV to identify targets.

More on the track I was originally pursuing; accepting ArmA II's aim system 'as is' (don't think BI is going to change it), we're still all on different systems, with different resolutions, render settings, quality of hardware, etc. -- these difference can be nearly as dramatic as the scale of the IFF problem itself for multi-player gaming.

While this is something Scale Render R&D is working on, in the mean time the best solution short of resorting to playing on a console with the biggest display available, or a 'Ghost Perspective Mod' for ArmA II (which might be a good idea) is for Game Developers to offer as many render settings as possible. Cutting FSAA out of the equation is a step backward in this regard functionally as well as aesthetically, in addition to all the issues already discussed.

:butbut: