There seems to be a work around to enjoy FXAA in DirectX 9 content:

http://www.forum-3dcenter.org/vbulletin ... tcount=118

In another thread, we briefly discussed FXAA and the other forms of "shader algorithm "AA"". I said there I thought all the methods looked bad, FXAA in particular.

Well, anyway I came across a SMAA injector and decided to put it to the test.

My test consists of STALKER ClearSky since (1) these stupid "AA" methods were supposedly developed in response to games that use deferred rendering engines like the STALKER series, (2) the ClearSky demo is very demanding, only Metro 2033 is worse in my experience on my system, (3) it is an easy demo to control settings wise.

What I did was record several videos which I am uploading and will post as they finish uploading (might take a while due to my internet "speed" and the time I have available for this sort of thing.

Here is the first of 3 - the SRAA video. An MSAA 4x video and a NOAA video to follow.

https://hotfile.com/dl/149538382/062ec88/SMAA.mp4.html

https://hotfile.com/dl/149543035/122e644/MSAA.mp4.html

https://hotfile.com/dl/149561747/5d4b88c/NONE.mp4.html

You should set your screen res to 1280x960 full screen to watch the videos, since that is how they were recorded in FRAPS.

contents of the injector.ini file looked like this:


Attached is an image of the log file from the injector showing it is actually active

Thanks for your work. Unfortunately, i haven't looked at your video yet and it will be awhile. i have my own issues with time (and bandwidth constraints) right now.


Probably *the* engine to actually test with is the Unreal Engine 3 since it is the most popular of the deferred shader engines and it appears that its mostly what FXAA was made for. i think that deferred lighting engines will get much more popular - not less.

You are also going to see more of this type of AA and hopefully it will get smarter. It is the future whether you like it or not. Your choice is to turn it off or force MSAA if you can handle the performance issues.

MSAA video added.

https://hotfile.com/dl/149543035/122e644/MSAA.mp4.html

Last video added, the no antialiasing video.

https://hotfile.com/dl/149561747/5d4b88c/NONE.mp4.html

Now everyone can see what SRAA looks like running in a game. SRAA is superior to FXAA, so if SRAA looks like crap (and it does), then why would anyone want to use FXAA?!

No-one wants to discuss this? People got bent out of shape over this post of mine about FXAA (among other things)
viewtopic.php?f=6&t=24350&start=410#p60204

I'd be interested in Picao84's reaction. I know apoppin is too busy to investigate at present.

About the only worth these software AA algorithms have in my book is the ability to slightly improve games where it is otherwise totally impossible to get any AA at all, like Godfather II and GTA IV.

link to SMAA injector I used and link to general SMAA info.

http://mrhaandi.blogspot.com/p/injectsmaa.html

http://www.iryoku.com/smaa/

I enjoyed this post and a nice chart/comparison on some of the methods:

http://www.rage3d.com/board/showpost.ph ... stcount=37

That was some nice work, thanks grstanford!

I had a small doubt myself, after recently seeing a picture showing that FXAA was too blurry compared to the new MLAA 2.0 method (which does it cheaply anyways, with there being more aliasing as a trade-off):
http://www.tomshardware.com/reviews/rad ... 148-2.html

BTW, on the next page in that article shows more blurry textures for HD 78x0 than HD 69x0 ( http://www.tomshardware.com/reviews/rad ... 148-3.html ) as to which AMD does not have an official response yet.

Using Clear Sky was a perfect example to compare the AA methods! So much foilage, nice trees.. awesome lighting over that cemetery and the buildings! A game like this looks like 10 times the power of what Xbox360 can handle! The blurriness isn't that noticeable after all.

SMAA is designed not to be blurry. It still can't do AA for sh!t though.