The Showdown: SMAA Vs FXAA

[styckx]

I know many of you use the FXAA injector I posted here around the time of the TS 2012 release.

Well, there is a new kid on the block.. SMAA Injector.. I will not link to it, because honestly I don't want to be remotely responsible or tech support for anything you do.. Just passing along some info. I I will make a few points though.

1. If you Google and use the SMAA injector and use the Steam ingame overlay you must edit injector.ini and switch a flag from 0 to 1 to be compatible. If you do not your game will be completely washed out in brightness.

2. If you use this and your game suddenly starts crashing or SBHH.. Do not email RSC.. It's not their problem you tinkered.. Delete SMAA files, verify your cache and leave it at that.

3. Do not use this with other games while playing on anti cheat enabled servers (may be detected as a cheating).

4. This option really should be only used if you're avidly disabling AA in the Railworks wrapper as it will provide AA with little to no overhead and better than the MLAA setting.

5. Bears repeating.. Use at your own risk.. RSC, Me, Bill Gates, or Adam Sandler are not responsible for anything outside of normal use you do with your Railworks installation. See #2..

About SMAA:
We present a new image-based, post-processing antialiasing technique, that offers practical solutions to all the common problems of existing filter-based antialiasing algorithms. It yields better pattern detection to handle sharp geometric features and diagonal shapes. Our edge detection scheme exploits local contrast features, along with accelerated and more precise distance searches, which allows to better recognize the patterns to antialias. Our method is capable of reconstructing subpixel features, comparable to 4x multisampling, and is fully customizable, so that every feature can be turned on or off, adjusting to particular needs. We propose four different presets, from the basic level to adding spatial multisampling and temporal supersampling. Even this full-fledged version achieves performances that are on-par with the fastest approaches available, while yielding superior quality.

Let's roll this beautiful bean footage.. I choose NEC to show the contestants. Railworks AA slider is set to OFF for all three of these.

Click all thumbnails for link to view fullsize.(action/view all sizes)

First a base: No AA injection at all and TS 2012 as is with no AA.


noaa by styckxx, on Flickr

Second: SMAA


smaa by styckxx, on Flickr

Third: The original.. FXAA


fxaa by styckxx, on Flickr

Now cropped key areas for good comparisons.

SMAA:


SMAA Crop by styckxx, on Flickr

FXAA:


FXAA Crop by styckxx, on Flickr

SMAA:


smaacrop2 by styckxx, on Flickr

FXAA:


fxxaacrop2 by styckxx, on Flickr

SMAA.. Winner?

[ightenhill]

Thanks Bill.. Giving it a check

[rabid]

I've been using FXAA for a while now, but I'll definitely give this a try later tonight, thanks.

BTW the link I found was this:
http://mrhaandi.blogspot.com/p/injectsmaa.html

[TheTazman]

I would like to give this a go too. But its very much a spot the difference competion.

[nobkins]

I have tested it and in my view SMAA is better quality will less performance impact (not that FXAA had much if any performance impact).

It might well be a case of spot the difference but the difference is there and it is better with SMAA

[andynwt]

hopefully the game will end up doing AA properly in the near future anyway.

[ightenhill]

Does it require switching on and off and can it be used as well as the 2x1

[Griphos]

Like FXAA, it's just a drop the files into the root folder sort of thing. If you use steam overlay, there's one value to change in a config file, as Bill said.

I've also been using FXAA, so I tried this instead and compared them both to the game AA set at 4x, using HSC. I could tell the AA difference between the game AA and both of these in normal play, but not between FXAA and SMAA. But, I could tell the difference between these two in terms of the way each handled ambient light, etc. For me, on HSC, SMAA was harsher. Not as harsh as in game settings, but FXAA was both smoother in terms of light gradation, and more realistic. Also, SMAA hit performance the most, wasn't very much of a difference in that respect between all three, actually, but it was consistently a frame or three lower with SMAA than the other two.

Interesting find. And results may of course vary with different systems. Shader values can be changed in both, so there is room for tweaking for those so inclined. I'm curious to see the changes in AA RSC has promised soon. In the meantime, I'll stick with FXAA.

[Shadders]

Hi,

I through together a comparison between all the different AA modes available. These were taken on an ATI 5970 with V11.12 of the drivers, combined with 11.11 Cap 3 profiles. All shots were taken from the WCML 2010 Early Liners Part 1 scenario. I started the scenario, closed the dialogue and pressed 3 to go to the back of the train. All shots were taken using Fraps.

** Warning this is a large picture (1920 x 2400) if you have a slow connection, it may take a while **


Click the image to zoom in


Hope it helps people in choosing.

Shad.

[TheTazman]

Government health warning...your eyes may explode looking at this image.

Hi,

I through together a comparison between all the different AA modes available. These were taken on an ATI 5970 with V11.12 of the drivers, combined with 11.11 Cap 3 profiles. All shots were taken from the WCML 2010 Early Liners Part 1 scenario. I started the scenario, closed the dialogue and pressed 3 to go to the back of the train. All shots were taken using Fraps.

** Warning this is a large picture (1920 x 2400) if you have a slow connection, it may take a while **


Click the image to zoom in


Hope it helps people in choosing.

Shad.

[Griphos]

That's quite impressive (what I saw of it before my eyeballs exploded!)

Thanks for putting that together. Can similar results be expected with nVidia cards, or does that change the game?

Edit: Well, to answer my own question, yes, it seems to be different, at least in terms of FPS performance. I tried SSAA 2x2, as that produced excellent results in your comparison. But whereas you seem not to have any FPS drop with that setting, I can't even get up to 14 FPS in the HSC Altoona Yard scenario I'm testing at that setting, whereas with MSAAx4 or even x8, I get 24 FPS or above (I'm locked at 30, so don't know how high it goes as I'm not using Fraps).

There was a visible difference between those settings (SSAA and MSAA) just in game play, I believe, but it was subtle. At least in the areas where I was working. Not the best setting for testing AA probably, but I don't have WCML or NEC, and the HSC is the most demanding route on my system (except for Ohio Steel, which I can't play at all).

[Shadders]

Yep, it's a bit of a eye scrambler when they are all stuck together like that

The frame rates are just what I get on my system with my in-game settings. I wouldn't like to guess on what anyone else would get. But the frame rates are useful for showing the overall cost in comparison to no AA at all. Quality wise, I suspect that there would not be much difference if the same in-game settings and AA levels were applied to an nVidia card though. Maybe someone with an nVidia card and some free time could perform a similar test?

Edit: I forgot, my in game settings are:
Particle Density = High Detail
View Distance = Highest Detail
Object Texture = Highest Detail
Passenger Density = Medium Detail
Terrain Texture = Highest Detail
Scenery Detail = 10
Shader Quality = High Detail
Water Quality = High Detail
Shadow Quality = High Detail
Low Detail Shadows = Off
Procedural Flora = Off
Adaptive Bloom = On
Depth of Field = On

1920 x 1200 resolution

[Griphos]

I have an nVidia GTX275. As I said in my edit above, there's a significant FPS hit on my system with SSAA applied. Huge, actually. From okay to slide show (24+ to 6-10).

Funny thing, on my system, no AA costs me FPS over MSAAx4 or x8. I lose at least 5 FPS with no AA set. And if I plug in FXAA or SMAA with no in game AA, I lose the same 5 FPS at least. I can run either FXAA or SMAA with in game set to MSAAx4 at least 5 FPS faster than with in game set to no AA. Weird.

[Parrotnut]

Somehow I got the best results with AA and performance selecting MLAA in the RW start menu
and in the nVidia Contol Panel, on the left, click "Adjust image setting with preview", then select "Use my preference emphasizing" and set it to "Quality"
screen res 1920x1200, GTX580

Frank

[mjoliver1]

I've now upgraded from the FXAA PPI Rev 37 to the FXAAToolHg4 and, using it at default settings for AA, Sharpen and bloom, with other settings turned off in the panel, I am delighted with the result already. The only negative, as yet - and I'm sure it's an available adjustment, is that the f4 controls are too high on contrast/gamma. Otherwise, the new version has stacks of adjustment parameters that can be managed directly from a panel rather than by changing lines in a .txt file, as was the former case.

Mike