bFXAAEnabled versus iMultiSample

[DoubleYou]

In testing and taking screenshots, I think it would be worthwhile to recommend all users abandon iMultiSample altogether. It is intensively performance costly, and the built-in FXAA appears to even be better. See the compares below.

 

 

iMultiSample=0 and bFXAAEnabled=0

iMultiSample=2 and bFXAAEnabled=0

iMultiSample=4 and bFXAAEnabled=0

iMultiSample=8 and bFXAAEnabled=0

iMultiSample=0 and bFXAAEnabled=1

 

If you think that these compares progress from worst to best, then you agree.

[oqhansoloqo]

What about iMultiSample=0, but then SMAA as a proxy through ENB versus FXAA?

[hishutup]

The last one looks overall blurry, in a bad way.

MSAA really stands on its own when it comes something like the boat sail ties.

 

Here is some from fallout 3, I remember one being MSAA and the other being possibly ENB Edge AA and something else.

Pay attention to the part of the image below. It should be very obvious which full image uses what type of AA.

 

 

 

 

I cant even really play skyrim anymore because of the framerate and the Aliasing. The only real viable solution is to downsample(1527x2715 -> 1920x1020) which is expensive.

[oqhansoloqo]

The last one looks overall blurry, in a bad way.

 

I thought I noticed a bit of blur as well compared to the one before it.

[DoubleYou]

This comparison is only meant to be between bFXAAEnabled and iMultiSample. Please don't send us down other AA trails, or we'll never get done!

 

The blur is the obvious downfall of FXAA, although I will say that it is more noticeable in the picture because of the way I have scaled it. I didn't even notice it in-game. I probably also should have disabled DOF for the compare.

[TechAngel85]

This will be subjective because you have a loss of detail anytime FXAA is used. This is because of how FXAA renders. Like MSAA, it applies the AA to the all the pixels on the screen, not just the edges. However, it does not analyze any of the models like MSAA would. FXAA only computes from the pixels, themselves, and the result is often this "blur" everyone is seeing. There's a reason it's called Fast "approXimate" Anti-Aliasing.  ::  Some users like this "blur" effect because it gives the game a softer feel, but others dislike it. This effect can sometimes be corrected with sharpening from an ENB. FXAA is lighter on system resources by taking very little processing power to compute, meaning it would be better for those that are on low-end systems or those wanting better performance than they get with MSAA. Still, I would recommend those low-end users try SMAA before recommending FXAA.

 

My 2 cents.

[Guest]

If you squint really hard, that's FXAA in real life.

[TirigonX]

 It is intensively performance costly, and the built-in FXAA appears to even be better. 

 

 

I can't agree with that at all. I don't have problems to run Skyrim with 4x MSAA at 60 FPS and I don't have an uber GPU. FXAA looks worse IMO. It just makes everything blurry. People need to test this for themselves and choose depending on their hardware and gaming priorities. To me proper anti aliasing is very important but I also want sharp textures. FXAA and SMAA just can't fulfill my expectations.

I also tried supersampling and downsampling, but they're causing too much of a fps drop on my machine. Compared to those MSAA is pretty performance friendly and is my current preference.

[Kesta]

From your pics, iMultiSample=8 look better than FXAA in my eyes.

 

No idea about the performance cost.

 

 

As for SMAA, it's a bad choice for STEP imo, as it make most of the custom fonts look hideous, which kinda reduce the possibilities of customization lot of users probably like to have when they're using STEP.

[DoubleYou]

Good discussion. The blur is quite noticeable on FXAA now that I'm looking for it. Naturally AA performance will be different on different video cards. If I put my card on supersampling, jaggies simply disappear and it looks awesome if only it didn't cost me ~10fps. iMultiSample at 4 is just a little more jaggy then FXAA, so it probably is a good equivalent.

[TechAngel85]

As for SMAA, it's a bad choice for STEP imo, as it make most of the custom fonts look hideous, which kinda reduce the possibilities of customization lot of users probably like to have when they're using STEP.

The only place SMAA is recommended is on the ENB Guide where ENBs are involved and MSAA can't be used. I personally use it combined with ENBSeries's EdgeAA instead of the game's MSAA and have no issues nor any jaggies. Still run 50-60FPS in all places as long as the a preset isn't active.

[Kesta]

The only place SMAA is recommended is on the ENB Guide where ENBs are involved and MSAA can't be used. I personally use it combined with ENBSeries's EdgeAA instead of the game's MSAA and have no issues nor any jaggies. Still run 50-60FPS in all places as long as the a preset isn't active.

Yep, I was just "answering" to oqhansoloqo's post 

[TechAngel85]

I would test the performance of every AA method, but I'm afraid my results would be rather bland due to my system. I'd not likely see any loss of performance for most of the AA methods until I got up to the higher settings. Someone with a more mid-range system would need to do that testing.

[DoubleYou]

I really think that it will vary per machine.

[GrantSP]

Comparisons such as these are going to be very difficult for users to decide which course of action to take because the variables are far too large.

Even if the comparison shots were short video snippets, the decision to go MSAA, FXAA, ENB or none of the above are not based purely on what you see, but as others have said on what you 'experience'.

 

Placing all these data in a table along with the minimum GPU, FPS-loss/gain, blurriness, etc. would come close to being an aid but even then it is still too subjective.

Really these sort of things fall into more of a 'marketing' discussion rather than a 'technical' discussion. It's a case of "do you LIKE this or not?", not "does this IMPROVE the game?"

 

My 2 cents worth.