Digital Foundry Article Technical Discussion Archive [2015]

Status
Not open for further replies.
Oh wow, not the AF debate again! It was added into Dying Light with absolutely no performance hit.

This AF debacle is hilarious. Sony need to fix that asap, kinda glad i don't buy multiplats on Ps4 :yep2:

And, wow Clukos, you prefer lower res, fps (usually) and other effects switched off just to have AF? lol
 
Oh wow, not the AF debate again! It was added into Dying Light with absolutely no performance hit.



And, wow Clukos, you prefer lower res, fps (usually) and other effects switched off just to have AF? lol

He has a good PC(970?) better framerate, AA and AF at least and more effect for PC games with GameWorks...
 
Oh wow, not the AF debate again! It was added into Dying Light with absolutely no performance hit.

Just curious - but do we have proof that Dying Light devs didn't spend more time optimizing/rewriting improper implementations for PS4 so that it could be added back in? Because if it was a simple as just flipping a switch, an AF patch would have been released within a 2-3 weeks after cert, not months after launch.

After 100K+ lines of code they don't know how to flip on AF? Really? haha... man developers are so dumb /s
 
Just curious - but do we have proof that Dying Light devs didn't spend more time optimizing/rewriting improper implementations for PS4 so that it could be added back in? Because if it was a simple as just flipping a switch, an AF patch would have been released within a 2-3 weeks after cert, not months after launch.

After 100K+ lines of code they don't know how to flip on AF? Really? haha... man developers are so dumb /s

Sony release a note in dev forum to remind how to do AF on PS4 and they change the API to made it easier to do...
 
Last edited:
Forza 6 has a low level (if any) of AF, and xbox one is known for having AF in lots of its games, so sometimes it must be for performance purpose.
Forza 6 has a pretty good level of AF for a console game:

fm6-indycaayn1wluh6.png


Forza 5 was shit:
fm5-indycaayn114uml.png
 
He has a good PC(970?) better framerate, AA and AF at least and more effect for PC games with GameWorks...

I a way the point still stands, Why buy any multiplat on a console with a PC like that!? Seems hardly worth mentioning AF when so many other features are missing! AF was the final straw over higher res, fps and less (other) effects!? ;)

Just curious - but do we have proof that Dying Light devs didn't spend more time optimizing/rewriting improper implementations for PS4 so that it could be added back in? Because if it was a simple as just flipping a switch, an AF patch would have been released within a 2-3 weeks after cert, not months after launch.

After 100K+ lines of code they don't know how to flip on AF? Really? haha... man developers are so dumb /s

Likewise I can't imagine devs would spend a great deal of time for no reward. And that's not a dig at devs.
 
Sony release a note in dev forum to reminds how to do AF on PS4 and they change the API to made it easier to do...
so this problem should have gone away months ago? Seems to keep popping back up every time we kill this topic. Under your assumption should there then be 16xAF on all PS4 games or the devs are just lazy?
 
Likewise I can't imagine devs would spend a great deal of time for no reward. And that's not a dig at devs.
fair - I'm just tired of reading about dev slagging. I really can't stand it.
And I'm going to reach for that 9mm. I'm done with this topic again.
 
If AF didn't cause a performance drop it would be on. And we see as written, Xbox has had as many low AF as PS4. The simplest answer usually is the case, looking for another reason is just a waste of time.

Yet we have some fairly high profile games like Dying Light where AF was added in patch 1.05 with no loss to performance. Or the case with GTA V, where AF was in there from the start, then accidentally removed by patch 1.08 (with no increase to performance) then restored again again with patch 1.10 with no detriment to performance.

I think the simplest answer is that it can be fairly easy to break AF on all three platforms.

As to why AF (in some scenarios) may have such a big impact and in other scenarios not perhaps, more research should be done there by the fellow gaming community; opposed to running off assumptions that because Q3DM1Test1 fps don't drop when I go from trilinear to 16xAF therefore this applies to everything.

In which games does AF have a big impact on performance?

fair - I'm just tired of reading about dev slagging. I really can't stand it.
You must be new here, it's well documented that all devs are lazy. :yep2: Especially the ones who post here when they should be working. :runaway:
 
so this problem should have gone away months ago? Seems to keep popping back up every time we kill this topic. Under your assumption should there then be 16xAF on all PS4 games or the devs are just lazy?

Between no AF and 16xAF, there is many options 2xAF, 4xAF or 8x AF. different level of AF for different surface and so on...

In PS4 exclusives most of the time there is AF...

It was the first PS4 game of the Dying Light developer too...
 
Yet we have some fairly high profile games like Dying Light where AF was added in patch 1.05 with no loss to performance. Or the case with GTA V, where AF was in there from the start, then accidentally removed by patch 1.08 (with no increase to performance) then restored again again with patch 1.10 with no detriment to performance.
So all games should be 16xAF on both consoles by this admission of evidence? I'm just trying to figure out where you guys are going with this.

In which games does AF have a big impact on performance?
See screenshots for Forza 5 vs Forza 6.
so did they just forget to enable it for F5?

Why is it when it comes to PS4 it's some conspiracy or it's a problem with devs or the SDK, but when Xbox goes through the same things no one makes a chirp.
And that's sorta my confusion with this whole argument. The discussion is occuring as if AF is free, in all situations for all games, for all platforms. When clearly it's not, cause even with games where PS4 has better AF, it's likely not 16xAF.

There are just so many holes, granted, it may not always be game performance based, but the fact is, Dying light still isn't 16xAF and neither is GTA i'm willing to bet. People get defensive over it too, but game performance isn't linked to hardware performance. Otherwise we'd never see year over year improvements in graphics.
 
Wow, not this debate again.

Dying Light, DmC DE, Project Cars and Super Street Fighter IV all received patches to add/increase AF on PS4 with no hit to performance. Many PS4 games that lack AF also have better performance on PS4, and some games are not demanding games at all (ie Strider), so not including AF due to performance issues doesn't make sense.
 
Last edited:
I'm cool with that
Wow, not this debate again.

Dying Light, DmC DE, Project Cars and Super Street Fighter IV all received patches to add/increase AF on PS4 with no hit to performance. Many PS4 games that lack AF also have better performance on PS4, and some games are not demanding games at all (ie Strider), so not including AF due to performance issues doesn't make sense.
Agreed, there are some cases where clearly it could have just been missed entirely - but I don't want to see this reason being applied to all cases - as in, lets not make this AF problem, only have one possible root cause.
 
Wow, not this debate again.

Dying Light, DmC DE, Project Cars and Super Street Fighter IV all received patches to add/increase AF on PS4 with no hit to performance. Many PS4 games that lack AF also have better performance on PS4, and some games are not demanding games at all (ie Strider), so not including AF due to performance issues doesn't make sense.

yup, there was also that PS3 game where you splattered paint to show the way...name escapes me, but clearly AF missing and game was not intensive at all.
 
So all games should be 16xAF on both consoles by this admission of evidence? I'm just trying to figure out where you guys are going with this.

No, because AF isn't free in terms of what the GPU is doing so applying it across the board (as you can force in drivers in Windows) isn't recommended although the performance hit still seems to be minimal. But even this performance hit can can be balanced with what assets of mipmap are available for any given texture - and this will probably depend the particular texture and whether it would be beneficial to have more levels of mipmap according to the article / technical director / graphics guy. In a driving game I'd argue it would be beneficial for most ground/road surfaces to have many mipmap levels and correspondingly higher AF during rendering. Mipmaps = textures = RAM. Just ramping up AF isn't going to do much in many cases except burn cycles.

This is why AF is set on a per texture basis.

Why is it when it comes to PS4 it's some conspiracy or it's a problem with devs or the SDK, but when Xbox goes through the same things no one makes a chirp.

Because people are crazy? There are those with theories like yours, although Nvidia appear to disagree. In their own words: "Anisotropic filtering exists to provide superior image quality in virtually all cases at the slight expense of performance." My emphasis.

And that's sorta my confusion with this whole argument. The discussion is occuring as if AF is free, in all situations for all games, for all platforms. When clearly it's not, cause even with games where PS4 has better AF, it's likely not 16xAF.

Interestingly, Nvidia also suggest:

With proper [anisotropic] filtering, the use of multiple mipmap levels in a scene can have no discernable impact on its appearance while greatly optimizing performance.​

Nobody is saying it's free, nothing is free. What many people here are saying and what most tests show is that enabling AF only slightly impacts performance. Nvidia state that when used correctly, and combined with mipmaps, it can actually increase performance.

But in the case of GTA on consoles, AF being enabled or disabled has no impact on the framerate. Or Dying Light.
 
Last edited by a moderator:
Okay lol I'm convinced: a good response. I know it is enabled on a per texture basis but I think even when looking at dying light Xbox vs PC: PC AF is still coming out on top for a lot of textures. To me enough slight performance hits must be adding up to something otherwise they'd just run it in almost all situations.
 
Okay lol I'm convinced: a good response. I know it is enabled on a per texture basis but I think even when looking at dying light Xbox vs PC: PC AF is still coming out on top for a lot of textures. To me enough slight performance hits must be adding up to something otherwise they'd just run it in almost all situations.
I don't think you'll find anybody who disagrees. And I wouldn't argue when things get stressed, AF could result in a lost frame because there has a to be a point/line where there simply isn't enough grunt to make everything happen at the desired framerate and wasteful use of AF may be the straw that breaks the camel's back.
 
I don't think you'll find anybody who disagrees. And I wouldn't argue when things get stressed, AF could result in a lost frame because there has a to be a point/line where there simply isn't enough grunt to make everything happen at the desired framerate and wasteful use of AF may be the straw that breaks the camel's back.
Well, I think there are a lot of people that sorta - unconsciously disagreed. People saw that Xbox One was running higher AF, and through the logic that XBO is inferior, therefore there is an issue with PS4. And that's sorta where, I had the issue. The performance of AF on XBO for a build developed for XBO should have no relation to a PS4 build. But people continually put them together.
 
and combined with mipmaps, it can actually increase performance.
NVidia is claiming that mipmaps improve performance*, not AF. AF requires takings more samples from larger mip levels, so it should require more work on basically all parts of the texturing system.

*Not using mipmaps means you're always sampling from the largest version of a texture, which is very bad from a coherency point of view. It also means that, if you were to pursue stable results without lots of texture aliasing, you'd need to take tons of samples and average them at each pixel; mipmaps are essentially just copies of the texture where much of that work has already been done.
 
Status
Not open for further replies.
Back
Top