Any updates on anti-aliasing work?

Well, I did a bit of FSR testing today and while I have few past points of comparison, it certainly doesn't look like the pre-FSR AA pass has improved:
ysDhyBL.jpg

COyDC2S.jpg

L7epdeX.jpg


That's 5k (5120*2880), with FSR ultra quality, plus SMAA. Extreme jaggies on specular highlights and other high contrast edges are very obtrusive, and even worse in motion.
I agree, there's a lot of work to do here still. My immediate reaction to u15 was in relation to crawl on orbit lines. That's been vastly improved. General jaggies are still present though.

General question though, at that frame rate, why are you using FSR to upscale when not doing so would look nicer and still be an acceptable frame rate?
 
As I understand it, FSR is needed for weak GPUs when you want to see 2K.

And average gpu's for 4k. Apparently the 4k scaled down (as common on consoles) provides acceptable iq. Its less invisible at lower resolutions imo. I'm still profiting from instead of upgrading one year, to instead choosing to not upgrade my monitors. 2560x1600 at 30 inches is great, and im still on a 10 year old pc :) 1080ti is fine, 2023 finally says im cpu bottlnecked, fixed by tweaking, will upgrade when rtx, a genuine new feature, is in say.. 3 games i want to play.
 
I haven't tested FSR in U15 yet, so cannot comment on any changes there.

I have noticed a reduction in twinkling of stars in the skybox; since the resolution hasn't changed, I assume they changed how it was being filtered, or the order in which the filtering pass is applied.

I've fairly conclusively ruled out a general improvement to AA.
I am cool with everything you've said about before and after at 4k, on kit that's capable of a 60fps floor with 4k.

Have you ruled out improvements to AA on kit that can barely manage peaks of 60fps at 1080? THAT is where the improvements are being reported.
 
I am cool with everything you've said about before and after at 4k, on kit that's capable of a 60fps floor with 4k.

Have you ruled out improvements to AA on kit that can barely manage peaks of 60fps at 1080? THAT is where the improvements are being reported.

Just in case you haven't read morbads posts, on his machines, he's analying drops from 250fps down to 120.. just the variation he has is multiples of the maximum i'm able to achieve so i wouldn't be surprised if there's some disconnect there. Also hes using AMD graphics.. I'm using nvidia.
 
That seems to support my point. Any bottleneck present on a Kabylake/1060 rig will have vanished into insignificance in those kinds of tests. 250fps! And yes AMD v nVidia (v Arc?!) likely to be a factor.
 
I am cool with everything you've said about before and after at 4k, on kit that's capable of a 60fps floor with 4k.

Have you ruled out improvements to AA on kit that can barely manage peaks of 60fps at 1080? THAT is where the improvements are being reported.

I've ruled out AA improvements on my RTX 4090, RX 6800 XT, and RX 6900 XT with recent drivers, as well as my 1080 Ti with two-year old drivers, at a multitude of resolutions/settings.

I tend to play at higher resolutions and I'm using higher resolutions for my comparisons because AA problems should be universally less apparent at higher resolutions. That the aliasing on specular highlights looks as bad as ever at resolutions exceeding 4k is a pretty damn strong indicator that nothing was fixed with regards to general AA.

Just in case you haven't read morbads posts, on his machines, he's analying drops from 250fps down to 120.. just the variation he has is multiples of the maximum i'm able to achieve so i wouldn't be surprised if there's some disconnect there. Also hes using AMD graphics.. I'm using nvidia.

I'm not analyzing frame rate performance at all when I'm talking about anti-aliasing. I'm analyzing the prevalence of aliasing.

Also, the screenshots above were taken on an NVIDIA setup. I have many computers.
 
Some of what look like specular highlights are actually coming from cube maps now, so that's not "anti-aliasing" as you'd conventionally recognise the term across the frame; it's a specific issue with rendering shadows and highlights (I assume mapping highlights of that sort is the inverse of mapping shadows of that sort.) So in fact that specific example is not "evidence nothing was fixed with regards to general AA" because it's not general AA. As of U15 the sun shining into the cabin is very clearly a dynamic cubemap implementation. When those look blocky it's usually something to do with the PCF implementation.

Turning back to the general AA-related symptom of "it looks jaggy" then that will be much more obvious the lower the resolution goes, because the jags are jaggier. A small improvement in fidelity here would have a much greater effect on perception at lower resolutions. Comparisons at 4k are relevant to a lot of the discussion here and in no way am I saying any of those are incorrect, but they are simply not relevant to "looks better at 1080 on my low-end gear."

That 1080Ti should have branch 535, current, drivers I think? Have you left those downlevel for a specific reason?
 
Some of what look like specular highlights are actually coming from cube maps now, so that's not "anti-aliasing" as you'd conventionally recognise the term across the frame; it's a specific issue with rendering shadows and highlights (I assume mapping highlights of that sort is the inverse of mapping shadows of that sort.) So in fact that specific example is not "evidence nothing was fixed with regards to general AA" because it's not general AA. As of U15 the sun shining into the cabin is very clearly a dynamic cubemap implementation. When those look blocky it's usually something to do with the PCF implementation.

I'm not talking about cube map reflections. I'm talking about those high contrast edges that are very prominent on many surfaces, which are the bulk of jaggies in the screens shots and videos I've posted. I'm focusing on those reflective edges because they are obtrusive, near omnipresent, and the main area where aliasing needs to be addressed. With almost any degree of illumination, those edges are white and blur poorly with what they're bordered by. AA elsewhere, with few exceptions, has been, and remains fine. The game's SMAA in particular works well on most other edges, without causing undue blur.

Turning back to the general AA-related symptom of "it looks jaggy" then that will be much more obvious the lower the resolution goes, because the jags are jaggier. A small improvement in fidelity here would have a much greater effect on perception at lower resolutions. Comparisons at 4k are relevant to a lot of the discussion here and in no way am I saying any of those are incorrect, but they are simply not relevant to "looks better at 1080 on my low-end gear."

It's not an improvement at 1080p either (one of the first comparisons I posted was at 1080p because I was also looking at CPU bound performance) and how fast the gear is irrelevant. One architectural generation of GPU almost invariably renders something the same way, no matter what segment that GPU is in. Even between vastly different architectures it's rare for games to look appreciably different (which is why the whole RX 6000 and 7000 thing is so weird...my 5700 XT produced images that looked more like--as in utterly indistinguishable from--my RTX 4090 than my 6800 XT did with the same drivers).

The only settings that matter are resolution (including any kind of super or subsampling) and the post process AA filter selecI'm not talking about cube map reflections. I'm talking about those high contrast edges that are very prominent on many surfaces, which are the bulk of jaggies in the screens shots and videos I've posted. I'm focusing on those reflective edges because they are obtrusive, near omnipresent, and the main area where aliasing needs to be addressed. With almost any degree of illumination, those edges are white and blur poorly with what they're bordered by. AA elsewhere, with few exceptions, has been, and remains fine. The game's SMAA in particular works well on most other edges, without causing undue blur.ted (which have all always worked, just with a very subtle effect that does almost nothing to the lines in question). I can't find any evidence for a general improvement to AA.

That 1080Ti should have branch 535, current, drivers I think? Have you left those downlevel for a specific reason?

Because there is little reason to update them and less reason to believe that drivers released years after NVIDIA stopped caring about Pascal will somehow make a Pascal part render an equally forgotten title better than Ampere or Lovelace parts (aside from frame rate, it produces the same image at the same settings as my RTX 4090 does with cutting edge drivers in this game).
 
I've have seen steady improvement in the drivers for my 1060, right up until now, and not just in Elite Dangerous either. Interesting though that Elite D isn't picking up any benefit of newer silicon though isn't it?
 
I've have seen steady improvement in the drivers for my 1060, right up until now, and not just in Elite Dangerous either. Interesting though that Elite D isn't picking up any benefit of newer silicon though isn't it?

Not sure what benefits we're talking about here.

This RTX 4090 is more than twice as fast as my RTX 3080 was which was in turn twice as fast as my GTX 1080 ti. Nothing looks different at the same settings because nothing is supposed to look different...when standards are even loosely followed the resulting render should be visually indistinguishable across any and all hardware that supports the API in question. Stuff looking different just because of a hardware change is good hint that something is broken or someone is trying to cut corners to cheat.
 
I'm not talking about cube map reflections. I'm talking about those high contrast edges that are very prominent on many surfaces, which are the bulk of jaggies in the screens shots and videos I've posted. I'm focusing on those reflective edges because they are obtrusive, near omnipresent, and the main area where aliasing needs to be addressed. With almost any degree of illumination, those edges are white and blur poorly with what they're bordered by. AA elsewhere, with few exceptions, has been, and remains fine. The game's SMAA in particular works well on most other edges, without causing undue blur.
The fact that they aren't shy about high contrast does bring me some solace, it does seem to indicate that they have a plan to fix that at some point.
 
Because there is little reason to update them and less reason to believe that drivers released years after NVIDIA stopped caring about Pascal will somehow make a Pascal part render an equally forgotten title better than Ampere or Lovelace parts (aside from frame rate, it produces the same image at the same settings as my RTX 4090 does with cutting edge drivers in this game).

Not that you would care, but it took me years to get off 436.48 and 511.79, but at least the recent one i found seems worth it. No effects to frametime as far as i noticed, and fixes seem very worth it, the difference between features in games actually working or not, or with elite at least going from broken unplayable to "it works on my machine" developer experience.
 
I'm not talking about cube map reflections. I'm talking about those high contrast edges that are very prominent on many surfaces, which are the bulk of jaggies in the screens shots and videos I've posted. I'm focusing on those reflective edges because they are obtrusive, near omnipresent, and the main area where aliasing needs to be addressed. With almost any degree of illumination, those edges are white and blur poorly with what they're bordered by. AA elsewhere, with few exceptions, has been, and remains fine. The game's SMAA in particular works well on most other edges, without causing undue blur.
I believe you are talking about something like this, right?
Elite Dangerous_2023.05.19-10.43.png

I believe it has something to do with the lightning of those edges of different objects. Should that be some particular broken shader program?
 
I'm not talking about cube map reflections. I'm talking about those high contrast edges that are very prominent on many surfaces, which are the bulk of jaggies in the screens shots and videos I've posted. I'm focusing on those reflective edges because they are obtrusive, near omnipresent, and the main area where aliasing needs to be addressed.
Aha! OK I agree those are terrible. That sort of environment feature is always an issue so maybe I was editing that out after 20 years of seeing the problem.

Your specific descriptions in the rest of that comment led me to find another really simple example - when you dock in one of the stations that has huge yellow girders. The edge of those girders is good old-fashioned jaggies. Even though pretty much everything else in the bay looks absolutely fine. Very odd.
 
The only acceptable AA is obtainable by using supersampling 1.5 or even 2.0. But the Game still runs really bad even if you are way above the Recommended specs so that is not an option. I really hope the next patch will fix that issue.
 
Anti Aliasing Not Working Correctlyhttps://issues.frontierstore.net/issue-detail/32228Over the course of multiple Odyssey updates, the team have worked hard to address as many of the reported issues around anti aliasing and this ticket will now be closed. We are grateful for all of the feedback you provided throughout.

 
I think we have to wait for u16 to see if it's finally taken care of. Otherwise I'm thinking a massive thread here on the forum for uploading screenshots and videos to illustrate aliasing as we come across it in the game. IMO, in u15 one only has to play the game and watch the screen to find plenty of examples. Not being a computer graphics specialist I wonder if it's all insufficient or bad antialiasing, or maybe the results of moving shadow, LoD, shaders, etc. Or maybe it's impossible to do it with such sharp edges and high contrast lighting, and they'd have to blur edges and decrease the contrast between lit up and dark areas.
 
Back
Top Bottom