42 minutes ago, St Mawgan said:

Ah well, there you go, I don't fly any of those so that might explain it. Only GA with and without glass and a few old school light-heavies such as the RJ and the F28. 

This pretty much explains it, exactly. So again: as long as not everyone clearly states his usercase, comparisons might get utterly difficult, sometimes even impossible. 

1 hour ago, PlumCrazy said:

When showing his graphics settings, he says "these settings are a little funky because I am using AutoFPS".

Thanks. That I missed.

7 hours ago, tfm said:

The guy in this video says he’s using “AutoFPS”. This means that the sim will paint more objects in the distance when it detects less overall demand. DLSS is less demanding than TAA so, unsurprisingly, the sim depicts more detail at distance. The guy says that this shows the superiority of DLSS, but all he’s done is give a demonstration of how AutoFPS works 😳

What am I missing?

Interesting, using TAA I have noticed my VRAM usage was increasing to the 90% range, whereas when using DLSS mid 40's to 50's.

Reading this thread is comical.

Here we have another know it all YouTuber being referenced that one way is better than another!

I won't  even waste my time watching this nonsense!

Every individual is using different hardware and equipment.

There is no one way that is superior for all.

This is obviously apparent by the debate created by this irrelevant garbage YouTube stuff being re-posted here!

 

Edited August 13, 2025Aug 13 by tpete61

30 minutes ago, tpete61 said:

Reading this thread is comical.

Here we have another know it all YouTuber being referenced that one way is better than another!

I won't  even waste my time watching this nonsense!

Every individual is using different hardware and equipment.

There is no one way that is superior for all.

This is obviously apparent by the debate created by this irrelevant garbage YouTube stuff being re-posted here!

 

Whats comical to me is how polar most people are on this topic… there’s pros and cons to both. And it takes 10 seconds to switch back and forth… test them both yourself and choose the right set of compromises for you.

And you can use both. You don’t need to stick to just one anti-aliasing technique. As I said earlier, in ny case I use both. I use TAA in the iFly 737 and DLAA for GA flights. 

8 hours ago, touchdown84 said:

Render Scaling > 100 does not "switch" from TAA to SSAA. TAA and Render Scaling are applied sperately for a combined result

Although not impossible, it would be highly unconventional to apply multiple AA passes using two different methods.

I interpret the Description differently from you “using more than 100% will Super Sample the rendering” … it doesn’t say “will ADD to…” or does not say “will combine…”.

Another odd issue with that graphics settings is why does the slider stop at 200 value?  When it comes to either TAA or SSAA a value of 200 has no practical meaning as applied to those AA processes.  

I’m guessing that 200 really means 2X SSAA based on the resolution value they calculate to the left of the setting?  Would be great if Asobo provided more information.

Here is a good reference to TAA: 

https://en.m.wikipedia.org/wiki/Temporal_anti-aliasing

Here is a good reference to SSAA:

https://en.m.wikipedia.org/wiki/Supersampling

As you can see, these are VERY different AA processes where TAA relies on prior frames to “smooth” out edges (jaggies) which as a result leads to ghosting on fast motion.

SSAA is a completely different (and more processing intensive) AA method that produces better images and is not subject to ghosting as it doesn’t rely on prior frames to compare against.  There is no ghosting with SSAA.

If you want to run a basic test, use TAA 100 and do some rapid motion view/camera or aircraft movement with objects relatively close (like clouds) … you will see ghosting.   Now enter values above 100 and you will NOT see ghosting when performing the same motion test.  This is why I believe the values above 100 are indeed “switching” to SSAA and not in combination with TAA.  

But, I would be interested to get some official word from the developers at Asobo in regards to this setting.

 

26 minutes ago, SayAgain said:

Although not impossible, it would be highly unconventional to apply multiple AA passes using two different methods.

I interpret the Description differently from you “using more than 100% will Super Sample the rendering” … it doesn’t say “will ADD to…” or does not say “will combine…”.

Another odd issue with that graphics settings is why does the slider stop at 200 value?  When it comes to either TAA or SSAA a value of 200 has no practical meaning as applied to those AA processes.  

I’m guessing that 200 really means 2X SSAA based on the resolution value they calculate to the left of the setting?  Would be great if Asobo provided more information.

Here is a good reference to TAA: 

https://en.m.wikipedia.org/wiki/Temporal_anti-aliasing

Here is a good reference to SSAA:

https://en.m.wikipedia.org/wiki/Supersampling

As you can see, these are VERY different AA processes where TAA relies on prior frames to “smooth” out edges (jaggies) which as a result leads to ghosting on fast motion.

SSAA is a completely different (and more processing intensive) AA method that produces better images and is not subject to ghosting as it doesn’t rely on prior frames to compare against.  There is no ghosting with SSAA.

If you want to run a basic test, use TAA 100 and do some rapid motion view/camera or aircraft movement with objects relatively close (like clouds) … you will see ghosting.   Now enter values above 100 and you will NOT see ghosting when performing the same motion test.  This is why I believe the values above 100 are indeed “switching” to SSAA and not in combination with TAA.  

But, I would be interested to get some official word from the developers at Asobo in regards to this setting.

 

Thanks for writing this. 

What about ghosting differences between DLSS and TAA ?

On my 65” 4K tv I see a little ghosting when flying close to clouds, but with TAA I do not have it.

The weird thing is that on the 2x 55” with DLSS I see no ghosting.

 

 

5 minutes ago, GSalden said:

What about ghosting differences between DLSS and TAA ?

TAA and DLSS are both based on prior rendered frames where DLSS uses a more sophisticated algorithm  (which you can set, low, medium, high, etc.) to determine the current frame render.  So both will have some ghosting but that really depends on your speed of motion … high rates of motion (keep in mind this is relative motion within the virtual 3D scene) will increase the likelihood of ghosting.

I haven’t done a lot of experimentation with DLSS because even at it’s highest quality setting the image quality was not as good as SSAA and the FPS difference wasn’t significant enough for me to sacrifice the SSAA image quality.  

Using DLSS at it’s highest quality setting should have less ghosting than TAA.  But that will also depend on your hardware (GPU), how heavy of a scene/location in the sim, how much motion, and what you’re actual “native” FPS is.  Fortunately we have monitoring tools now that will show native FPS as well as DLSS FPS.  If you enable DLSS, your “native” FPS actually goes down in most cases (which is expected).

30 minutes ago, SayAgain said:

TAA and DLSS are both based on prior rendered frames where DLSS uses a more sophisticated algorithm  (which you can set, low, medium, high, etc.) to determine the current frame render.  So both will have some ghosting but that really depends on your speed of motion … high rates of motion (keep in mind this is relative motion within the virtual 3D scene) will increase the likelihood of ghosting.

I haven’t done a lot of experimentation with DLSS because even at it’s highest quality setting the image quality was not as good as SSAA and the FPS difference wasn’t significant enough for me to sacrifice the SSAA image quality.  

Using DLSS at it’s highest quality setting should have less ghosting than TAA.  But that will also depend on your hardware (GPU), how heavy of a scene/location in the sim, how much motion, and what you’re actual “native” FPS is.  Fortunately we have monitoring tools now that will show native FPS as well as DLSS FPS.  If you enable DLSS, your “native” FPS actually goes down in most cases (which is expected).

If we're talking about AA techniques, then DLSS shouldn't be part of the discussion... that's an upscaling solution.

DLAA is the equivalent term for anti-aliasing. It's based on previous frames like TAA but uses Nvidia AI and tensor cores on recent GPUs to optimize performance.

Things to look at with DLAA vs TAA are distant objects as the guy in the OP video shows, chain link fences shimmering, distant light poles (not getting rendered at all with TAA), etc.  And as I mentioned, and most people here realize, DLAA doesn't work properly with the HTML screens in glass cockpits resulting in less sharp text than TAA.

Edited August 13, 2025Aug 13 by Virtual-Chris

I’ve hidden three posts that contribute nothing to the discussion. If they continue I’ll lock the topic.

6 hours ago, Virtual-Chris said:

If we're talking about AA techniques, then DLSS shouldn't be part of the discussion... that's an upscaling solution.

I’ll disagree with you on that, wikipedia says it’s an AA technique … anything that involves upscaling is an AA technique.

https://en.m.wikipedia.org/wiki/Deep_Learning_Super_Sampling

DLAA (Deep Learning Anti-Aliasing) is by the name implies, also an anti-aliasing technique:

https://en.m.wikipedia.org/wiki/Deep_Learning_Anti-Aliasing

They are both classified as AA techniques just with different algorithms.

Agree with you that DLAA is better with distance objects or fences or poles or power lines etc.

3 hours ago, SayAgain said:

I’ll disagree with you on that, wikipedia says it’s an AA technique … anything that involves upscaling is an AA technique.

https://en.m.wikipedia.org/wiki/Deep_Learning_Super_Sampling

DLAA (Deep Learning Anti-Aliasing) is by the name implies, also an anti-aliasing technique:

https://en.m.wikipedia.org/wiki/Deep_Learning_Anti-Aliasing

They are both classified as AA techniques just with different algorithms.

Agree with you that DLAA is better with distance objects or fences or poles or power lines etc.

I guess you’re right, but when you run DLSS you’ve got a lot more going on that is going to influence the image quality than just the anti-aliasing. But good point 

Guy's, just try preset F...yes you lose clarity but ghosting is much better then with K. 

McDan out 

14 hours ago, SayAgain said:

Although not impossible, it would be highly unconventional to apply multiple AA passes using two different methods.

I interpret the Description differently from you “using more than 100% will Super Sample the rendering” … it doesn’t say “will ADD to…” or does not say “will combine…”.

Another odd issue with that graphics settings is why does the slider stop at 200 value?  When it comes to either TAA or SSAA a value of 200 has no practical meaning as applied to those AA processes.  

I’m guessing that 200 really means 2X SSAA based on the resolution value they calculate to the left of the setting?  Would be great if Asobo provided more information.

Here is a good reference to TAA: 

https://en.m.wikipedia.org/wiki/Temporal_anti-aliasing

Here is a good reference to SSAA:

https://en.m.wikipedia.org/wiki/Supersampling

As you can see, these are VERY different AA processes where TAA relies on prior frames to “smooth” out edges (jaggies) which as a result leads to ghosting on fast motion.

SSAA is a completely different (and more processing intensive) AA method that produces better images and is not subject to ghosting as it doesn’t rely on prior frames to compare against.  There is no ghosting with SSAA.

If you want to run a basic test, use TAA 100 and do some rapid motion view/camera or aircraft movement with objects relatively close (like clouds) … you will see ghosting.   Now enter values above 100 and you will NOT see ghosting when performing the same motion test.  This is why I believe the values above 100 are indeed “switching” to SSAA and not in combination with TAA.  

But, I would be interested to get some official word from the developers at Asobo in regards to this setting.

 

Just use 105 Render Scaling with or without TAA, there is an OBVIOUS difference in aliasing, especially visible on sharp edges (e.g. wings in external view), TAA is doing the heavy lifting there.

It is applied in two completely different stages of the render pipeline, Render Scaling affects the initial internal render resolution and TAA is a post-processing effect.

By the way, combining different AA methods has always been a thing, like MSAA+SGSSAA, and especially when one of the passes is a post-processing effect like FXAA or TAA.

Edited August 14, 2025Aug 14 by touchdown84

I've stopped watching these videos. It's nearly as dangerous as watching golf youtubers