Ghost shadow on the water

[Murmur 3,944]

56 minutes ago, efis007 said:

Sorry but I don't understand what the problem is.
Are you saying that without RTX it will be impossible to have artifact-free reflections?

RTX is not involved at all. Screen space reflection does not require RTX or ray tracing.

56 minutes ago, efis007 said:

There are "old" simulators that worked without Nvidia RTX, and the reflections of objects on the water were generated perfect and without artifacts.

As far as I understand it, screen space reflection is a newer technique used by recent flight simulators, which allows much better performance at the cost of some artifacts (like this one). The old technique did not have these artifacts, but had worse performance.

56 minutes ago, efis007 said:

So this bug is not excusable in the year 2023, they need to fix it. 😒

As I said, this artifact should be common to simulators using screen space reflections, and the competitor flight simulator has the same artifact. So I don't know if it will be fixed or simply accepted as a limitation of this technique.

Edited September 1, 2023 by Murmur

[blingthinger 649]

2 hours ago, efis007 said:

RTX

Oh come now. For someone so well-versed in CGI and cockpit lighting techniques as you, it's amusing to see you troll with RTX :)

Doesn't simzilla suffer from the same reflection derp?

[efis007 894]

I did some research and I confirm that MSFS has the bug too, this video will show it. 👍

On the MSFS forum I also read an explanation of how the SSR system bug works.

"This is NOT a bug, this is just how “Screen Space Reflections” work. It is already in the name: it can only refelct what is shown on the screen, so things obscured can’t be refelcted. The pro is that they can refelct the UNOBSCURED environment pixel perfect, the con is it can’t refelct the obscured environment.
Now there are two solutions to mitigate that and you wil see in a minute why Asobo decided to go with SSRs:

Possibility 1: Cube maps: to be accurate (and contain moving items) the engine needs to render the whole scene → this is very hardware intensive and thus cube maps usually have a much lower resolution or even omit some geometry. → not that accurate

Possibility 2: Ray tracing → that didn’t need much explanation … this can be done only with very high end hardware

Conclusion: So SSRs are currently the only middle way to get somewaht accurate reflections with not blowing up the computer".

This is the explanation they wrote on the MSFS forum.
Which is basically similar to what some users have written here.
I understood the explanation.

However I continue to have many doubts about this SSR technology. 🤔
On the MSFS forum they write that the SSR is to increase the framerate.
And I say "we discovered hot water!"
If part of the graphics is inhibited by the calculations it is obvious that the frames increase! but it seems to me a very old-fashioned way of gaining frames.
Someone on the MSFS forum calls SSR a "modern technology".
I call it a real limitation.
Truly "modern technology" should do the exact opposite, it should generate stunning graphics WITHOUT any viewer limitations.
If, on the other hand, I make limitations... for example I remove the reflections on the water (to increase the framerate)... or I remove the shadows of the clouds (to increase the framerate)... what kind of "modern technology" is it?! 😳
So at this point we remove everything, we also remove the realistic light and we will increase the framerate even more.

The second point I disagree with is the argument that SSR "increase the framerate".
Fortunately I can directly compare XP11 and XP12 in the same visual weather conditions and the exact same scenario every day, and I can guarantee that the framerate of XP11 is at least 3 times higher than XP12.
How is it possible?
If XP11's reflexes were so heavy to compute that they would blow up my PC (words written on the MSFS forum), why do my field tests confirm the exact opposite of this theory? 🤔
Why does XP11 Vulkan run blazing fast at 90fps on my hardware, and XP12 Vulkan run at 30fps?
If XP12's SSR technology is meant to increase framerate, why does XP12 run 3x slower than its namesake XP11 without SSR?
Apparently words do not follow actions. 

However one thing I think I understood: for XP12 it will not be easy to get rid of this bug.
He gave us new things, and made others worse.
Modern technology? 🙄
I don't think so... modern technology shouldn't work like this.

However I am confident in the developers, I hope they fix it. 🤞

[C2615 322]

Older sim basically render another mountain under the water, they can set to only reflect certain things like bare terrain and clouds and maybe user aircraft, you can actually select what you want to be "reflect" in some game's setting.

The performance cost could be relatively controlled by limiting how many things you reflect, bare terrain didn't cost much to render anyway.

more fancy way like one in XP11 is render a "environment map"- screen shot a spherical map without the aircraft, and reflect that image.

XP11 only update the "environment map" like every few minutes, if you taxi a shiny aircraft from near the gate to a remote taxiway, you may see it still reflects the terminal up close when you are already miles away.

Also environment map can have problem on get the reflection angle totally correct, in your screenshot of XP11, the reflected terrain actually isn't mathematically correct, but I know, it looks conversing enough. But when I approach at night to a coastal airport, I can see some lights from the city reflected from the water in a impossible angle (it somehow looks better at first glance. but still, mathematically not correct) .Another way to see the effect more clearly is just put a shiny airliner on a taxiway, move the camera around, and you'll noticed the yellow taxi line reflection under the belly move in a very strange way, both in XP11 and P3Dv4/5.

And it also can't reflect fine details, like seeing the wings reflected in a funny distorted (but realistic) way on the fuselage, that's relatively new effect you can only found on MSFS and XP12 (IDK if P3Dv6 got one).

Maybe some sort of mixed approach could bring better result, or maybe AI could save the day, I'm not a game or CG developer, hopefully they can came up something cleaver soon.

Edited September 1, 2023 by C2615

[blingthinger 649]

54 minutes ago, efis007 said:

I call it a real limitation.

I'm sure Elon would looove for you to tweet (to X?) these and other similar thoughts to him.

 

1 hour ago, efis007 said:

Why does XP11 Vulkan run blazing fast at 90fps on my hardware, and XP12 Vulkan run at 30fps?

I'm confused as to why you're confused. The update frequency and fidelity of reflections is now improved overall. This is possible because of SSR. Precisely because it's only having to calculate what's in the frame. In XP11's engine, asking the system to create the same reflection fidelity would turn into an intractable problem. Same goes for shadows and lighting and clouds. You either spend compute cycles calculating the actual view created by the geometry and lighting all around (and either light fire to your system or essentially demand full RT) or you spend cycles culling out surrounding geometry/lights while trying to decide which component to include in the render. SSR eliminates all of those calcs. Why isn't it faster as is? Because calculating real reflections is expensive. This is more a discussion for the Intel/AMD/NV forums. LR and Asobo are merely the victims.

And in 12.06, most folks are now very CPU limited because of the rendergraph. Sped up GPU side quite a bit. Now on to even more multithreading in 12... and more FPS.

That said, NV did recently introduce this:

https://developer.nvidia.com/blog/improve-shader-performance-and-in-game-frame-rates-with-shader-execution-reordering/

Shader execution ordering causes a lot of the backup in these calculations. If-then-else (and other branching functions) are currently a real roadblock for speedy rendering on the GPU side. SER looks to be a good way of actually using all the GPU compute cycles.

[efis007 894]

7 hours ago, blingthinger said:

I'm confused as to why you're confused. The update frequency and fidelity of reflections is now improved overall. This is possible because of SSR. Precisely because it's only having to calculate what's in the frame.

Forgive me, but in front of this image how can you write that the "reflexes are now overall improved"? 🤔

In the photo the reflections are all wrong.
- To the left of the image is an unrealistic blue reflection.
- There are strange white halos on the reflection of the main mountain.
- There is the horrible phantom shadow of the Cessna in the middle of the scene.
- And the reflections of the clouds on the water are also missing (the water was very calm, you should have seen the reflected clouds too).

Are you confused why I'm confused? 😊
Sorry, I'm the one getting more and more confused. 😁
What I have seen and photographed has no correlation with the words "reflexes are now overall improved".
I don't see any improved reflections in that XP12 image, I just see a lot of bugs and bad reflections.

I see the improved reflections in this other image: XP11.
Those are beautiful reflections! 👏

XP11's reflections will be old... they will be outdated... they will be unrealistic... but visually they are much nicer than what I have seen XP12 do above that lake....(moreover, the reflections of XP11 look very similar to reality).

When XP12 can get reflexes done right then my confusion will vanish and I will gladly accept the statement "reflexes are now overall improved".
But now I don't accept that sentence because it doesn't correspond to the truth.
The evidence is in the pictures. 🧐

Perhaps XP12 has focused more attention on the puddles on the squares and the reflections on the metal sheets.
However, a few puddles or shiny bodywork are not enough to assert that the reflections of XP12 are overall improved.
XP12 has actually improved some parts (the ambient lighting for example I really like), but it has ruined others (the reflections on the lake are bad compared to XP11).
He took one step forward and one step back.

However I am not making a criticism of XP12. 😉
MSFS also has the same flaw, I read about it on their forum and saw it on video (I don't have MSFS!).
So Xp12 is not the only sim to have that problem.
And this thing is curious.
It is curious to know that XP11 makes reflections on the lake better than the powerful MSFS and XP12. 🤭

[blingthinger 649]

3 minutes ago, efis007 said:

reflexes

You mean "reflections", correct?

Improved everywhere else. Higher fidelity, higher refresh rate, etc. In particular airport lighting at night. And it's not only reflections that were changed w/ screen-space. So even though I applaud your sleuthing screenshots and your:

4 minutes ago, efis007 said:

criticism of XP12,

I don't agree with your overall

5 minutes ago, efis007 said:

truth

that screen-space was the wrong direction to go. For me, the other benefits far outweigh the cost of having a derpy reflection from time to time, particularly given that I spend most of my time in the cockpit.

[efis007 894]

48 minutes ago, blingthinger said:

You mean "reflections", correct?

Affirmative. I mean the reflections of objects on the lake, in this specific case the reflections of mountains on the water without any graphic artifacts.
XP11 does it very well while also maintaining a very high framerate.
I wish XP12 did exactly the same.
It currently fails, so I modestly cannot accept the verdict that Xp12's reflections are nicer than XP11.

As I already wrote, if we talk about reflections on puddles, reflections on metal, PBR improvements and ambient lighting ... everything is ok, I agree that XP12 has taken a step forward in this area.
But if we talk about mountains reflected in the water currently XP12 is lower than XP11, so it has taken a step backwards on that specific sector.

 

49 minutes ago, blingthinger said:

I don't agree with your overall

You can agree or disagree, that's fine. 🧐
But I showed pictures.
Those photos prove that the lake reflections of XP12 are worse than XP11.
And they prove that XP12 creates artifacts in the scene where XP11 never creates them.
Those photos prove I didn't write any wrong words.
Words can be interpreted.
Words are just words.
But not the photos!
Photos prove things better than words.

 

 

Edited September 2, 2023 by efis007

[Pascal_LSGC 309]

7 hours ago, efis007 said:

But I showed pictures.
Those photos prove that the lake reflections of XP12 are worse than XP11.
And they prove that XP12 creates artifacts in the scene where XP11 never creates them.

No. Those screenshots don't prove anything. You cherry-picked the one case that doesn't work in XP12, and one case where it "works" in XP11 (and that's debatable, the perfect mirror effect is surreal), while ignoring the number of cases where thing are artefactually wrong in XP11 (typo intended).

In case you don't remember, XP11 has only one plane of reflection for the visible region. That means that when some lakes are at different altitudes, the reflections are truncated, and regularly jump as you progress. And the reflection are not affected by light scattering. And the 1 fps reflections on the plane, and the performance cost, etc...

So yes, the screen space reflections are a huge progress. Except for the one case that obsesses you.

Real time computer graphics are always a trade-off. But you should know that.

[C2615 322]

On 9/2/2023 at 4:04 PM, Pascal_LSGC said:

So yes, the screen space reflections are a huge progress. Except for the one case that obsesses you.

To be precise it's not the Screen space part make it better, it's the reflation part. and the screen space part is mainly make it fast enough but also to be blame for many of the strange artifacts.

[mSparks 2,094]

On 9/2/2023 at 1:00 AM, efis007 said:

XP11's reflections will be old... they will be outdated... they will be unrealistic... but visually they are much nicer than what I have seen XP12 do above that lake....(moreover, the reflections of XP11 look very similar to reality).

It is unfortunately correct that it is a limitation of SSR with no known fix or workaround.

XP11 uses cubemaps, different technology, different limitations. the main ones being cubemaps can't do any of the reflections in

The way cubemaps work is to render the entire scene multiple times for each reflection you want - so two different reflections = 50% less fps. This is opitimised by excluding a lot more than just a spot in water reflections where the aircraft obscures the "view".

Complaining about this one is very king Canute. SSR is better, its not perfect, the issue you highlighted has no known fix, so the only way it can get fixed is if you to create your own tech demo showing it fixed and sell it to the entire software industry.

___

While Raytrace is generally "raytracing everything" - you can see what that means for yourself by installing

https://store.steampowered.com/agecheck/app/1089130/

The concepts are gradually being adopted bit by bit for difference aspects of graphics engines, screen space reflections and volumetric clouds are two example of that bit by bit.

But until almost everyone has a graphics card capable of running Q2RTX at 200fps in 4K, a full raytrace pipeline is never going to be profitable.

Edited September 4, 2023 by mSparks

[efis007 894]

On 9/2/2023 at 10:04 AM, Pascal_LSGC said:

No. Those screenshots don't prove anything.

Those screenshots were taken with the same default Cessna aircraft, with the same camera angle, flying over the same lake (Swiss Geneve), with the same "calm water" setting to favor the reflections of the mountains and clouds.
Those screenshots weren't made "artfully" to detract from one simulator (XP12) and enhance another (XP11).
They were made to compare two "Xplanes" simulators in the exact same graphic/geographical situation and observe how they behave.

If you don't like my tests above the Swiss lake, please provide me another geographic location of another lake for further comparisons.
I will gladly compare the two simulators and show you (again) that XP12 SSR gets the reflections of objects on the water wrong, and also generates artifacts behind the aircraft. 

On 9/2/2023 at 10:04 AM, Pascal_LSGC said:

You cherry-picked the one case that doesn't work in XP12, and one case where it "works" in XP11

It's not absolutely true.
That day I was flying with XP12 over the Swiss Alps with the intention of looking for the famous "anomalous clouds" that had been reported by some users.
So the purpose of my flight was of a completely different nature: to find anomalous clouds and, if I found them, to photograph them.

Subsequently the weather became bad, I was flying too high, the visibility was poor, so I lowered the altitude with the plane.... and it was ONLY in that instant that I saw the bug of the reflections on the lake. 😳
That absurd white shadow of the Cessna projected on the scenery... I had never seen that thing with XP10 and 11.
How was it possible that the powerful XP12 generated those horrible reflections?
I knew nothing.
I didn't know that bug existed.
I didn't even know what SSR was. 🤔

I was happy, I thought I had discovered a new bug!! 💪🥇
Instead I discovered another sad truth: ALL the forum members already knew about the existence of this bug !!! 😄
Only I didn't know him!

Now.... you're telling me that that bug doesn't exist and that my evidence isn't evidence? 
You have a great imagination man.

On 9/2/2023 at 10:04 AM, Pascal_LSGC said:

and that's debatable, the perfect mirror effect is surreal

When the water of a lake is very calm, the mirror effect is easily generated in nature.
This effect is much more common than you think, especially in lakes in the morning and in the evening.
When the water of a lake is calm, the reflection of the surrounding environment on the water is so evident and defined as to make one lose orientation, it literally seems to see the world upside down.

On 9/2/2023 at 10:04 AM, Pascal_LSGC said:

while ignoring the number of cases where thing are artefactually wrong in XP11

I'm not ignoring anything.
I am well aware that XP11's reflex system is a compromise, and as such has some flaws compared to XP12.
But here in this topic I'm not talking about "accuracy of calculation", but about "compromise in the best visual aspect".

This photo below with this white shadow of the Cessna was calculated by the SSR.
And most likely the calculations are 100% accurate as well. 
But even if the calculations are 100% exact, the visual aspect is very ugly, it is not realistic at all, in nature the white shadows of the planes in front of the scene are not generated. 👎


In the XP11 photo instead, the reflection is probably not accurate... probably is not calculated with the same precision as the SSR... maybe it's even a totally invented reflection.
But the visual aspect of that reflection is more realistic than XP12, and above all it doesn't have those terrible white aircraft shadow artifacts caused by the SSR.
Therefore, between the two sim reflections, the visually more "realistic" one, (i.e. the one that comes closest to a TRUE photo of a lake's reflection) is undoubtedly that of XP11. 🏅

On 9/2/2023 at 10:04 AM, Pascal_LSGC said:

In case you don't remember, XP11 has only one plane of reflection for the visible region. That means that when some lakes are at different altitudes, the reflections are truncated, and regularly jump as you progress.

I will verify this theory in other flights with XP11 and 12.

On 9/2/2023 at 10:04 AM, Pascal_LSGC said:

And the reflection are not affected by light scattering.

I repeat, I don't care in the slightest how a reflection works and if it is influenced by scattering and other parameters.
I am only interested in the result I get visually !
At the moment XP12 loses the comparison on what concerns the environmental reflections on the lake.
Do you like seeing the ghostly silhouettes of a ghost plane on the water?
I honestly don't like.
Since XP12 came out, people keep talking about "realism realism realism realism" and then accept the wrong reflections? 🤔
 

On 9/2/2023 at 10:04 AM, Pascal_LSGC said:

And the 1 fps reflections on the plane, and the performance cost, etc...

Hey Sir, on my system XP11 runs at 90 fps with reflections set to max, and XP12 runs at 30 fps.
Until today the "performance costs" I have them only with XP12.

On 9/2/2023 at 10:04 AM, Pascal_LSGC said:

So yes, the screen space reflections are a huge progress.

So what do we say to all the XP12 and MSFS users (yes, them too) who complain about having reflection artifacts everywhere caused by the SSR?
Do we tell him it's a good thing?
Do we tell him this is progress?
Only after all those reflection bugs disappear from the scene we can loudly say it's progress.

On 9/2/2023 at 10:04 AM, Pascal_LSGC said:

Except for the one case that obsesses you.

Obsession? ... Me ?? 😄
I don't have any XP12 obsession. 
I have only shown photographs showing that XP11 does lake reflections better than XP12.
 

 

[mSparks 2,094]

1 hour ago, efis007 said:

white aircraft shadow artifacts caused by the SSR.

caused by only raytrace reflecting what is rendered to the screen, rather than the whole scene. Given

1 hour ago, efis007 said:

and XP12 runs at 30 fps.

You'd be running at 3fps if they did much more.....

Also

what GPU do you have? I'm getting 60fps in VR on an RTX3070 and pretty high settings, its absolutely sublime.

1 hour ago, efis007 said:

I honestly don't like.

Its difficult to say this without sounding insulting, but...

just in case you hadnt thought of it

Stay in the cockpit maybe?

1 hour ago, efis007 said:

and then accept the wrong reflections?

Lord, grant me the serenity to accept the things I cannot change

Courage to change the things I can

And the wisdom to know the difference.

1 hour ago, efis007 said:

Only after all those reflection bugs disappear from the scene we can loudly say it's progress.

Which is why RTX silicon is "game changing". The problem now is the widespread adoption of the silicon chips capable of doing it, not the software to do it. ETA 5 to 10 years.

At least there is an ETA now, back in 2018/2019 the most we had to look forward to now was maybe getting 60fps with existing graphics.

Edited September 6, 2023 by mSparks