Are clouds the best ever now? I'm impressed.

[Noel 4,501]

I had to reinstall Win 10 and lost all of my screenshots that I failed to save except a couple I had put on PostImg for linking--one of those is below.  The differences between then and now are in some ways subtle and mostly manifest with clouds at closer range.  I see some really ugly clouds at certain times now that never saw initially and I've never captured these because not pleasant to look at but I'm going to start to help illustrate where there is room for improvement. 

Squalls near my home near Denver CO.  Notice my old 2070 Super is running at 90% utilization, but still no trouble maintaining the requisite 30fps by vsync.  Very likely today this very same scene with the 3080Ti would be around 55% utilization--so there is clearly some headroom to exploit.

[sightseer 895]

2 hours ago, Krakin said:

People keep talking about downgrades now being present but they hardly ever show pics of when things were different

One thing that changed was the lighting or maybe just cloud shading when they introduced raymarching.  I thought it was supposed to make things better but now Im not so sure.

This pic shows two pics - the one on top is the new lighting and the one on the bottom is the original lighting.  Its a weather theme at same location and time etc.  The raymarching just makes things more harsh and took out detail on the face of the cloud.

I dont think voxel density or whatever actually changed.  I think they changed noise algorithms because people complained about pixellation and with raymarching we got blobby clouds.

Maybe the noise could be higher res.  but there is definitely a difference.

 

[HiFlyer 9,359]

31 minutes ago, sightseer said:

This pic shows two pics - the one on top is the new lighting and the one on the bottom is the original lighting.  Its a weather theme at same location and time etc.  The raymarching just makes things more harsh and took out detail on the face of the cloud.

I suspect that if you gave a location, somebody could very quickly fine tune a sun position and season to match that picture today.....

Or... maybe not.

It certainly would be an interesting project.

[sightseer 895]

19 minutes ago, HiFlyer said:

I suspect that if you gave a location, somebody could very quickly fine tune a sun position and season to match that picture today.....

Or... maybe not.

It certainly would be an interesting project.

Clearwater Florida. August. around 11am I think.  using a custom weather theme derived from 'Scattered Clouds". 

Can it be done without filters?  Something else that changed along the way was the 'fix' of SDR and HDR.  Usercfg used to say HDR 1 for me but now it says 0(last time I checked).  I tried changing it to 1 but it made no difference.  I think its the raymarching that makes the shading difference.

Edited August 2, 2022 by sightseer

[MDFlier 2,031]

5 hours ago, Cpt_Piett said:

The graininess is definitely infrequent. It seems to be related to lighting conditions, among others. I'm not that well versed in the technical bits of how clouds are actually rendered. But it seems to me that, every now and then, some clouds appear much more grainy than others. 

I'm not seeing graininess in SU10, but I have seen a comb like pattern of vertical lines in clouds, but only when looking out the side window of the airplane. I never see it in clouds in front of me. I saw it briefly on my last flight, and it reminds me of the artifacts that I've seen when taxiing past light poles and wind socks. Like ghosted vertical lines as you move by them. I don't think it is a problem with the clouds per se, I think it is related to the lighting.

But grainy clouds, I no longer see. I used to see some blocky edges to the clouds, but that went away a while ago. No idea why.

Edited August 2, 2022 by MDFlier

[Alvega 2,124]

On 7/31/2022 at 2:12 PM, Noel said:

 

Just horrid!  I wouldn't harp on this so much if there weren't already hardware resources to exploit which there most certainly are.  Give us the slider to impact cloud resolution in particular such that in the screenshot above you could dial out this low-density rendering, and fully, as the hardware permits.  Then continue to work on subtleties and cloud morphology, lighting and so forth. 

 

Don't know what's wrong with your sim, but even with max zoom looking at close clouds I don't see anything like this.

Here are 2 screenshots, second one is max zoom.

 

 

 

 

Another one just about to enter the cloud

 

Edited August 2, 2022 by Alvega

[Alvega 2,124]

Some people also say clouds are transparent. I'm inside that cloud, can't see anything to the sides, how are clouds transparent?

 

 

 

[Alvega 2,124]

Also, for the ones that say we shouldn't see the wingtips when inside the clouds, check this: https://youtu.be/lriTnm-E_o8?t=338

 

 

[asanosho 55]

It seems that the problem of cloud granularity is not related to the voxel density of clouds at all, as Noel incorrectly assumes it:

I looked at how the Unreal Engine 5 represents "volumetric" clouds. According to their documentation, "volumetric" clouds in their game engine, and presumably in all other game engines, are not true 3d particle systems that fill a given volume, as is the case in the major 3d rendering programs such as 3dsmax, Maya, and especially Houdini. Rather, they seem to be simple meshes or metablobs that have merely been given volumetric textures. This makes my argument of increasing render time with the third power of the number of voxels to be rendered moot.

MSFS does not use the unreal engine but the technique used to render the clouds should be very similar. So, the problem is not with the voxel density but with the volumetric textures applied to the cloud meshes and the parameters of the raymarching algorithm.

Whether and to what extent this shader network can be interfered with via sliders that the end user can operate is, however, beyond my knowledge. 
 

[Noel 4,501]

8 hours ago, Alvega said:

Don't know what's wrong with your sim, but even with max zoom looking at close clouds I don't see anything like this.

Here are 2 screenshots, second one is max zoom.

Not sure, but it's not only mine, and anyway that is not the point as I continue to emphasize.  I have views w/o this voxel grid appearance as well,. zoomed in, but as I say it's a tip-off to the resolution of clouds in MSFS.  Here is your screenshot without the voxel grid appearance--to me this is just looks blurry, looks essentially out of focus.

[March Hare 720]

11 minutes ago, asanosho said:

It seems that the problem of cloud granularity is not related to the voxel density of clouds at all, as Noel incorrectly assumes it:

I looked at how the Unreal Engine 5 represents "volumetric" clouds. According to their documentation, "volumetric" clouds in their game engine, and presumably in all other game engines, are not true 3d particle systems that fill a given volume, as is the case in the major 3d rendering programs such as 3dsmax, Maya, and especially Houdini. Rather, they seem to be simple meshes or metablobs that have merely been given volumetric textures. This makes my argument of increasing render time with the third power of the number of voxels to be rendered moot.

MSFS does not use the unreal engine but the technique used to render the clouds should be very similar. So, the problem is not with the voxel density but with the volumetric textures applied to the cloud meshes and the parameters of the raymarching algorithm.

Whether and to what extent this shader network can be interfered with via sliders that the end user can operate is, however, beyond my knowledge. 
 

I wouldn't use Unreal Engine 5 as a reference. It makes use of a common volumetric cloud rendering technique in games. But it isn't the only way.

In fact, Seb literally said himself in a live stream that the clouds are built from voxels, and indeed applying the same perlin or simplex (and other) noise techniques as UE5 uses in its method, but to a voxel density grid instead, is part of a volumetric cloud rendering method published in a research paper over 10 years ago, and that's probably closer to what Asobo are doing.

None of us can know exactly how they are created and we likely won't find out. Asobo's engine seems far more advanced than the off-the-shelf methods/engines already.

Edited August 2, 2022 by March Hare

[Noel 4,501]

8 minutes ago, asanosho said:

It seems that the problem of cloud granularity is not related to the voxel density of clouds at all, as Noel incorrectly assumes it:

According to others these are voxel grids and you can certainly discern the 'slices' they are made of.  You see this depending on angle, of course, as they are slices.  Here's a screen shot from a while back that illustrates this to some degree:

 

[Noel 4,501]

8 hours ago, Alvega said:

Also, for the ones that say we shouldn't see the wingtips when inside the clouds, check this: https://youtu.be/lriTnm-E_o8?t=338

To not see wingtips from point-blank range certainly can happen depending of course on cloud density.  I've ridden my motorcycle into a fog bank that gave me just that--about 20 feet of visibility that was it.  Whether or not MSFS cloud rendering can duplicate this kind of density remains to be seen.

Edited August 2, 2022 by Noel

[Keto Ketchup 240]

8 hours ago, Alvega said:

Also, for the ones that say we shouldn't see the wingtips when inside the clouds, check this: https://youtu.be/lriTnm-E_o8?t=338

For the ones that say we should see wingtips inside of clouds clear as day, check this:

I've been on enough commercial flights to know that this video is accurate. And I live in a mountainous region, and when we're literally "in the clouds", I often cannot clearly see a wing's length away walking around my own house. The foggy mist is right there in my face, something that I don't get from the current version of MSFS. In MSFS, I never truly feel "in" the clouds.

Edited August 2, 2022 by Keto Ketchup
clarity

[Noel 4,501]

9 minutes ago, Keto Ketchup said:

The foggy mist is right there in my face, something that I don't get from the current version of MSFS. In MSFS, I never truly feel "in" the clouds.

This current rendering model does not seem capable of delivering the kind of density required.  Doesn't matter if it's voxels or pixels it's simply low resolution, very low, and out of character with the rest of the sim.  Lest people who feel the need to come back in to defend against this contention I'm impressed with where it's come so far, can tell it's very difficult to budget performance resources and so forth, but as I say until the low resolution is improved we'll be stuck with something like this going forward.  I have ample GPU resources left to exploit, and hardware resources continue to improve steadily, so this does bode well for the future I feel.  I hope Asobo can implement a slider to improve density, or utilizes a better algorithm to increase density the closer one gets to clouds, as I have said in this thread it's a win win win for the ever-increasing hardware capability that is real.

Edited August 2, 2022 by Noel