23 minutes ago, RXP said:

It is not vertical though: it you have a very elongated object horizontally, its bounding sphere will be much larger vertically than the object height, and therefore will be culled later than another object of the same height as the elongated one, but with a smaller bounding sphere thereof. 

It IS vertical, with vertical being the ratio between the object bounding sphere and its apparent size compared to the screen height. So, basically, when the Sphere is touching the screen Top and Bottom borders, the LOD value for that object will be 100. 

And yes, of course using a spherical-based calculation WILL penalize object with elongated shapes, but it's also probably the fastest method because, if the sim had to spend so much time to calculate a precise culling based on the object actual shape, it might as well just draw it..

Edited July 28, 20214 yr by virtuali

5 minutes ago, virtuali said:

This applies to lots of DEFAULT library objects and, not all of them are made as good as they could, for example the default taxi_light_02 object, which is very ubiquitous

Which confirms what I was suspecting, thank you for the confirmation:

1 hour ago, RXP said:

This also requires if I'm not mistaken that the modeler sets this size in the model properly.

4 minutes ago, virtuali said:

with vertical being the ration between the object bounding sphere and its apparent size compared to the screen height.

But the apparent size of a 10mx1m object with a bounding sphere of 10m, and the apparent size of a 1mx1m object with a bounding sphere of 1m, will make the 1mx1m object placed nearby disappearing before the 10mx1m object, even if both are 1m heigh and occupying the same amount of vertical pixels isn't it?

Now I understand authors could set their elongated objects bounding sphere closer to the less prominent axis so that it keeps displaying properly, but the problem is that it could end up with lots of discrepancies when mixing add-ons and objects, or from one airport author to another airport author. 

If even Asobo is not setting both light poles with similar LOD, is this by mistake, for artistic reasons, automatically set by the modelling tool by default and in this case a difference could be due because they didn't author the objects with the same 3D modelling tool version and there was a bug?

Visual consistency should require consistency in the LOD based culling and this is a very hard subject, but I believe there should be automatic and heuristic ways to do it by default and only let the models overriding this on purpose for the most complex shapes.

Edited July 28, 20214 yr by RXP

5 minutes ago, RXP said:

1 hour ago, RXP said:

This also requires if I'm not mistaken that the modeler sets this size in the model properly.

Sure but, assuming the object has been exported correctly, the sim won't even have to "calculate" its bounding box, because I think is included in the .GLTF file ( same as it was in the .MDL file ), that's why is the fastest approach.

And why modelers can even "cheat" the system, by adding extra small polygons, (usually underground so they can't be seen), which will trick the sim to assume they are larger than they really are, so they can escape the automatic LOD cleaver...

Edited July 28, 20214 yr by virtuali

3 minutes ago, RXP said:

But the apparent size of a 10mx1m object with a bounding sphere of 10m, and the apparent size of a 1mx1m object with a bounding sphere of 1m, will make the 1mx1m object placed nearby disappearing before the 10mx1m object, even if both are 1m heigh isn't it?

Of course. That's what I meant with "elongated object are penalized"

20 minutes ago, RXP said:

And this is certainly explaining to me what I've been reporting for quite some times already: trees disappearing then reappearing when you move toward them.

If this is the underlying reason, it would be because depending on the distance the game would be using different set of similar looking trees but with different bounding sphere radius and/or different W/H ratio, therefore culling the same looking yet different tree objects at different distances.

Hey brother. I know AVSIM as a commercial guy probably isn't your main course of rational but I just wanted to say I'm thoroughly enjoying your "science" behind why what is happening is happening. It's good reading despite completely out of my league.  

Edited July 28, 20214 yr by styckx

I don't know about you guys, but I have a feeling RXP and Virtuali could have fixed this issue in an afternoon if they had access to the code. 

And I'm not being sarcastic, I'm just impressed.😄

1 hour ago, ckyliu said:

Very droll. It's hardly a tweak, it's just a cfg setting to see if it improves a very specific problem "you have that no one else has mentioned." MFS needed a tweak before SU5 to make it perform better, it just didn't exist.

You stated that no one else has mentioned the reduced clarity in the DC 6 VC.. Here is a quote from the PMDG DC6 forum, that I will copy and paste. 

 

Unfortunately currently nothing can be done about the lower resolution in the VC. This is due to Asobo changes on how they handle the textures. We have inquired them but have not yet received any answer if this is fixable.
 

Edited July 28, 20214 yr by Bobsk8

3 minutes ago, Bobsk8 said:

Unfortunately currently nothing can be done about the lower resolution in the VC. This is due to Asobo changes on how they handle the textures. We have inquired them but have not yet received any answer if this is fixable.
 

Wow…. Now if that doesn’t contradict what Asobo said about not degrading visuals for Xbox, I don’t know what does… that’s disappointing. I noticed the VC in all aircraft look worse, along with the anti aliasing on the screens. 

Just now, pvupilot said:

Wow…. Now if that doesn’t contradict what Asobo said about not degrading visuals for Xbox, I don’t know what does… that’s disappointing. I noticed the VC in all aircraft look worse, along with the anti aliasing on the screens. 

Its like the old FS9 days when you tried to get better performance by rolling sliders to the left. The worse it looks, the better the FPS. That is how Asobo increase performance for their X box base, I'm guessing. If you took Su4 and moved all sliders to the left, you would see better FPS too. 

4 minutes ago, pvupilot said:

Wow…. Now if that doesn’t contradict what Asobo said about not degrading visuals for Xbox, I don’t know what does… that’s disappointing. I noticed the VC in all aircraft look worse, along with the anti aliasing on the screens. 

Yes the first thing I noticed is how the cockpits look worse in the new update.

4 minutes ago, Johnny Wyatt said:

Yes the first thing I noticed is how the cockpits look worse in the new update.

Yep. Everything looks sorta of.... Assassin Creed now. It's not quite dull, it's not quite detailed either.. It just sort of exists. 

Edited July 28, 20214 yr by styckx

12 minutes ago, styckx said:

Yep. Everything looks sorta of.... Assassin Creed now. It's not quite dull, it's not quite detailed either.. It just sort of exists. 

The DC6 VC gauges and lettering used to be crystal clear. Now you have to zoom in to accurately read the gauges. A real PIA. 

3 minutes ago, Bobsk8 said:

The DC6 VC gauges and lettering used to be crystal clear. Now you have to zoom in to accurately read the gauges. A real PIA. 

Yeah. When I first loaded it up I literally said out loud "Why is FXAA enabled now?" But nope..  My resolution is 4k but everything seems to be rendered at 720p. This sim used to be simply mesmerizing.. Now I don't even want to show it off. It's just.. Dull.. Dull is a great word.. Nothing stands out anymore. 

Edited July 28, 20214 yr by styckx

So can we assume if we increase our quality settings we can achieve the level of detail we had previously at the obvious expense of the "gained" performance from SU5?

I completely gave up on the sim until they fix the temp issue, until then I'm trying to come up with a plan to try and restore what they changed especially with visuals..

Edited July 28, 20214 yr by pilatuz