LOD_RADIUS=15 but it's NOT about distance

[Guest]

I'd like to open a discussion about LOD_RADIUS and what it really means.  Motivation for this came about from using MSE Ultra City Boston.  I noticed what seemed like a much smaller radius of HiRes textures.  So I started down the road of trying to figure out why.  

 

All these years I thought the LOD_RADIUS setting was about distance from my view point or "size" of the area around my view point.  After some testing, I noticed that LOD_RADIUS is really about "LOAD" (texture data loads) and not about distance at all.  The "distance" is really a visual result of texture data loads for any given location ... so this implies the actual size of the cone that will produce best quality textures around one's viewpoint will vary based on texture data loads (the scenery BGLs).

 

So one could fly over an area with where you might get 24 miles of max res textures around your viewpoint, while in another area (like MSE Boston) you might get 2 miles of max res textures (all with the same LOD_RADIUS value).

 

For example, threads like this http://www.flightsim.com/vbfs/showthread.php?219354-Lod-radius-question indicates a "distance", one person suggested 39km for LOD_RADIUS 8.5.  I'm not trying to pick on anyone, because that's what I thought also.  However, it's not the case at all - distance is the result but the value does not mean distance, it's load limiter.

 

Also heard that LOD_RADIUS above a certain value is ignored ... by my testing, this is NOT correct.  See the series of images below and you can visually see the difference from LOD_RADIUS 4.5, 9.5, and 15 (I was able to hit 16, but 17 would result in an OOM prior to display).

 

 

 

 

 

 

 

 

 

 

 

 

 

If others have already discovered this and it's well known, then my apologies (I just never found this info anywhere with my extensive google search).  For me, this knowledge will help me plan my LOD's better and manage the many versions of my FSX.CFG, DLL.XML, and EXE.XML that I use for specific aircraft and flight plans.

 

Thoughts?  Input? 

 

Rob.

 

[simmerhead 885]

The easiest way to see the difference is in outside top-down view. There you can actually see which tiles are sharp around the aircraft, so I'd say it's about radius too. I read a blogpost about this from one of the ACES developers a while back, but I can't remember who (not Phil Taylor).

 

I've tried to go as far as 18 (unflyable), so high numbers aren't ignored as you say. You'll notice that when the flight loads. Going from 4.5 to 18 makes loading take forever.

[boez 85]

To get these images I changed the mipmap of every single spring texture in my texture folder to a different solid color. Where White = full 1024x1024 mipmap down to Dark Red = 64x64(IIRC). This allowed me to accurately tell just what mipmap is loaded and how far away from plane it is loaded.

 

These pictures below are from my Top Down view during spring at EGNT. As you will see from the attached images changing the LOD slider proportionately increases the diameter of each LOD ring which means that the area of the mipmap coverage gets larger at the outer rings (square law). In fact its hard to see in the zoomed out images just what effect the slider has on the inner rings so I've attached two more at a higher zoom level.

 

The conclusion is that 1024x1024 mipmaps (White) are only visible from the cockpit for a severely limited range.

 

Here's what LOD does: slider at Small

 

 

Large:

 

 

And zooming in on the center:

 

Small:

 

 

Large:

 

[odourboy 1,289]

Interesting thread - I'll be watching this. If I'm reading this right, Boez's findings don't conflict with Rob's but it appears that Rob has come across an added dimension - the the 'radius' isn't a distance radius, but a complexity one. For a given scenery set/location, increasing the LOD_Radius will give a consistent distance increase in the viewing radius. But that same LOD_Radius at a different location may provide a larger or smaller distance radius, depending on the complexity of the local scenery. Is this a correct interpretation?

[skwaL 36]

OOM chance ------ small ..................................... large  :lol:

[boez 85]

@odourboy. Another way of thinking about it is that the lod slider simply controls two things...the overall area or number of tiles that has its scenery bitmap loaded from disk (instead of the generic 'globe' texture) AND the mipmap that is used to display that bitmap. If I repeat the test at more 'complex' areas the same set of rings appear. Note also that the autogen slider has zero effect on this pattern (as expected).

 

 

I can post some looking from the cockpit later.

 

Sent from my HTC One S using Tapatalk now Free

 

 

[boez 85]

This is Plockton in Scotland which is good hilly terrain to show the differences that the LOD slider makes to the cockpit view. For the Small radius even quite close hills are already blue (at a low 128x128) while the Large radius keeps higher resolution mips (Magenta) at the same point and so of course will look more detailed when going back to the standard bitmaps.

 

Small

 

 

and from cockpit:

 

Large:

 

 

[boez 85]

Finally just on a similar topic here are a set of pictures which show the absolute best image detail that a 1024x1024 texture file can produce.

 

This is a single pixel set black in the white bitmap and how that looks in the simulator both with no and 8x anisotropic filtering. Note how big a single pixel looks on the ground!

 

This is why even on the best PCs available ground textures at low altitude can still appear blurred. Nothing to do with LOD or mips but the limited tile resolution.

 

 

 

 

 

 

 

[Howellerman 104]

This is *really* interesting, particularly what Boez (only 7 posts? I am glad you spoke up here!) did with the colorization. And now I don't feel so bad about my rig producing blurry ground textures below a 1,000 feet!

 

Per simmerhead's comment, did either of you (Rob and Boez) "time" the differences in LOD loads? I also wonder if this would lead to big-time stutters if the sim is struggling to load distant terrain maps at the higher LOD settings.

 

Finally, and this was noted by Boez, Rob's first set of images showed no visible change on autogen features. I had always thought LOD was related to autogen "popping", but apparently not.

 

Interesting!

[simmerhead 885]

Nope, makes no difference on popping in the far distance.

 

Higher LOD_RADIUS is unflyable as it will fill your VAS pretty fast. An ACES member explained the LOD algorithm (how many tiles are loaded at once), but I don't have time now to find the link. Many of the FSX devs used to blog about these things, and also shared quite a bit of info here on AVSIM before the trolls took over and sushed them away.

[ellipirelli 1]

Very interesting topic, one question tough, can somebody please elaborate on LOD slider. I thought LOD_radius was changed by editing the fsx.cfg only.

[Guest]

 

 

But that same LOD_Radius at a different location may provide a larger or smaller distance radius, depending on the complexity of the local scenery. Is this a correct interpretation?

 

Yes.  It's about memory allocation.

 

 

 

This is why even on the best PCs available ground textures at low altitude can still appear blurred. Nothing to do with LOD or mips but the limited tile resolution.

 

Impressive work!  That really demonstrates LOD_RADIUS and confirms my theory in well defined test.

 

 

If I repeat the test at more 'complex' areas the same set of rings appear. Note also that the autogen slider has zero effect on this pattern (as expected).

 

And yes again.

 

 

 

For the Small radius even quite close hills are already blue (at a low 128x128) while the Large radius keeps higher resolution mips (Magenta)

 

Exactly!  

 

 

To get these images I changed the mipmap of every single spring texture in my texture folder to a different solid color.

 

However, one additional variable will be compression of the texture by the graphics hardware ... is this a factor?  No matter what I do, I can never seem to exceed 1.6GB to 1.8GB GPU memory, I will OOM before I can reach 2GB.  So in the testing you did, do you have any numbers of actual VAS and GPU allocation?  I would expect that solid colors would compress down considerably reducing the data allocation (certainly on the GPU side) and hence providing a large "visual" radius of high to lower MIPs.  I don't think this invalidates your testing, it just might make LOD_RADIUS larger than it actually would be if you had the smaller MIPS rather than solid colors?

 

Thank you Boez, your contribution has been invaluable.

 

Rob

[pmanhart 176]

While using such a large LOD_Radius for photoreal scenery works, it will always kill your FPS and make OOMs and CTDs much more likely if you are using any Autogen, especially in areas where the autogen count has been increased over FSX default (such as UTX or Orbx Region scenery).

 

Each of those tiles that are loaded have a certain number of autogen that load with them, and with a large LOD_Radius passing 6.5, in even a less-dense urban area you can have upwards of 600,000+ autogen objects fighting your CPU to display.

 

A good rule of thumb is for every one increase of the LOD_Radius value above FSX large (4.5), you should move your autogen slider down one notch (i.e. when increasing the LOD_Radius from 4.5 to 6.5, lower your Autogen Density slider two notches). Even following this basic rule though, does not trend correctly, and you may have to reduce your Autogen Density slider even more.

[simmerhead 885]

Very interesting topic, one question tough, can somebody please elaborate on LOD slider. I thought LOD_radius was changed by editing the fsx.cfg only.

 

It can be changed inside the FSX display settings menu, but maximum will be 4.5. You have to manual adjust it in fsx.cfg to make it higher.

 

In the settings/display/scenery menu it think it's called Level of Detail radius, and can be adjusted to small, medium and large.

 

 

And a big thank you to Boez for that very detailed and nice series of explenations!

[odourboy 1,289]

Note also that the autogen slider has zero effect on this pattern (as expected).

 

Yes, but does the LOD_Radius setting have an effect on Autogen as pmanhart is claiming - or are they completely independent?