I think so, yes. From what I can tell, Objects LOD is restrictive to buildings and trees - stuff above the ground. Terrain LOD is everything at the ground - including ground textures and elevation mesh. I admit I haven’t played around much with these settings, so if someone else can explain better, feel free to chime in.
looks like it’s use maximum, with some percent out, because my 2080ti eat 10700 with 2.0 2.0 with 3440x1440 same all ultra
quality 0 = maximum
ah ok thx very much
Where is the airport you took your screenshot from?
innbruck airport, rwy 26
Ah, right. I should’ve known. Thanks, I’ll look at this tonight. Seems like a perfect example for LOD tests.
Mine is 16 for Graphics
Mine is 4 for GraphicsVR
“Preset Ultra” change this to " Preset Custom " without the quotes
Save and it should work on lod9
I did a few LOD tests at Innsbruck to highlight the visual differences and performance hit for each LOD step. You can see in this example, the visual difference between the default 2 and 4 is significant. The rendering distance of the trees and terrains is significantly increased on the mountain directly down the centerline. The difference between 4 and 6 is less noticeable on the mountain directly in front, but the more distant peaks and the relief directly to the right of the runway is better defined with more sharpness. At LOD at 9, the difference is barely noticeable, as any changes are probably well outside the viewing area. The difference at LOD at 9 is probably noticeable at altitude, but I’d argue the viewing distance is so far, you probably won’t even notice those additional trees or terrain rendering.
For the performance hit, I’d say about 3-4 fps hit per LOD step of 2. Surprisingly the VRAM usage doesn’t increase significantly until about LOD = 6. System memory increased for each step, but not really significantly until LOD = 6. I still maintain that for my system, the sweet spot is at LOD = 4 or 6 (I prefer 6). Others may have problems running beyond LOD = 2, but this type of setting is entirely personal preference of performance vs visual quality.
It should be worth noting that LOD is a radius defined around your aircraft - so a circle in simple terms. If you just assume the area of a circle (πr^2), each step of LOD (the radius) increases the area rendered in the circle exponentially by a factor of 2.
System specs: Ryzen 5950X, Asus TUF 3090, 64 GB 3600. Resolution set to 3840x1600, render scaling at 120%. Ultra preset (customized) with texture supersampling at 8x8, and both shadow sizes set to 2048.
LOD = 2:
LOD = 4:
LOD = 6:
LOD = 9:
Very succinct summary. Well done!
I completely agree.
Thanks for the data!
I would have thought it might be much higher at somewhere like London City where dense buildings are spreading out in all directions, rather than just 2 directions through the valley at Innsbruck. However from memory i think even at that location my numbers there were in a similar ballpark to yours (something like 9gb at 4 Lod)
Edit - I see you have similar in your new York screenshot in OP
That might be a function of rising terrain, meaning less drop off in quality than you would get with flat terrain. In other words more ground closer to you the more vertical the terrain is.
I agree, another aspect of the testing could be done at an altitude in a large city. I’d suspect the resource usage will be significantly higher than at Innsbruck, but might vary. The nice thing about Innsbruck is you can see the effect on the rendering distance relative to the elevation mesh and the trees in the same screenshot.
However, I fully acknowledge the mountains are potentially hiding some of the rendering at higher LODs.
I was one of those people between FS98 and FS2004 and now I can’t understand how I ever considered that “smooth” !
Back to the topic, we’re all getting very random results on VRAM because we’re all running this game at different settings. I only have a GTX 1070 and have been getting an average of 40+ fps with LOD4 and never exceeding 3/4 of my total VRAM. But then, I’m using custom settings between Mid-high and Ultra.
Yep, the settings matter here. I’ve maxed out all the settings I could to show the worst-case scenario. Clearly MSFS will utilize additional VRAM given the opportunity, but this seems to be most sensitive to high LOD settings.
However, I don’t think higher VRAM usage would translate into an actual measurable performance uplift. I think the only real benefit of higher VRAM would be reduced stuttering and less framerate spikes, as assets can be kept in VRAM and not offloaded to the higher-latency system memory.
And yes, 20 fps back in the FS2004 days was like a god-tier system rig…
… I would beg to differ on this. More RAM usage should translate into less data transfer (in certain cases) and that will always result in better performance. I tried to outline this in a different thread recently:
Now if “performance” means “FPS while maintaining the same, slow moving, camera context” then a RAM increase might have zero impact. But especially the switching between the aircraft and the drone can benefit dramatically from more (V)RAM usage.
Yes, I agree. This was the point I was trying to make in my last post. You’re effectively eliminating those framerate drops introduced by panning, but not necessarily having a measurable boost in FPS because higher VRAM usage.
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.