I’m setting up a landscape, and I’ve determined what I think the LOD Distribution should be. It’s working fine, good fps.
But I have noticed a loss of detail in the landscape. Some of the hill slopes become angular and pointed. This is even more evident when I introduce PGC. At LODs 4 and 5, the PGC rocks are hovering over the less-detailed landscape.
I’ve tried pushing the LODs back, but I would have to push far, which will diminish the optimizations. Also adjusting the culling distance of the PGC rocks is an option, but I like the far view.
Another brute-force option is to just bury some rocks on those hilltops.
Does anyone know of a good solution for this?
FEATURE REQUEST: The ability to adjust the LOD Distribution of an individual Landscape Streaming Proxy in the Details Panel.
The only real solution for perfomance would be not to use landacapes.
You could convert your landscape to a mesh, reduce the polygon count analytically, place it in the level, and call it a day.
The performance cost would be next to insignificant compared to the landscape - inspite of the massive size of the mesh and the overall tris count it will bring to the scene.
You miss out on editing, and on the grass/paint layers.
For a mesh you have to first bake your material down, and then add grass as folige in a procedural foliage component.
As for your feature request.
Working on the landacape is like beating on a dead horse at this point.
It just doesn’t work, never has, and in all sincerity since they are adding Nanite and effectively converting it to a static mesh, it will become even more irrelevant.
Also, they wreked the systm in 4.18 when some asshat decided that the landscape has to work LODs based on screen size instead of Distance from player.
Because the size on screen of a component is variable based on the overall size of the landscape, there is no good distribution values to be had for maps 4km to 8km (^2) in size.
Being able to say “don’t LOD down this component” (your feature request) would be nice, but also impractical.
Delete the component from the landacape. Replace it with a mesh.
Goal achieved. Way better performance…
This is a good idea. More work than I had wanted to do though.
1 Like
I agree, but it still beats re-doing the whole thing. Particularly when the component probably has a more managable number of layers compared to the whole landscape.
1 Like
Actually, this looks like it’s got a mention in the “Looking Forward” Public Roadmap.
https://portal.productboard.com/epicgames/1-unreal-engine-public-roadmap/tabs/94-forward-looking
See HLOD in Editor.
If so, it may be an option.
I think the one that interests you is “next generation terrain”.
The problem with that is that its epic making it, so they’ll base it on the 30 year old system they have in place that shouldnt exist 
If they don’t, well then there’s a chanche to actually have something that works in the engine for a change…
Sadly their track record on this has always been to do the least possible to it and preserve the system instead of what they should do - scrap it.
Nanite brings a new/interesting approach to it because it lets them not scrap it, while also converting it entierly to a mesh that doesnt cost an arm and a leg to render meanwhile also allowing what they ought to have done 10 years ago to it - an analytically created mesh.
I hope im wrong and that they finally scrap it / start over with better underlying base ideas…
1 Like
