I know that we don’t know a lot of details about the upcoming Nanite tech yet, and that answering my question requires a lot of guess work and gut feeling, but I’m sure there are people who have a better grasp of what’s coming, so here goes:
I’m wondering what Nanite will mean for Landscape rendering. I’m working on a game requiring huge landscapes (think racing game in real world settings). One way to realize this would be to use World Composition with UE4 landscapes (although the origin shifting is tricky in multiplayer mode). But I’m now considering using high-poly meshes and Nanite instead. Since my players stick to roads, I don’t really need a uniformly detailed mesh, but could rather use high-poly areas for the terrain mesh close to the roads and fewer details farther away.
So the bottom line question ist: In your opinion, will Nanite also change the way we do landscape rendering? In what way?
I don’t believe so, Nanite uses tech similar to this demo. The UE5 tech demo was a best use case scenario, reusing a lot of the same highly detailed models.
Just how long is your race track?
I mean the world’s longest closed circuit is only 16 mile (Pescara IT). Even this should hardly need shifting.
If we are talking about offtrack/rally races, then 90% of what you need is the landscape, so your assetion that you dont need them becomes odd.
Either way Hopefully the new GI does bring some change to what we can do in real time. For both ray trace and deferred. However i doubt that things will cost Much less.
For that to happen the Unreal team would have to scrap the current landscape system and make a new one that isn’t so limiting and expensive to begin with.
since they just put a lot of time&effort into the new system with layers, i seriously doubt that any sort of change like what would be needed will be implemented anytime soon.
Also, in terms of reality and features lacking in 1:1 implementstions you can add Eart’s curvature to the list.
the current landscape implementation does not allow for spherical maps, or just curved maps that adapt based on character position like IRL.
You can fake this by using a spherical shader on the landscape, but then the landscape cost is even greater and collisions past 1km from you would always be wrong. Plants would float. Etc.
thanks guys for your input. I am talking about something more alike to rally tracks, where a map is around 50x50km in size with tracks per of about 30-60km. And of course I would need a “landscape” or terrain for this. But the standard UE4 landscape (or Unity terrain, or Cryengine terrain, or…) is heightmap based, with a uniform pixel/vertex density over the whole map (even the displayed vertices will vary due to LOD. The idea was therefore to use meshes instead, which I can optimize at design time since the possible points of view, viewing distances etc a known in advance (since the player is confined to the streets). But this still means streaming meshes with millions of polygons, and Nanite sounds like something that would fit this bill.
I’m also heavily considering using mesh based terrain via Nanite for our game.
It seems like a perfect fit and I’ve always disliked the heavy performance cost and of the Unreal Engine landscape system. If you want even a half decent looking landscape.
By far my biggest concern is: would there be visible seams where different terrain meshes meet?
For example if one terrain mesh is still at a lower resolution in the distance than a closer one.
This seem highly likely with the tech that’s being used for Nanite, right?
Since our terrain will only be something like 3x3km in size I’m hoping Nanite will be able to handle it, one way or another.
With a 50x50km terrain I could definitely see many more issues happening.