Landscape Love?

Will Landscape get any love any time soon?

  • Tools are all CPU based, and slow with larger brushes or landscape sizes
  • The edit layer system gets slow at larger sizes or higher layer counts
  • Blueprint Brushes seem to be only suited for smaller landscapes, and seem left behind in general
  • The landscape material layer node blending system seem out of date, especially when compared to the newer material layer asset system. Naming and grouping parameters for layers is often a mess.
  • Landscape is often just a hog in terms of rendering and memory performance.

In general, landscape just seems left behind. Its sluggish and just isn’t very pleasant to work with outside of materials.

1 Like

You don’t have to use Landscape Material Layers. You can simply use anything in the material editor instead. You can combine it with landscape painted masks, but you don’t have to :slight_smile:

But I agree in general that Landscape leaves a lot to be desired. Especially now in UE5, with tessellation being deprecated, virtual heighfield workflow being still clumsy and unfinished, the landscape resolution is just not on par with the rest of the next gen UE5 feature set and leaves much to be desired.

Almost entire UE5 is migrating to the next gen, high fidelity graphics, but with landscape, we still have the good old 1 meter / landscape texel resolution, which just doesn’t cut it these days even with help of normal maps:


and especially with either ray traced or virtual shadowmap shadows, which make it way more difficult to hide the low res topology of landscape up close.

There are some clumsy workaround such as scaling down the landscape, but then it requires many more landscape components to cover large areas, and the performance goes down the drain even on high end hardware.

It’d be nice to get some new nanite-style landscape tessellation which could get us at least 10CM resolution, instead of 1M, while maintaining at least comparable performance.

Yup… hoping someone at Epic can at least acknowledge there’s a future or some WIP.

Of course, but they could really make it more streamlined…

It’s been left behind for a decade.
Octree and a procedural mesh with max samples is probably better performing when custom coded by someone who just red this and is like.
Really? Lms.

The virtual textures had started off well, and died in the gutter unable to raster foliage at runtime into the material.

And now they probably got rid of tessellation, so 3/4 of the marketplace stuff won’t work anymore.
(Not that it’s a bad thing, tessellation was one of the leading causes for the performance being equivalent to the speed of a slug in a slug race.)

While at it, I’ll also add the fact that it’s completely flat as a factory defect.
Any world of any type has a depth falloff based on its sphere shape.
Usually at a distance of 5km you loose 2m.
By 10, you are down to 8m.

Meaning any Tower like object you place in game should sink into the terrain at a distance.
While if the setup curved the map this would happen without any thought put into it.
Yet they still push the “archviz” bs when one of the basic principles of architectural design cannot be analyzed within the engine without coding or making the map properly curved (and large enough for it) yourself.

Let’s also add the fact you have to do magic to edit it at runtime.
Yes you can. And it performs too. But it’s literally a month of figuring out how if not more. And no collision change.

2 Likes