After reading the technical guide about landscapes (https://docs.unrealengine.com/en-US/…ide/index.html), I am still struggling to fully understand the advantages/disadvantages of having a landscape with 1 section vs. 2x2 sections per component at same resolution. For example those to setups for 505x505 resolution. They are both the same size:
Why would I choose one over the other? They are both the same size/resolution. What is the advantage of having less components, but more sections. I kinda understand that with more components there are more draw calls. But what is the disadvantage of having less components but more sections (why not using that all the time). Most guides and tutorials never really explain why I would choose one over the other. They only always say something like, depends on the situation, your world, etc. But I still try get a grasp on why I want to choose one over the other.
Could someone please elaborate on this a little?
Thank you, highly appreciated.
So, I made some tests. I created two landscapes with the specs mentioned above using the same height map (just some hills, rather flat landscape with noise and hills, no real mountains). Both 505x505 resolution, one 8x8 components with 1x1 sections, the other 4x4 componets with 2x2 sections. I ran test with with empty landscape, landscape with landscape material applied, and landscape with landscape material applied using the Grass Output Node with very dense grass and flowers. For the test I looked at the Mesh Draw Calls.
So I noticed significant draw call reductions when using the 4x4 components. With 1x1 sections I had between 1000-2000 draw calls, with the 2x2 sections 950-1700 draw calls (depending on distance). Without the grass with only landscape materials applied it was 30-80 draw calls for 1x1 and 25-30 draw calls with 4x4.
So, now I understand why I would choose to use more sections and less components at same resolution. This would reduce the amount of draw calls quite a bit and leave more draw call budget for other stuff.
But what I still don’t understand is why would someone choose to use only one section per component. What is the downside of using more section and the upside of using less?
As per the technical guide: Each component has a render-thread CPU processing cost and each section is a draw call.
Did you measure the FPS?
I have done tests and using 1x1 gives more FPS than using 2x2, although it has more draw call, the truth is that this is confuse.
2x2 is always more expensive and has other drawbacks, like the size of the landscape section at LOD0 being larger (twice) the one for 1x1.
You need to bench on the end target machine.
The idea is to find the optimal compromise between needed size in KM^2, number of components and MS cost of rendering.
Or just don’t use Landscapes. That’ll fix performance right away