Map max size with world partition

Hi, I would like to understand why is it not possible to have a bigger grid than the max native one in unreal 5 (8km x 8km) when the partition system is supposed to solve the floating number issue.

Why is it not possible to create a 1000km x 1000km landscape in UE5 and then divide it in partition?

Does it mean that if I want a map of this size, I’ll have to create 1000/8, so 125 separate level that I would manage via level instancing? Would that work with multiplayer as well?

At last, I have read that the limit is 20 x 20km in UE4, any reason why it’s only 8 x 8km in UE5?

What would be the most practical way to create a big map for multiplayer?

So for those who might have the same question, I was told on a discord that the double numbers aren’t included in the early access, explaining why there is still a grid limitation to 8km. If the doubles are included in the final release, I assume the limitation for a landscape creation will be far greater that 8km.

Blockquote I’ll have to create 1000/8, so 125 separate level that I would manage via level instancing?

Yes :sweat_smile: , I think there’s a way to have low res LODs of separate levels (so you can see distant geometry), but I don’t know how to do that

I have dealt with 1444 or something tiles. It’s not fun.
It drags on performance because every tile is a call.

8k size probably drags on performance still in UE5 like it did before.

And that’s because instead of adding and managing more components, the landscape system makes the components bigger.
Leading to large areas of terrain to have really high vertex count.

I haven’t tried using ue5 because it’s worse than sh*t on sample and empty projects. There’s really no reason at all to use it.
But, I’d wager the situation is just worse with landscapes in it, in spite of the new mesh tris management system.

Yes, well, the old world composition system was built to do so.

Whatever they are doing in ue5 might not. Can’t really know. As I said before. It’s a useless vanity plate until they decide to fix performance.

“And that’s because instead of adding and managing more components, the landscape system makes the components bigger.
Leading to large areas of terrain to have really high vertex count.”

Hello, thanks for your feedback, could you elaborate on the quoted sentence? What do you mean by the landscape system makes the components bigger? If you are speaking about the cells, you can fine tune the size of each cells.

No. You cannot.
To reach 8Km^2 you are forced to use the maximum at 2X2.

I don’t see the square getting bigger when going from 1x1 to 2x2 so I don’t understand your sentence when you say that it doesn’t add more components but makes them bigger?

“The square” you are looking at probably isn’t a component.
Delete a landscape section. That is a component.

Well these squares are labelled as “section per components” and “number of components”. So I assume that’s what they are. When I increase those values, it increases the amount of components, not their size as you were claiming, so can you elaborate on that? Also what do you mean by delete a landscape section? How does this have anything to do with the component size?

It doesn’t have anything to do except for the fact you can only delete components. And you can see the actual size of them.

Sections/components are mislabeled in the UI.
The delete function calls them sections. The code calls them components.

The landscape section LOD reduction works on Components, which the map system calls sections.

I’m going to say it again, so maybe it sinks in.
The bigger the landscape size, the bigger the component size. The higher the number of components. The lower the performance.
Particularly when you sit at the intersection between 4 components and all of them are LOD0.

Of course the larger the map size is, the more components you’re going to have. Now are you saying that for the same map size, unreal 5 is less performant despite world partition? If yes, what’s the difference in percentage? I guess you had a look at the cpu, gpu and ram performance to get to this conclusion?

The landscape system in UE5 has not been changed from that of Ue4 as far as I’m aware of.

Ergo, the benchmarks done for ue4 are still relevant even if ue5 is it’s own beast with a worse rendering engine performance.

Yet, if the components weren’t also made bigger the system would probably perform more evenly.

Worth mentioning:
You have 2 options for 8km^2 maps.
One has incredibly large components, the other is increasing the size of the 4km^2 version.

They obviously perform differently.
You need to bench each in a worse case scenario.
(Corner of 4 components, in which the whole rest of the landscape components - fustrum culling - is visible).

Yes, but It’s not the only way.
Aside from being a logic issue too, it’s also a math problem you can solve. And there’s other implications you may need to do that for, like memory footprint.

Landscapes built with an Octree mesh system blow the sorry excuse unreal put togeter for landscapes out of the water.
Even the Voxel plug-ins do a better job at it really.

The only sort of downside (sort of. Because you can use procedural foliage and it’s better anyway) is that you cannot use the landscape grass Node unless you use the Landscape.

In 99.9% of all cases, the unreal landacape system is a fly trap that needs to be properly leveraged (for games).

  1. Create the landscape with layers and / no functionality. Limit to 4 layers per tile.
  2. paint and sculpt to taste.
  3. Bake the landscape down to mesh impostors.
  4. Move all the meshes out to a new world composition level.
  5. change all landscape meshes to use the same mater material instance, and provide functionality to the landscape:
    5.a) Pys Mat based on a color texture (export layers from original landscape to generate pack file).
    5.b) apply procedural meshes based on the same layers.
    5.c) create a world height blend texture to color match all landscape tiles based on world position.

The list goes on. There’s a lot to doing thing right…

Another problem is that, once your terrain is bigger than 8 km, the curvature of the planet is starting to actually matter, and Unreal is still very much a “there’s a single gravity vector and up vector” game engine.
Perhaps that will get better, too, over time – at least with Chaos, they have the code and CAN make it better, which really wasn’t a good idea to try to do on top of PhysX.

Actually that matters at 5k.

But unreal is a flat earth’s dream box.
There’s absolutely no built in functionality for sphere curvature approximation and the awful heightmap system is terrible at doing this as well.

To calculate the proper curvature in 120km^2 plus maps you have to take the correct version of the ESRI approximation and re-do all of your max height calculations based on the offset the sphere math provides.

If you do it correctly, then for small spaces the earth’s curvature can look somewhat right.

The further the 2 points, the larger the possible error.
Also, the more the distortion.

This is more of a GIS/archeology subject matter.
For games pretty much no one worries about it.

As always, the base of the problem is that we like to represent maps in a flat surface.
When some day we switch to Lidar point clouds this won’t be an issue…

Extra point:
Because world composition is essentially based on a single map to generate seamless transitions, the whole system has to be scrapped and re-done with Geocentricity in mind.
And I’m not positive but Google Earth may have some good code transitions for this.
It surely looks a lot more spot on than anything you can make with unreal and 20mDTM

Sorry but I am not convinced. I asked you clear questions about performances, it seems you didn’t do any benchmarking. Now you say UE5 is ■■■■ in matter of map performance, but it’s not worse than EU4. I guess you’re not a UE user then? What engine do you use then?

Then before you could even make a point on performance aspect, you now bring a supposed issue with curvature. Let me remind you than no matter how far you can see on a given earth place, you’re not supposed to perceive any visible curvature. I don’t see how a flat world would be an issue in EU.

So just to wrap it up. UE5 is still way performant than UE4 since world partition, will you admit? I don’t see how it could be the other way around. Earth curvature an issue? Please explain why with math since this discution is already all over the place, without any data to back up your claims.

Maybe since you have never done anything and others here have done things that are on the market - including Games - you should defer to their expert opinions instead of being antagonistcal like a novax or flat earther on FB…

It’s not like @jwatte or I go around making ■■■■ up, or get paid to take the time to answer questions like this.

Here’s your “poor man” explanation (Like for flat earther’s…):

Place a tower across an unrealistic earth desert without dunes or anything but flat.
Move far enough away and the tower will disappear.

This isn’t magic, the tower doesn’t suddenly disappear. It’s the earth’s curvature that makes it go out of view.

Now try doing the same in engine with the regular landscape.

And for the record, UE5 is sh*t across the board.
That’s partly because it’s in a development stage, and partly because epic DGAF about performance and optimization of the engine code.
This is a FACT that Epic has amply demonstrated by the changes between ue4.24 and .25, The ignoring of bug reports, and a ton of other BS you probably won’t be privy too being you don’t know the engine.

Last thing:
My benchmarks mean NOTHING to you - or anyone.
Every system is different. That’s why people don’t share benchmarks results, but they share a scene to test specific system performance with.

What you can do, depends on what you need to do it on.
And In the case of UE5 also at what native resolution.

Yeah, yeah…Long story short, you’re just full of sht.

Please, as you said I am a beginner, therefore do not waste my time with your subjective whining here. Credit isn’t something you beg for, that’s something you earn. So far you had no data to share whatsoever. Just your feelings. That behavior doesn’t earn you any credibility.

Maybe you are right, but you weren’t able to coherently explain it nor demonstrate it. You would look like a clown in any business meeting. You can’t blame people for not giving credit to someone who can’t back a single of his claim with data and who discard benchmarking and performance analysis as data driven evidence.

I would also point that a few large scale flight simulator were made on UE and they seem to work pretty decently. If you know any other somewhat free engine that gives you super performance on large scale and as much programming power and flexibility as UE, please share your knowledge with me, it will be more than welcome.

Also, I have no clue about what I am talking about, but I would suggest that reproducing the effect of a globe earth shouldn’t be that difficult. First of all, you could program a blueprint to change the perspective of a given level when you are at a distance. You’re tower could be lowered into the ground when you are not in that cell, so that it gives that horizon effect from far away. I am pretty sure that UE would also allow to play with rendering and camera effect, without having to move the object.

Secondo, I understand that you are definitely not a large scale guy on UE so I don’t really understand how earth curvature would affect you on a scale of less or equal than 5 x 5 from a visual perspective. After all, unless you are on the ocean, or on a super large scale flat plan like you don’t have so many in the world, you wouldn’t even be in a place to see that horizon effect.

It seems to me that your past frustration with a few projects on which you probably couldn’t overcome the challenges, might have obscured your vision. I had a look at your market place page and didn’t see any full game release there by the way…

Who is not supposed to perceive the curved horizon? The player of a pirate-on-the-ocean game? The player of Microsoft Flight Simulator, Or you, in your particular game?

If you’re OK with fully-flat maps, that’s totally OK for you! But we can’t, in general, assume that that’s an OK state of the world for “games,” in general, much less “3D visualization” which is where Unreal Engine is trying to make inroads.

Anyway, glad everything’s working great for you!

Something I’ve learned the hard way: If you don’t actually understand the math involved, it’s probably better to not propose something until you work it out on paper.

Of course I would prefer if UE allow a fully scaled model of earth and the solar system. I was just questioning whether it’s really an issue for the type of game most people make. Now the horizon effect start to be visible at 5km if you are standing at man height on a perfectly flat surface.

Usually the view distance of the player stop way before 5km for performance reasons or just because most gameplay don’t allow to engage other players at this distance. I have played a lot of different simulators and generally the curving effect you see at an altitude is modelized by the camera to give the “fish eye lens” effect. It never shocked me or put of off the game to not see the ground curving as you don’t see it curving either when you fly at 40 000 feet in a liner jet.

For the perspective blueprint, I don’t think there is any math to re invent. Those geometries formulas are pretty much all known by astrologists and I am sure you would find all the references you need on nasa’s website.