Feedback Any plan for a better terrain system?

I have been on Unity for 5 years now, worked for many studios and I'm about to release my first game which is an open world.
All those projects had something in common, the terrain felt really heavy, and it's a very old system. Off course, there are solutions to optimize it like scene streaming, LODs, pixel error or simply converting everything into meshes, etc.

But with all the Dot/ECS stuff and Unity aiming for a more massive world and big content. When will we have a new terrain system that will be really performant? I've read a few things about Dots terrain but if it's real when we will see anything about it?

 

Yeah the terrain tools are subpar at best (compared to some of the asset store assets or unreal's terrain)
It would be cool to have a terrain system that won't require buying a lot of third party stuff just to make it decent.

 

This has been long over due.

 

I second it. Unity's terrain performance is one of the major bottlenecks at the moment.

 

Some points regarding the current terrain system:
- Terrain maps are pretty big in size, Tiny terrain did resolve this very well using lossy/lossles compression
- Custom culling solutions can't access terrain chunks the way umbra can, so terrain API seems to be limited in this way
- Limited textures number (currently using single texture instead of texture clusters, so only 4 textures per splatmap)
- Lack of performant LOD solution
- Lack of a performant details renderer (e.g. Vegetation studio/Nature renderer on the Assetstore)
- Lack of virtual texturing
- Lack of non-destructive terrain editing tools, (e.g. spline terrain editor tools in UE4, Microverse/Gaia on the Assetstore)

etc...

For the existing solutions on the Assetstore, some are impacted negatively by a limited/problematic integration wit Unity.
This affects development workflow, duration, and experience, and in some cases can suppress some important features.

So in the end the lack of an official complete and performant solution, isn't fully resolved, and this is affecting Unity's image as some voices are not recommending it for big games, for this reason

Hope to see some focus towards the terrain system, as it's an important part of modern games.
If Unity is working on a solution in silence, i would be glad to have some news

 

yeah, gonna echo what everyone else has said here. The Terrain system is quite antiquated. It works, but its a bit of a performance hog and its wildly outclassed by the terrain systems in other engines (EG Unreal).
A overhaul is pretty overdue at this point.

 

What I would like to see in 2023 (2023.2) for the terrain system in Unity ^ ^
https://forum.unity.com/threads/imp...ke-to-see-made-to-the-terrain-system.1362709/

 

https://forum.unity.com/threads/will-we-get-splines-in-new-environment-system.783467/

It's supposedly a thing since ~4 years but still no result at all, so I wouldn't hold my breath. Even if they announced it tomorrow, we would still have to wait 1-2 years for it to become a more-or-less stable & performant solution

 

For editing you can take a look at MicroVerse:

 

In fairness,the new SRP Terrain details renderer is the fastest grass renderer i have ever used in HDRP,
it's faster than Nature Renderer and GPUInstancer.

but it just needs to support LODs or separate culling distance for each detail,or it won't be good for Highend quality huge world because the performance will be wasted for unnecessary far distance small details.

This scene with SRP terrain detail renederer is running at 1440p 60fps on a 60w rtx3060maxq.
while GPUInstancer running at 52-55fps.
but with fine adjustment in GPUI,you coud get better performane by decreasing unnecessary details culling distance and detail shadow culling distance.

 

Even if things like this or the much needed NavMesh fixes are in our future, it’s still way too far ahead, multiple years kind of far. Unity has been too busy doing nonsensical stuff as well so I doubt we’ll ever see it. They’ll probably buy another AI start-up or smt and work on implementing that into whatever they can next.

 

It's good that we are all in agree about how bad Unity's Terrain really is.

If we all ask for a solution that works, maybe they are going to listen and make us happy in Unity 2023. Else I am not planning to use Unity >= 2022 at all, so they better have something big ready to go or they're runtime fee won't apply to anyone.

Onestly I think they should just hire Jason Booth to rework Unity's Terrain system, him alone would make the component 100 times more efficient, while also increasing the quality of it.

 

LOL, I feel your frustration. I'm working on building a mesh terrain prefab library so I don't even have to deal with that anymore.

 

Yea, for me it has been really frustrating, I put a lot of effort into building a good looking Terrain, which doesn't run anywhere near acceptable performance.

I wish I had known MicroSplat before...

 

They've only just hinted that they are working on better terrain despite the requests for it going back years.. maybe might actually see something by Unity 7

 

Yeah mesh terrains are the way to go unless you're building a small level and it doesn't really matter. There are plenty of conversion tools. Really figuring out your grass is the biggest thing to worry about and there are also many ways to work around that.

I've spent stupid amounts on assets trying to improve the terrain system performance. You can throw glitter on poop and make it a little prettier, but at the end of the day it's still poop (performance-wise compared to mesh terrains). It's not even close, there is only so much a texture array shader can do. Jason makes good shaders, but it's absolutely not even close to the best for making open world games that run fast in unity.

 

i have my hopes on this plugin, "Storm" terrain replacement:

https://www.procedural-worlds.com/products/enterprise/storm/
https://www.youtube.com/playlist?list=PLckEMv5tz0E5HLUIGOHYW7zbRVUzVdDXg

 

Nah, just use this https://assetstore.unity.com/packages/tools/terrain/terrain-to-mesh-195349 it's what the serious open world devs have all been using for many years. It's really simple stuff to get your world running fast, beware of over-engineered solutions.

 

Nah, I don't, I heard Adam talking about it, based on what and how he said it, it will be cost-prohibitive for any indie project. Since they are reserving this for AI and Meta-crap I expect the license will be real b2b priced, $1k+ and expect well into the thousands.
I have no doubt it is a great piece of tech and all, I generally love what Adam and Procedural Worlds do in general.

I'm sorry, but who are those "serious" open world devs?

 

https://assetstore.unity.com/packages/tools/terrain/world-streamer-2-176482

This is who I learned about mesh terrains from an earlier iteration of their first open world streaming asset.

Also, my own assets run faster than anybody else's on the store. Just to be clear, I hope we're not debating whether a mesh terrain runs faster than the unity terrain system. A mesh terrain performs more like a single cube than any system on top of a system for raw unity terrain.

 

I got World Streamer in 2016. What's your point? You said all serious open world developers. I asked who are those serious open world developers, and you're just avoiding answering whom you're talking about and start to rumble about your assets and Naturemanufacture. Just to be clear: I asked one question which I'm curious about: who are those serious open world developers. No less, no more.

 

Pro tip: Don't hold your breath waiting for Unity to do something official.

They've been doing an official overhaul on Terrain since 2014 and there have been some improvements but nothing super cutting edge. Just some basics added, perf improvements, stability, stuff like that. It's still mostly the same.

Interestingly, most people don't need wildly complex terrain tools and serious devs have plenty of resources so do it properly on their own anyway so it's not really a hot feature. Beginners on the other hand are a large vocal group and immediately assume that realistic terrain, foliage and infinite draw distance grass is the bare minimum for a simple game.

 

My own levels built from scratch and profiler stats speak for themselves. Right now one of my levels is being tested in a 32 v 32 VR multiplayer game. I already answered your question. Mesh terrains are best. You asked my point. I've repeated myself multiple times so please just stop with the saltiness.

This a thousand times. Thank you.

 

I did not. I asked who are the "serious open world developers", which appears you were talking about yourself, so you're the "serious open world developers". Is there a link to your released open world games? I would love to see them and learn.

 

Are we talking that I've executive produced or that my work is in?

 

Absolutely, I love talking about this stuff and sharing my research.

 

Dude is trolling me about stuff that has nothing to do with profiler stats and performance lol

 

The games industry seems to have moved away from height mapped terrains as they're not very efficient and it can look bad when assets like rock faces are simply placed on top of terrains as the intersection between terrain and the asset is often very obvious and unsightly.

The trend seems to be to create a world from mesh clusters where everything blends nicely together. But the problem with this approach is that it's very time consuming and so only the big studios can afford to use it.

So while terrains may no longer be suitable for big studios, the indies probably still need them. But are gamers still willing to purchase, play and appreciate terrain based games?.

 

I'm not so sure that studio size matters so much because any indie can build their terrains with any of the excellent additional tools available. But for optimization. there's really nothing keeping anybody from converting it to a mesh for a massive performance boost.

The Unity terrain system is awesome and I personally see no reason to change it.

There have been many successful games in the past and will be many more in the future that use the unity terrain system. Depending on the game it's not always necessary to do many optimizations. To answer your question, I'd say probably yes.

 

Anybody got any rendering bottlenecks they need help solving?

 

Got a video to share?

 

Any material or example publicly released is welcome by me.

I'm not sure whom you're talking about, but if it is me: I'm not trolling. You made a highly exaggerated claim about "serious open world developers" without any data backing your claim up, I asked for clarification which you're dragging your feet at. I'm genuinely curious what have you released that you praise yourself openly in this manner. Also, I am always open to learn.

 

I think that if they added something like this: In open worlds when creating many terrains and editing them with their respective mountains, valleys and etc... Add a button to generate low poly versions of those terrains and depending on how far away the camera is ( a long distance) the original terrains are hidden and the low poly versions of them are shown, and when you are even further away they are hidden, and vice versa when you get closer to a terrain it shows the original terrain and the low poly is hidden. The idea is to create a system similar to how the LODs groups system works but focused on terrain, I think something like this could minimally compete with the nanite or rather the world partition of the unreal engine

 

No I don't because I'm not the one testing it. There's literally tons of stuff available to the public like screenshots with stats on my assets linked in my profile. I was actually thinking about asking you if you wanted a key and you can see how fast my work is yourself.

The VR project that my customer informed me that they are testing my work in I just found out about this week and I haven't talked to them about sharing the details. I'm not the type of person who posts their resume online. There are plenty of examples available that prove any "wild" claims I've made. And I'm willing to help solve and rendering bottlenecks if possible.

Side note:

I'm ignoring Lurking-NInja. They straight up sound like they're dealing with some issues and I'm not going to be their punching bag for whatever their deal is. I get spiteful vibes from them and how they twist words around. Acting like I owe them money or something. There are so many reasons to not share any more info I have particularly with the tone in which they are addressing me. Everyone I know knows that it is industry standard to use meshes for terrains. If someone doesn't agree or believe, I simply don't care because that's their choice if they don't want to take 5 minutes to test something.

EDIT: Just one example. Lurking-Ninja asked me what my point was. I told them my point and then they said they never asked for my point.

Me: You asked my point (I explain again)

Yeah I'm good on that. Wow mate. Have a nice day!