Unity is not worth using right now

You can place them on the terrain (manually, or using any automatic method you use currently) and allow GPUI to use the terrain placement data, OR you can populate positions in GPUI through code (which is more flexible and what I personally do).

 

You can even just paint using the terrain system, and then replace those trees positions with prefabs:
http://answers.unity.com/answers/1644322/view.html

 

So I openend Unity again after 5 years and was a bit overwhelmed with all the new options and increased complexity. I definitely agree with the sentiment of having different versions. However this is not something I wished only for Unity, but also other software. For example Photoshop, Blender or even some Office software has become much more complex (or bloated) over time and it's often tedious to get into it again after some pause. Because of that I wished there were always 3 versions:

a) a basic version for beginners and where you just get in quickly
b) an intermediate version with more advanced features
c) an expert version with everything in it

But yeah... naturally I understand that it's unlikely to have that because of the maintenance required. But it would be nice.

 

Actually this sounds like the kind of thing I was hoping DOTS would move into, moving DOTS processes onto the GPU and allowing for much higher core counts to unlock the processing power of our GPUs.

If you look at the new Mathematics libraries they are very similar in style to GPU shader code, I was hoping that DOTS would be able to work across the CPU and GPU unlocking the full potential of both for game developers, whilst the developers would only need to learn to use one programming language and API.

 

I agree a little in the fact that it does seem like Unity builds something out, runs into hurdles and dead ends and then due to scaling and performance has to completely backtrack to a different solution. However, this is true for any framework. To mitigate this I try to avoid being on the bleeding edge and adopt only out of necessity. Let others be the guinea pigs. Also, don't just update without backing your project up and fully testing. Make sure it supports features you need and does not deprecate others you rely on. If your game is running fine on an older version be wary of updating.

I do find at times Unity can be finicky. I have on numerous occasions beat my head and click finger against a wall for hours until a little Unity reboot magically fixes everything.

Overall, I very much enjoy Unity and it has brought game development to the masses. It is an ever-growing and amazing tool that consistently gets better. Yes, it has issues at times but it is a great engine.

 

If there was a heat map for Unity what would be your hot zones of problems, e.g. DOTS-ville, Terrain-heights, Mechanim-place, HDRP-boulevard, Burst casino...?

Note: budding level designers build a map of Unity...

 

Come on Unity, Unreal is dropping some amazing Terrain tools out of the box.

So if anyone wants to draw a fast Unity landscape for a heatmap maybe it would be quicker to user another toolchain...

 

Monobehaviour Metro, Prefab Province, Asset Store Arena, Singleton Slums

 

The problem with Unity terrain isn't the tools. In fact, they are 100% C# and there are many great extensions and replacements out there, like World Creator and Unity's own terrain tools package. The problem is the underlying tech, which doesn't scale to larger worlds without significant performance issues on consoles and mobile.

The terrain culling and rendering backend runs entirely on the main thread, including foliage and trees. The terrain uses a monolithic height/splat/normal map with no streaming capabilities whatsoever, so an instanced 8k terrain uses almost 1GB of RAM, not counting the actual textures to make it pretty. Non instanced will use almost 4x that since each vertex is unique.

Why would you use non-instanced terrain? Well, you have to if you want to use tessellation to get away with a lower res terrain, because surface shaders don't support instancing and tessellation at the same time. I'm not even sure if the SRPs support instanced terrain at all, being the incomplete messes they are right now.

When replacing the terrain with a plain old static mesh that has no LOD whatsoever runs significantly faster, you know there is something wrong.