Official The road to 2021 - Q&A

In regards to multiplayer networking, are there any plans to create an example replay system?


Answer:

 

Hey Unity Team! I want to know that in future versions of unity will you please Add a Built in Dynamic clouds like Unreal. And Optimized volumetric lighting in Universal Render pipeline??? Please it will be helpful if you implement these features.


Answer:

 

Something that will (most likely!) give Unity performance by default but still being an easy engine to work with is an upgrade to .NET 5/Core. DOTS is super powerful sure but it's also quite restrictive for beginners. So if we can keep the traditional Unity with GameObjects and MonoBehaviours and still get good performance, that would be great. Many performance tests have shown that .NET 5 is becoming really fast.
So are there any strides towards ditching mono in favor of a complete .NET integration? Especially now when .NET is becoming so cross-platform.


Answer:

 

Thanks for the roadmap, guys!
Important question: Will URP get Realtime Global Illumination in 2021?
I need me my realtime light bouncing.

Answer:
https://forum.unity.com/threads/the-road-to-2021-q-a.950158/#post-6200858

Second (more definitive) Answer:
https://forum.unity.com/threads/the-road-to-2021-q-a.950158/#post-6202638

 

Will Environment System come in 2021? No mention of it in the blog post and it's a rather important addition imo.


Answer:

 

With the focus being on stability, current systems are being prioritized. Meaning we're working on terrain bug fixes and bringing terrain tools package out of preview

A recap what we've done lately (since having a terrain team):
2018.3 terrain performance improvements were shipped (many things from the CPU to GPU). 2019.1 introduced terrain tools package (blog)
2019.2 improved terrain tools package (blog)
2019.3 terrain holes were added to the core terrain system

Since then we've been focused on a new environment system. We did announce it too soon, and our release target was too ambitious. Recently (last couple weeks) we have pulled engineers off of that to bring current terrain into a better state. A better state for current terrain means bringing our bug count down, proper URP/HDRP support, and we're working on improving mobile performance. We're also working on getting terrain tools package out of preview - that work is bug fixing, some UX improvements, and moving APIs out of experimental. We're also hiring so once bugs are under control, there will still be a dedicated person for terrain (JD, should any of you be interested or know a good fit!). We want to bring current terrain to a stable state and keep it there - so we can work on a new modern system (as we can only facelift a system built ~13 years ago so much).

While engineers are focused on terrain stability, we do still have UX designers, technical artists, and some tools engineers working on the new system as we are committed to pushing it forward.

 

I would love to see wacom tablets fully supported out of the box with pressure and tilt across all the unity editor tools

• terrain texture
• terrain tree size
• terrain tree density
• decal painting
• texture painting
• prefab painting ( polybrush )
• etc

Any professial DCC tool has had this for ages and I really miss this in unity


Answer:

 

What, exactly, is mean by "Iterative improvement of the core experience for URP and stable HDRP. In the Asset Store, enabling users to discover URP and HDRP content easily through tagging and filtering. "

For instance, is MicroSplat compatible with URP? It is if it has the URP module, but not by default. Is the streams module? Again, there are dependencies here, due to the nature of SRP introducing massive compatibility nightmares for asset developers. Also, since it's not possible to ship things so they "just work" across render pipelines, the initial experience of downloading MicroSplat for an SRP (and most other assets) is pink shaders, which leads to a bunch of support requests essentially because SRPs were developed as if the asset store didn't exist. It seems like these things need to be addressed before we start rolling out filters, and guaranteeing to users that content works with their pipeline when the initial experience is of a broken asset.

Also, if you are admitting HDRP is not stable, then why is it being advertised as production ready? I would argue that it cannot be stable until each upgrade doesn't break half the shaders in the store.

 

I'm concern about HDRP default performance (300fps in an empty scene). There will be some improvement?


Answer:

 

Thanks for blog post and this topic

Regarding networking:
There are network transport solution that can handle 1K-2K connections. As I know new transport tested on 32 connections. Looks strange.

Can we expect Unity transport to be able to handle 1K-2K or more connections?


Answer:
https://forum.unity.com/threads/the-road-to-2021-q-a.950158/#post-6201200

 

Will Splines still be a core feature of 2021?


Answer:
https://forum.unity.com/threads/the-road-to-2021-q-a.950158/#post-6203361

 

Rendering pipelines

We never have ability to make custom culling solution into rendering pipeline, but many projects can benefit from it because they have specific camera and culling can be made faster and easier. Even more, we often need additional camera (now in camera stacking) and dont want to cull world again and waste 1ms of frame on mobile for each camera, we want to filter existing cull results or in many cases of UI we just need to draw all ui renderers without any culling and waste of time.

Various culling solutions in asset store have ugly integration into rendering pipeline. Having builtin interface for culling integration will make things way better.

So actually do Unity have plans to support custom culling and/or additional culling filtering for SRPs?

Optimisation
Unreal use 2 stages of image upscaling to achieve good performance. We now have camera stack and really want to render world on say 520p then upscale and render UI on 1080p. Better will be solution with few upscaling step like Unreal has.

Will SRPs support upscaling camera in camera stack or upscaling render feature?


Answer:
https://forum.unity.com/threads/the-road-to-2021-q-a.950158/page-2#post-6221019

 

Will SpeedTree 8 be finally supported in all pipelines? SpeedTree says Unity has to implement it. As a SpeedTree customer I'm at a loss here, purchased a lot that doesn't work at all. Just got an eMail about SpeedTree 9 Sneak Peek. The whole SpeedTree issue and non-existing support has become frustrating. Is there an official statement?


Answer:

 

Hi, thanks for opening this QA thread!

TLDR and main question:
Can you share some technical details and plans, if any, to speed up enter playmode with script changes (so script compilation times). Many, including myself, have asked before but often the answers have left me with more questions so it would be great to hear any more information you're able to provide or some sort of summary of your key plans in this area.

Background:
It appears there has been a ~40% compile time regression which appeared in 2019.3 with assetdatabase v2
This has been brought up in other places including this thread mentioning editor start time.

Recent insights and more discussion on compilation times:
Recently Joachim mentioned:

This sounds great, however, it confused me a bit; Does Unity recompile all it's packages at the moment every-time there is a script change...? Also, are there plans to take this further and only compile code that is changed rather than the entire assembly?

Something else I read, related:

Is this actually the case already for Unity packages, or is this coming in 2020.2? Does this require the reference assemblies feature Joachim mentions above?

Community insights and tests:
Unity compilation visualizer
From the conversation it sounds like Unity currently has a bug recompiling packages more often then it should. So I guess that will help lots when/if it's fixed.

Any more info appreciated! Echo previous comments that it was a good blog post.


Answer:

 

Workflow and Stability

For now we have strange workflow in few areas:
1. [SerializedReference] May be I miss something but as know it very unstable and using it is somewhat dangerous
If we serialize so object to it and then those objects classes will be refactored to different assembly or just change names then we loose everything. Do Unity have plans to support some way to keep everything work? may be [FormerlySerializedAs("OldNamespace.OldClassName")] on class?


Answer:

2. Prefabs. They change every time we go into play mode, they create overrides and dont allow to remove it (mostly in uGUI), they sometimes just ignore changes in base prefabs and we need to reimport all prefabs on dev machines or build machines, they allow to override any part of prefab, dont provide any control over it and this lead to various hard to find bugs. What steps Unity plan to do in this regard?


Answer:
https://forum.unity.com/threads/the-road-to-2021-q-a.950158/#post-6201203



3. Addressables/AssetBundles. Still has bad workflow and stability overall. Crash almost every build, so we need to rebuild all addressable groups from ground up before every build. There even Addressable importer asset to somewhat help. Default api to load bundle through AddressableReference load entire bundle into memory instead of loading only bundle header and load assets only when necessary. UI is just shame. Sorry.
In short we have some tools to work with assets in runtime but dont have any tools and workflow to understand what will be part of build and why.... What plans Unity have in asset management area?


Answer:
https://forum.unity.com/threads/the-road-to-2021-q-a.950158/#post-6201188

 

- What's in store for the realtime GI solution that was supposed to be 2021? Is it delayed, on hold, or even dropped for something else at a later time?

Referring to this -
https://forum.unity.com/threads/enlighten-deprecation-and-replacement-solution.697778/


Answer:
https://forum.unity.com/threads/the-road-to-2021-q-a.950158/#post-6202638


Another question.
- Any news on HDR Output for HDRP or even URP?

I know it's at a research phase for both SRPs the last time I checked the features table for both. I thought this would be somewhat closer to being implemented, especially with the upcoming consoles and HDR displays being more common.

 

Thanks unity I can clearly see the Bright future of you!

 

The Animator Controller is out dated! You should change the way of using animator controller or change the UI of it.

 

We switched to the HDRP and lost SpeedTree support will this be remedied?


Answer:

 

Hey @Jes28 !
Sorry you're experiencing these issues with Prefabs. It looks like various issues are mentioned, some sounds like misunderstanding others could be bugs. For us to plan to do anything about it, we need to know about the issue.

Can you please post in more detail in our forums? Thanks!

 

From the blog post:

I am wondering, which OSS multiplayer community are you referring to and which contributions/collaborations are planned?


Answer:

 

This is just last time I have some number on forums from Unity member. And it was claim that 32 connection fully tested and work and nothing more. Just want be sure what numbers was actually tested and work fine.

 

Will the Streaming Virtual Texturing package support UDIMs?


Answer:

 

What is the timeline for ECS API stabilization?


Answer:
https://forum.unity.com/threads/the-road-to-2021-q-a.950158/page-2#post-6204291

 

Hello!
I am happy to see QoL fixes becoming an important focus for the Unity team.

I have a question about the node based scripting system. We heavily use visual scripting (with Flow Canvas) in our projects. But it's not about being able to code via a visual tool. It's about providing our designers and artists with custom, high level nodes that allow them to do pretty much everything.
Some examples:
- A EntityGoto node. Two parameters: the destination and the entity. Few exits for this node: On destination reached, on destination near.
- A PlayAudio node. Allow to play sounds/musics with our custom system.
- A BeginDialogue node. Allow to start a dialogue.
- A EntityTrigger node. Triggered when an entity walk through a trigger.

This allow us to give all the tools needed to script all the game, without asking designers to understand how the systems works. They just say: ok, when the player's entity enter this trigger, I want this NPC to go to this point, then play a cinematic and start the next dialogue at the end of the cinematic, etc...
Having the ability to have custom nodes is the most important imo. Afaik, Bolt is not very friendly with custom nodes (if it is even possible).
Things I love with FC is how easy it is to add new nodes. Here are some random examples.




It's also easy to implement custom GUI/inspector for each node.
We also use the same idea for our dialogue system. Everything is custom with high level node.

So, the questions are: Will it be possible to create (easily) new nodes? If not, will it be easy to use the graph API to create our own custom visual scripting tool?

Thanks and keep up the good work!


Answer:

 

Hey. This is great to hear and the open source networking community would be better with even a tiny amount of help from Unity

I think there are exactly two things that you can do to help every single user & developer:

1. Fix Mono Networking Bugs. In particular, mono sockets are broken and async doesn't scale past around 100 CCU. Every single open source networking solution has to be built on a broken foundation which is Mono. Right now, we need to build native C transports for every single platform if we want MMO scale, since Mono's sockets are broken and there is nothing we can do about it. We can fix everything else, we can't fix mono.
   
2. Donate to support development. You guys don't have much time, obviously. Supporting the open source solutions would go a long way. Mirror & MLAPI both have Github sponsors set up, and both solutions took over all of UNET's users without any of UNET's funding.

Answers:
https://forum.unity.com/threads/the-road-to-2021-q-a.950158/#post-6201527
https://forum.unity.com/threads/the-road-to-2021-q-a.950158/#post-6201692

 

Currently SpeedTree 7 and SpeedTree 8 are supported in URP, and we're working on adding some additional lighting support (GI & shadows). For HDRP, this work is part of improving current terrain, to be done post bug fixes and stability.

 

We don't have a date for that work yet - right now the team is focused on terrain bugs and increasing stability. Once stability is there, we will be working on SpeedTree support for HDRP.

 

Currently the biggest problem I have at unity is the use of LODs, I would like there to be a way to generate the lods automatically to improve the workflow (For example through impostors). Is there a plan for this in the future within the SRPs?


Answer:
https://forum.unity.com/threads/the-road-to-2021-q-a.950158/page-2#post-6221037

 

Sadly DOTS is not on your 2020 or 2021 main list... so I guess it means 2022?

Better Roadmap & User Feedback
It's a little confusing at the moment...
The Roadmap page is a good idea, but I don't really trust it. DOTS is not there either. Looks like there's just one new feature for 2020.2, and everything else is for 2021 or later. The most recent Archive version is 5.6.0, where's 2017 to 2019?
I recently discovered the Graphics Roadmap, it feels like you want to fix the late Unity Ideas (which was also a good idea, but terribly broken). I really support that, any plans to merge both roadmaps?
Please make something like that for the DOTS ecosystem!
Also, the Roadmap main page should have all the Unite Roadmap update videos including the PDFs.


Answers:
https://forum.unity.com/threads/the-road-to-2021-q-a.950158/page-2#post-6204177
https://forum.unity.com/threads/the-road-to-2021-q-a.950158/page-2#post-6204291


Better Package Tags
I'm sure you all seen the feedback from the preview package visibility changes, and the main consensus was that if packages in development had better tags, like Experimental, Alpha, Beta, Released/Verified and Abandoned, there would be no need to hide anything.
Unity is known for listening to the community, have you listened?

And a quick rant... I keep trying to convince myself why I should not move to Unreal. Unity is morphing to a new Engine, since I have to learn a new unfinished Engine, why not learn the other that is in a finished state? The main reason is that I hate working on Windows, and I don't want to throw away years of intellectual investment. That article didn't give me too much excitement, the goal seems too far away yet. But I'm conting on Unity, I'm sure it will be the best again.


Answer:
https://forum.unity.com/threads/the-road-to-2021-q-a.950158/page-2#post-6204249

 

Providing the best possible .NET runtime performance is definitely an important focus area for us and we have done some initial experiments with CoreCLR, parts of which are detailed in this Porting the Unity Engine to .NET CoreCLR blog post by @xoofx. We do not have any concrete plans to announce at this time, as we are still in the process of assessing the benefits of migrating to CoreCLR for typical Unity workloads.

 

Now that the Environment System might not be in 2021 and Splines was a core part of it, might we see Splines being back-ported to 2020 (e. g. as package) or will they be bound to 2021?


Answer:

 

How should i request a feature? I Means do you have any page or Website where we can request a feature from unity?


Answer:
https://forum.unity.com/threads/the-road-to-2021-q-a.950158/page-2#post-6203634

 

Your experience with being left with more questions than answers is not an uncommon one. The script compilation and assembly reload process is complex and we do not provide much insight into it. We are trying to do better in this area and first step is the release of the Editor Iteration Profiler here on the forums.

One thing to note and that often confuses people is that after each successful script compilation we always perform an assembly reload. This leads the misunderstanding that script compilation is slow, while it is often the assembly reload where the majority of the time is spent. You can also see this with the Editor Iteration Profiler or if you look in the Editor.log where we log the script compilation and assembly reload times. I would say a typical scenario is that compilation time of a small assembly is between 0.1-0.3 seconds and afterwards the assembly reload is at least 2 seconds, some times much more, depending on the project.

One way to think about assembly reloads is that it is a complete shutdown and restart of all C# code and backup/restore of all serialized data in the editor, which includes both your projects C# code/data and the editors C# code/data. C# methods and callbacks in the project is invoked during this process and their performance is often the main culprit of long reload times. In the Configurable Enter Play Mode thread on the forums there is a diagram that provides an overview of this process.

We are actively working towards minimizing the impact our own features have on assembly reload time and we will increase our focus on this for the 2021 release and beyond. As well as providing better docs and best practices for how to keep assembly reload time to a minimum.

Unity does not recompile all packages when make a script changes. It recompiles the assemblies that have a dependency on the assembly that was modified, recursively. Since no package assemblies have a dependency on user assemblies, then they will not be recompiled.

Assemblies are the most fine-grained unit of compilation for C#. However, we run the C# compiler (Roslyn) in a server mode in Unity (VBCSCompiler.exe process) and it internally caches details about the assembly, which is why small assembly compilation is often only 0.1 - 0.3 seconds after the first compilation.

The reference assemblies feature in 2020.2 make it so that if an assembly is recompiled and it's public metadata does not change, then any assemblies that depend on it are not recompiled. What this means in practice is that if you only modify private fields/methods/etc. or make changes to a method body, then we will only recompile the assembly that was modified, as the changes you have made have no effect on the other assemblies that depend on it.

Please report a bug if you experience any issues with compilation.

 

Will there be approaches to not have meta files for every single file in the project? There are are other techniques like a common registry file that doesn't duplicate the total file count.


Answer:
https://forum.unity.com/threads/the-road-to-2021-q-a.950158/page-3#post-6225714

 

Thanks for clarify

You say that average target is 300 connection per session. it is absolutely OK for one session but how many sessions can be held on battle server side.
We already have around 1K-2K connections (16 connection per session) on old buggy UNET and our bottleneck is network. We event dont yet apply DOTS Jobs and Burst for server so it can handle may be 5-10K players but what about network?

Libraries on AssetStore like dotsnet claim support of thousands of connections and looks good for server workload


New UnityTransport claim support of only 300. May be this is 300 per session? What then total number for one server with hundreds of sessions?

P.S. In our case one server is one unity server process

 

You not fully covered question.

Bugs is mostly fixed as I understand you, and we just need to go to last LTS it is OK thanks

What about plan for exported properties on prefabs so users of prefabs can only change exported ones and can nor override other ones, so we Developers can focus on creating things and not fixing bugs of accidentally overriden properties?


Answer:
https://forum.unity.com/threads/the-road-to-2021-q-a.950158/page-3#post-6225777

 

Thanks for reply

Just tested last available package version dont see any valuable changes in ui and workflow

As I understand current Workflow of Addressables looks good for Unity and we will not see any progress in near future (prior 2022).

Am I right?


Answer:
https://forum.unity.com/threads/the-road-to-2021-q-a.950158/page-2#post-6217653