DOTS Netcode Copenhagen updates

I did a game jam last weekend and was trying out https://github.com/Unity-Technologies/multiplayer but soon realized it is 3 months old. A lot of the stuff demoed in Copenhagen is there, but there are some critical pieces that were different that I couldn't exactly work out.

Any rough estimate for when we could see the fall update to the repo? No rush, just checking in.

DOTS is really coming together nicely now that all the modules are ramping up.

 

Does this next DOTS Netcode update come with DOTS sample already?

 

Any news?

 

Updates?

 

Do you guys know what ~ means?

because I dont! why is there no Animation DOTS in Package Manager window yet?!

 

Hopefully it won’t break too many of the current implementations

 

https://github.com/Unity-Technologies/DOTS-training-samples/blob/larus-labrat-netcode/Ported/larus/LabRat/Packages/manifest.json

 Access to 'https://artifactory.prd.cds.internal.unity3d.com/artifactory/api/npm/upm-candidates/com.unity.burst' was denied

 

It's exciting, patiently waiting.

@pal_trefall Nice try

 

DOTS survey

What version of the Entities package are you using?
<0.1.0
0.1.0
0.1.1
0.2

Yes please

 

I'm impatient and yes, it's VERY exciting. I'm doing VR with it!

 

Along with the new beta release (2019.3.0b11), I would have thought that the new packages would be released, this does not seem to be the case. And as such, does anyone have a clue as to when to expect to get our hands on this?

I really can't wait!

 

I want them to get it right because I want to base production code on it. I have waited quite a few years for Unity to get to a point where they can support my business goals.

If it means I wait another month even, so be it.

I've supported Unity since start, and I have faith this is where it all lands. This netcode + DOTS takes a lot of talent and engineering to pull off so I'm very excited.

You are not alone!

 

Lets organize a network stress test in here once this launches!

This online third person shooter game will be a pretty good proof-of-usability for DOTS. If we can have 100 online players, all with animated meshes, and the framerate stays over 60 on a decent-but-not-great gaming PC, and the networking holds up... then we can all rejoice

And I'm guessing there's still a lot more optimizations coming to DOTS in the following months/years so it'll only improve with time

 

After watching the Unite video, I gave up on my hope. Here are the reasons.

1. Delivering something a year late and it still looks like a mess. I'm not even sure where they are headed.
2. Unity's track record tells me that it's not going to be much different in the near future.
3. The API looks so convoluted and low-level that I don't think anyone can wrap their head around easily.
4. DOTS is still in fluke and the Netcode is in the same boat.
5. Even if everything is perfect, your game needs to be DOTS to take advantage of it. And I'm not sure DOTS conversion is possible.

My take is that I'll give Unity another 5 years and if they don't abandon it by that time, then I'll consider.
Good luck and there goes my hope, again. Unity is Unity and they never failed to disappoint me.

 

I'm gonna speculate, but I think the whole reason why it's been hard for Unity to give us good tools in the past is because they were stuck with tremendous technical debt and they had millions of users that would complain if any breaking changes happened

I don't think there's any realistic way to salvage the monobehaviour workflow and make it good for high-grade games at this point.

DOTS and SRPs are the solution for this. A completely new tech stack that throws all the flawed legacy stuff out of the window and is being made by a team that has a bunch of AAA game industry vets in it (Mike Acton, Andreas Fredriksson, Tim Johansson, and Sebastien Lagarde to name a few). These are people who worked on Frostbite (Battlefield) and Insomniac Games' engine

At the very least, we can say it's in very good hands

 

The plan is always nice and I was sold DOTS promise, but I cannot say the same about their execution.
I always say I'll be happy to be proven wrong but it hasn't happened yet.

 

LTS is about keeping people supported, yet Unity still fails to deprecate stuff properly. The rampant mess of XR in Unity is clear evidence of this. Unity should really rely on LTS and customers need to learn what those letters mean. What's the point of long term support if nobody uses it then complains when things change?

Unity should be a bit braver with changes I think.

Bit off topic but I get what you mean.

 

Built-in renderer, enlighten and old lighting system should be removed and URP replaced as default.
HDRP should be made as extension of URP, if needed.
Old scripting workflow should remain the same while keeping DOTS as optional for learning project.

LTS release should be made modular. If I need to use SRP v8 in 2018.4LTS, I should be able to do it.

 

‘3. The API looks so convoluted and low-level that I don't think anyone can wrap their head around easily. ’

On the contrary. Because it is so low level, you can have such huge performance.

Let those who can’t wrap their head around easily stick to the classic way. Not everyone can make extremely performant, unique game. Not everything is meant for everyone

 

But also,

I'd be curious to hear what you didn't like about what was shown

To me it looks like the lag-compensated, predicted, interpolated, cheat-proof, highly efficient online framework that we've all been waiting for for a very very long time. Unlike UNET, Photon, Mirror & others which gave us quick starts for games with simple needs, but reached their limits really quickly

 

Yes and they'll have Vivox and a BAAS so that even my crappy little game will be able to support player storage and accessibility. I only hope the Unity gods will help me get this going cost effective though

 

From my private experience the URP is not really ready for Production it presents a lot of bugs.
Weird rendering not supporting all devices,

in my case: i have a different bug on each device more details here

 

To be fair though, DOTS subscenes and hybrid renderer aren't supposed to be production ready yet

 

As a matter of fact, I like ECS and that's why I had waited so long. The problem is that Unity cannot deliver on its promise in a timely manner. They promised "Transparency" but it's just speaking from the bottom orifice. I don't think they even have design specs written down yet. They show some stuff but right now it's too low-level and it involves too much plumbing. Waiting for a year is long enough and I don't think I can trust Unity to deliver on its promise however I have to work on schedule, unfortunately.

 

Uh dude we're done here. DOTS is in no way advertised to be finished. In fact I thought it was supposed to be done in 2023. If you want some strong stability look for 2018.4 LTS - any professional would.

 

I think you are just trolling, as you have just given vague accusations of it being unworkable but it sounds more like someone who is just watching from the sidelines and waiting for a final release and getting impatient rather than someone who has attempted to learn it.

 

I don't think he's trolling. He just seem really frustrated. The biggest problem here is that Unity is increasingly highlighting/promoting tech in preview or experimental status (past few Unites and GDCs are full of them). I'm not saying it's 100% a bad thing. But it is the current trend and has its downsides. One of which is frail user expectations.

The upcoming Netcode should be fairly limited at the beginning too. For example, according to one of the talks, they are only doing ReliableOrdered packets now (Sequenced or Unreliable are not supported yet). But that's fine since it's in preview and all.

What is not fine is taking the LWRP/Universal RP out of preview without realtime point light shadows (okay I'll stop myself right there before going into a rant).

 

I mean they do like a million Unites per year, so they obviously have to show stuff that isn’t ready, or even exist at all.

But, hear me out, there is a solution to all this. Unity needs to realize they should go ahead and live their best life and turn into a party planning company and be finally freed from all those preview features that will forever be preview features even if the preview tag is removed.

With all that said, the netcode talk seemed promising to me, even if the majority of the audience seemed to be disappointed.

 

I try to only complain about something when I have a better solution to suggest.

What's the alternative here exactly? You want Unity to not develop new tech, just because it's taking longer than expected? You want to stick to the old tech stack forever? You think Unet and Photon are fine for esport? You think you can develop DOTS faster than Unity does?

If you complain, then you must give me a clear solution. And "have better time estimates" is not a solution, because anyone that had to do big software estimates in their life would know that it's impossible to make this an exact science, or even to make it semi-accurate. They're just doing what they can

 

@PhilSA He must not. Any solution can be developed on top of ECS / Job System in less than 1/2 year. Delta snapshots or eventual consistency, you choose. Their C implementation of the transport is their bottleneck as well as scheduling a job to dispatch events.
You will have your game running at 30 fps hence the networking at 30hz because of the main thread sync point. There is no proper I/O, their replication model is poor and single threaded. Unity failed to deliver their promises many times: RakNet, UNET, now DOTS. On presentation, he runs 4 test clients and his server loop takes 25ms, do not expect 100 clients at 60 fps.

For any experienced user new DOTS network will be no more than just a marketing strategy.

Cheers.

 

And how exactly do you know that this is the final product?

Can you seriously tell me with a straight face that unity will ship a networking solution that runs at 30 fps for 4 players?

 

I do not think they will ever ship it.

 

Moderation is sadly in effect. Replies must be on topic here on out, sorry guys! (you can talk about DOTS networking, not argue with each other)

 

I wouldn't get so hung up on performance.
I've reached 60 fps even with the FPS demo back then and 10 players. Above that, for 15+ players in the same channel it started to be a problem. Biggest problem was the bandwidth and this got better in the current netcode where I have no issues with performance at all. Bandwidth will always be a problem, it's the nature of the beast when developing multiplayer.
But I feel like some are expecting to make a new Fortnite or MMO and those will be surely dissapointed because that needs a whole new front of optimizations that are (I think) not the goal of the current netcode.
It'll be very interesting where the differences are to the current netcode.

 

It's not always about scale. I have a multiplayer VR game and it's crucial to have very good latency compensation and very high performance, esp mobile VR like an oculus quest.

So it's not just about 100 players in battle royale. Sometimes it's 6 players on a device that must not overheat, that is wireless, so it has to bounce around the room's wifi before hitting the regular internet. Each player has to update 3 points in space at very high accuracy at all times with rotation.

When you have:

• Mobile hardware
• VR response times
• No direct cable connection

Then DOTS networking is hopefully the best solution for it. I have not yet tried it but I am hoping to (it's a closed beta at the moment as far as I know, for bigger teams).

 

Really sorry to bring this up again, but does anyone have any updates?

 

Well, we're only 4 days past the original ~3-4 weeks estimate, so it's not like they're super late yet, and they told us about having shipstoppers just a week ago, so I'd give them a couple more weeks.

 

You have your months mixed up. It has been a bit over a month since that estimate.

 

Not really though?

 

Unite was September 23-26, we are a few days away from that being 2 months ago.

 

Yes, and what does that have to do with anything?

 

Something substantial is coming in 2020.1.

 

Yeah... I don't think Entities 1.0 is coming in any good state in 2020.1. I'm concerned about any hopes of stability in Unity after 2019.2 releasing with 18 known issues, insane compilation times (~3x 2019.1), and still having 12 known issues as of patch 12. As if that wasn't enough, 2019.3 is supposed to be released this year (1.5 months left) and still has 43 known issues in the latest beta (11). Additionally, Unity isn't exactly known for delivering, as has been pointed out.

 

That estimate was given in this thread mid October, it is now mid November.

 

1. All engines have known issues. The latest official UE4 release has 300+ known issues.
2. "Insane compilation times" is one of the reasons why Unity is making DOTS; so we don't have that problem anymore. When working with DOTS, it is recommended to have the "fast enter play mode" option
3. You seem to have a very selective memory regarding what Unity has delivered. In recent years they delivered the URP, HDRP, ShaderGraph, VFXGraph, Addressables, PlayablesAPI, Timeline, New Input System, Package Manager, new 2D tools, IK package, Cinemachine and more. Please recognize that you are exaggerating

We even got confirmation in the other NetCode thread that the Unite NetCode presentation was using a WIP version with a lot of things being non-jobified (which will be fixed for the true release). It was also being played in-editor with 4 simulated clients. Tests were made with 50 players at a tickrate of 60. If I'm not mistaken, that's twice the tickrate that games like Battlefield 4 and Fortnite run at. Things are looking good.

It's called an "estimate", not a "vision of the future". I once estimated 2 weeks for something that took 5 months. I once estimated 2 months for something that took 3 weeks. This is 100% normal and happens all the time.

You can't know how long it's gonna take to solve a problem you haven't solved yet

The 4 weeks estimate would've landed on November 10. So we're only 9 days past that

 

And how many days are we past the 1 month estimate we were given?

 

It comes out when it comes out. It doesn't matter when it was "supposed" to come out.

They're solving hard problems in epic ways, so let's just be grateful and wait. It will release as soon as it's ready to release.

Bashing unity helps nobody, they're solving the hard problems for us so that we can focus on making games!