Official Enlighten deprecation and replacement solution

Oh that's really unfortune. So it seems like they didn't really manage to fully remove it yet. It seems like all I have left is hope that they won't be able to, because it actually works are does it's job very well in HDRP 8.1, without it there's a year without Realtime GI. Thanks for the answer @Bordeaux_Fox !

 

I'm curious: You say Enlighten coupled with HDRP works still in the 2020 Alpha? You can still bake the scene again with Enlighten?

 

Yeaa, there is just one bool in HDRenderPipeline files that enables/disables Enlighten in 2020.2. It seems like someone just wrote false instead of true and thought that it removes enlighten entirely, but it's still a part of the engine and works just like before. Setting it to false removes option to choose Enlighten, but all the indirect intensity/albedo boost options are still there so you can just turn it on and set file to read only so Unity won't be able to replace it. It was the same in Unity 2019.3 HDRP and they fixed it so I thought that they just made the same mistake again.

 

I think you mean this?

 

No I tried it but it doesnt work for 2020.2

I mean this one

 

Pretty cool it still works.

 

Will the new real-time global illumination be used with urp as well too?

 

See the official roadmap of URP: https://portal.productboard.com/unity/1-unity-graphics/c/23-realtime-gi-support

It's under consideration, means it may will not ever come to URP. I just know HDRP will get one new Realtime GI system for sure. But again, this does not mean it will be good. And so far I did not see anything promising from Unity in this regard. At the moment, URP is not the right choice if you actually care about indirect lighting because there is even no light proxy volume. Seems URP is heading into an almost baked direction.

 

I'm extremely sad for Unity when open-source Godot can make performant realtime GI in like, two month?.. While we have to wait 2 years for it and only pray that it will support URP... That's the power of commUnity right there

 

This is why splitting the graphics pipelines into three was a terrible mistake. Now every new feature needs to be implemented in the different pipelines, or simply not support it.

 

Probably only HDRP gets realtime GI, maybe URP at a later time, but build-in doesn't.

 

What really frustrates me is that unity refuses to share any information but was quick to cut Enlighten out.

 

Yes, but I also like when she announces as a surprise. it always gets me excited, but until then, it's really a little frustrating to run out of information

 

I used to be surprised but I've realized that Unity rarely delivers on time so, I have been off that hype train long ago. Unity have been fitting releases and announcements to their event days (Unite, GDC etc) lately and have had very little to do with readiness of implementation. GPU lightmapping by Unity is still wonky after all this time and yet there is one dude on asset store who still did a better job, faster.

It is strange as Unity seems to take an awful long time to implement something and when it is finally revealed (usually a few years late) it is an early implementation and takes another few years to make it barely usable. Look at the terrain, DOTS, Hybrid render, raytracing, GPU baker, Navmesh/AI etc. Then they are forced to abandon it after all that time because the tech by that time is too old.

They are just amazingly underdeveloped compared to the time they have been working on it - supposedly with a team.

If you are having trouble understanding where I am coming from then remind yourself of all the demos, especially the Megacity and now look at the Unity executable now and run it. It is pretty much self explanatory.

Let's not forget that HDRP doesn't have grass shader yet which is sort of a joke at this point.

 

There are some answers related to GI for URP and HDRP in the "Road to 2021 Q&A" thread.

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

https://forum.unity.com/threads/the-road-to-2021-q-a.950158/page-2#post-6213113

 

This is interesting. GI solution in Eevee. Visually it makes emissive materials emit indirect lighting almost realtime.

 

UE4 got DDGI plugin https://developer.nvidia.com/rtxgi

 

We don't need RTX, we want GI solution that scales down from mobile games up to procedural open-world without any baking nor high performance cost

 

They already working on DDGI regardless you(we?) need it or not. Personally, i would be very happy because i dont care about mobile games.

 

Sorry, you've got me wrong =)
RTX is great for visualization and stuff, but for games it fits only next-gen consoles and high-end PCs in 1080p@60fps.
Most of us just don't have that super expensive hardware to run it.

It would be great if there was an option to switch between some SDFGI and DDGI and change rendering pipelines at runtime.
That would be an amazing level of scalability.
But in reality, Unity SRP does not allow that (yet, I hope) - you either chose performant wide-range URP where SSAO was only added in last version, or overweighted high-end HDRP that has no scalability, even when every graphic feature is turned off

 

I agree. I think given the history of what we have seen with URP and HDRP that it is likely we might see two different GI implementations.

 

FYI DDGI is not a high-end effect that requires a high end GPU. It doesn't even limit the application resolution or framerate. All it requires is a modern GPU (i.e. one that supports raytracing), which includes the current RTX cards, the next generation of consoles, but also the next generations of mobile devices. In my opinion the only platforms that are screwed are current phones, current consoles, and ageing PCs (which is still a lot, but I guess it's a problem that will 'fix itself' with time).
In my opinion, if you want something that scales from mobile to high end PC and supports entirely procedural worlds, DDGI is the way to go.

 

Which is a whole market now, and would still be for the next several years.
And.. DDGI on mobile? Best luck, wake me up when it'll run on budget xiaomi at 60fps

 

For budget xiaomi you need to bake GI. There is no dynamic GI that can run on such devices and

People for decades trying to solve this problem. I dont know why do you expect this from unity. Even most powerfull PCs couldnt run such GI before raytraced hardware acceleration.

 

Enlighten was a decent solution to that problem that worked on yesterday's mobile machine, and their methods is known. Gi isn't magical, you can run approximation in old hardware.

I'm currently working on a proof of concept that tries to run on open gl es 2.0 phone, by turning the problem into a sampling problem and caching ray hit in a texture as uv. So you can just have the texture sampling itself to resolve GI.

 

Can you share more info on that one?

 

I have a thread in gi firum called exploration of hi compute more or less. I'm currently re factoring the code to get result asap.

 

While I wish this was the case, I have realistic expectations

A GI system that scales from mobile to procedural open world? Sounds a bit impossible. I mean, sure, I'd love it, but maybe we need to dial our expectations down a bit.

Anyway, still no news? Is Unity hoping that raytracing GI (which is a mess at the moment) will eventually just work out?

The recent SSGI was horrible.
baked GI is still full of baking artifacts
Raytracing GI is well...try it for yourself...

Please, say something.

 

Enlighten (by baking the visibility factor (which they name form factor) of each pixel and maintaining a list of the most potent of them).

Okay doesn't scale to procedural open world, but you can have an implementation that does it with similar method (and less discrimination over potency since couldn't bake). Given the sparseness of some current real time method, I think it's possible.

 

Like already said - GI is not something magical.
There's so many different ways to achive it on low-end hardware.
Just watch this hacky talk that firstly inspired me:

They've used light probes and per-vertex visibility maps baking and it worked out good enough for procedular runtime

 

It's not real time gi and it's bake offline. It's for modding purpose, ie create a map, then bake lighting contribution before distributing it. It doesn't adjust bounce to different lighting condition.

In general i see a confusion between baked gi contributions and real time resolution of gi.

Enlighten don't help because it's baking data but to resolve in real time, ie bounce color and light intensity is real time. It basically maintains a weighted list of contributing pixels and propagate the light from the list to get real time gi.

Solution like above freeze the computation in a data structure like lightmap and lightprobe and add real time light on top with no propagation.

My own method try to bake hemispherical approximation above the pixel with uv capture of the environment in cubemap, then gather lighting after the lightpass and accumulate the result to approximate gi. Ie it's similar in idea of enlighten in that it's a light graph.

Advanced techniques try to gather the light contribution of the hemisphere in real time, by resolving the sampling directly, which allows contribution of dynamic objects. They send ray or ray approximation (generally cones) to find geometry over the hemisphere each frame, while enlighten and my techniques just cache the static geometry hits and don't bother with dynamic (though i have some solution that are possible by overlaying uv capture of dynamic on top of the uv capture of the environment).

 

Neoshaman, lets be honest, Enlighten worked for shadow baking and GI for small static scenes. It did not scale at all. In fact, it often just straight out did not work. I think we are giving more credit to a broken system now that we are missing it.

 

It worked when you knew what to do, like not having high Res to map below 1 texel per 2 meters in most open big space because you is a low frequency phenomena. It's even in the manual and unity's tutorial.

 

Any information on when to expect news about a replacement? 2020 is in it's final weeks...

 

And yet AAA games have no trouble with Enlighten.

Unity botched the integration. It took them 3 years to make it somewhat work even for small scenes, and a couple more years later they deprecated it.

P.A.M.E.L.A. devs barely had enough time to complete their game, after fighting through Enlighten issues and bugs for years. That's the lifecycle of a Unity feature.

When they say "Enlighten had a good run", they mean 3 years in what was practically "preview" state, and then a couple in beta.

Makes you think what other features are about to have a "Good run". PLM has been in a preview for a while, maybe it's nearing "good run" status.

 

a prebaked lighting system that required 1 texel per 2 meters...that took forever to bake...

I am gonna stop there.

 

@Jesper-Mortensen , perhaps you want to add something at this point.

Thanks. Good question.

 

Just informational for anyone curious. I just did a count of rtx cards on Steam Graphics cards survey. Rtx cards currently account for just over ten percent of users at 11.33%. Rtx cards currently have a growth rate, at least on Steam, of less than one percent per month.

Obviously the growth rate of Rtx cards will increase over time but it's very slow at the moment. That's not even to mention that decent looking bounced lighting, and other raytrace stuff, on a medium complex scene still doesn't give enough framerate, in my opinion, on first gen rtx cards.

Enlighten, even having to twist its arm, tweak settings, and wait on precomputations/baking, was still better than nothing. It looks pretty good and gives better framerate than almost anything else out there for realtime gi. I know some of us in here want a drop in solution and Enlighten wasn't quite that, but without Enlighten we have limited options. Mostly voxel tracers.

Soo more information dump of other options, some likely already mentioned in this thread. I'll put NeoShaman first because he/she regulars this thread

Realtime global illumination options:

RTGI by neoshaman (not publicly available?)
https://forum.unity.com/threads/exp...n-for-open-gl-es-2-0-and-weak-machine.752600/

SEGI by Sonicether, lots of voxel tracers based off this (tested working in Unity 2020 standard pipeline)
https://github.com/sonicether/SEGI

Nigiri almost-dynamic voxel-based by ninlilizi, this girl cracks me up, known for hijacking other global illumination threads. She's got a newer version in the works that I'm highly curious about.
(old version but tested working in Unity 2020 standard pipeline)
https://github.com/ninlilizi/Nigiri

SRP-VXGI by Looooong (tested working in Unity 2020 standard pipeline)
https://github.com/Looooong/Unity-SRP-VXGI

HXGI by Lexie (not publicly available), really curious about this one as well.
https://forum.unity.com/threads/hxgi-realtime-dynamic-gi.472486/

SSGI (ScreenSpace Global Illumination) by Raul_MadGoat (will be paid - not publicly available yet)
https://forum.unity.com/threads/wip-madgoat-ssgi-screenspace-global-illumination.849214/

UPGEN by mm_ASH (Paid - Downloadable hdrp demo on store page, standard pipeline also available)
https://forum.unity.com/threads/upgen-lighting-standard-hdrp.898526/

REAL-TIME Ray Tracing by Parsafari (not much info, no rtx card needed?)
edit - dang darn Unity forums forcing youtube embed!

Unity RTX raytracing just to list it (tested working on Unity 2020 hdrp)
https://docs.unity3d.com/Packages/c...@10.2/manual/Ray-Tracing-Getting-Started.html

 

Well, some of these solutions are... not really what you could consider a solution.

• neoshaman's RTGI is still in too early days of development, so hard to tell where the journey is heading.
• SEGI is dead - and personally I wouldn't touch any voxel-based GI solutions due to their performance hunger.
• I had high hopes for Lexie's HXGI, but with no recent updates (neither in his thread nor on his twitter account) it's safe to assume that's just another abandoned product.
• MadGoat SSGI looks interesting, but SSGI in itself is not really RTGI in a traditional sense.
• Parsafari's GI system apparently uses "Ray tracing hardware acceleration on NVIDIA GPUs that have compute capability 5.0 or above to calculate lighting" (src). It works in HDRP & LWRP only (src) and the release is TBA (src).

 

Alright so I have been waiting for some news on Unity's Realtime GI and was getting very suspicious as the claimed "2021.1 date" is not that far and we have had almost zero news.

And I see this comment on one of the blogs.
LeonhardPsay, August 18,
We’re actively working on defining the solution for our customers for real-time GI, fully understanding how impactful dynamic global illumination is to modern games and many of our customers’ use cases. While we are not on track to provide a replacement in time for 21.1, we are still aiming to share the plans by the end of 2020.
https://blogs.unity3d.com/2020/08/13/the-road-to-2021/

So why was this IMPORTANT message not updated on this post, in fact not announced properly so that we are all aware? Instead you guys choose to sneak this info in as part of a reddit reply?

Am I the only one who missed this very important message or did everyone?
This is really not good business practice...really bad...

So you guys are planning to share the plans by the time you promised to provide a replacement....

I mean I should have known when the Roadmap 2021 post was just full of candifloss and no mention of new GI system...but this...this is another level...

 

If a Realtime GI solution is pushed out until who knows when, maybe it's time for Unity to at least keep Enlighten going in HDRP until you actually have a replacement? At this rate, there won't even be an LTS version of Unity that supports both HDRP and Realtime GI.

 

Let's be honest. Anyone who believed a ready-for-production GI system was going to be built between the time they said 20.1 while refusing to give any details and the date supplied ... has not paid attention to the track record of new features.

They'll share when they have something positive to share is my guess. Until then, consider it hopes, dreams, pixie-dust, and possible vapor-ware. That way itll be a nice surprise, as opposed to a string of disappointments as we wait.

 

I was just wondering, as a paying user of Unity, if I requested development information regarding Realtime GI to Unity formally, are they legally obliged to provide information?

I am open to the idea of taking this to another level if Unity continues to provide zero information let alone keeping their initial release date promise. No information for what? and for what purpose?

If some asset store author did this (going dark about a promised feature) then I'd accept it...but Unity?...seriously this is not professional nor morally correct. If you knew already that you couldn't keep up to your initial promise, then announce it properly so that everyone can get on the same page and provide information on as to why and how it will be handled.

This was what was told to the community:
We are also fully committed to delivering a real-time GI replacement solution in 2021.1. The Unity team has a solid plan to solve this complex problem the right way, with great artists workflow and optimal runtime performance for 2021.1.

 

No, I highly doubt it. As a paying customer who owns multiple versions + subscription + many thousands on the asset store, even when Unity had broken features which caused problems with launching projects, I was told they would not provide any feedback into the status of said features unless we paid 10k a year for multiple seats of tech support. I'll leave it as it was the most offensive customer support I have ever had, anywhere. No one cares. Ultimately, they just want money. Years later, different posts about this made, forum threads asking devs to respond, e-mails to various Unity supports ... and I never got a response outside of the single representative who basically told me I wasn't worth their time or money, even though I was a paying customer with a project blocked by an unknown status of a known bug they refused to comment on after a dev had already said it was being fixed and then disappeared.

My request was basically what you are asking for, only it was about a feature that was already launched but horribly broken on android while being massively marketed. It had many forum threads, a dev response stating the bug was being fixed before he disappeared to never show up again, etc.

So no, they aren't going to hold themselves to anything. Seemingly its against company policy, even if its about their own broken features they are heavily marketing. If not for that, then definitely not for this, which isn't even real yet and no one can claim its affected them short of basing their projects on their projections.

My advice is to just assume Unity is a company out to make money and will market whatever they can to achieve that, but happens to also provide a great product. However, those two things should be viewed as different worlds ... the marketing, and the actual product you use. Not mixing the two makes things a lot more palatable.

 

DGordon, I definitely share your frustration there. It seems the downfall of a decent experience with Unity was all started when an Ex-EA CEO took over to get the company ready for an IPO. They constantly talk to shareholders about expanding into other industries while leaving the developers that got them here flailing with broken, missing, or half-baked features.

 

I actually don't believe that one CEO is ruining everything like this. My gut feeling based on my past IT work experience is that everything in Unity is compartmentalized. It makes running a company easy and somewhat okay on the surface. The only major problem is that no one is really held responsible.

For example, the whole realtime GI deprecation that continues into this current mess. Why isn't the graphics head coming out and apologizing? Why isn't there a corrected roadmap? Why isn't anyone coming out and saying "sorry we messed up, and this is what we will do to fix it."

Because no one is "really" responsible for the feature.
Sure there are dev teams allocated, but they are probably swamped trying to meet deadlines, and usually it is not their fault. The issue is the PR/management thinking "aghhh too bad". They think like this cuz they know that they will not get fired for this stuff. No one is really running the team to meet the needs. The problem doesn't end there. When a feature is dependent on another, then the whole loop exponentially increases.

Want another example?
Let's talk about DOTS...I mean, WTF? Seriously, how long is this going to take? 10 years? Now the hybrid renderer is just pretty much stalled at infancy. It says version 2 but its more like internal prototype 2.
The terrain system? Seriously... there are no ECS based detail placement cuz DOTS is not ready. Grass is like all dead at this point.
Progressive mapper? phhhh....I don't even know why.
Are you aware that ECS is slower than the current Gameobject base unless the whole game is restructured? And the performance benefit happens in very very thin slices of situations?
Raytracing? NavMesh? Kinematica? Input System?
Seriously, they are all below average at best.
And why is Probuilder and related tools breaking every third update....

The only good implementation that I saw recently was the revamped TAA for HDRP and Animation Rigging (after a long list of updates, but still good)

I am aware that dev tools will always have issues and that game dev's job description innately includes working with an imperfect tool, but recently, Unity has been comfortably crossing the line and I am very worried.

These things just go on and on and on. Some of the things they are working on, are really long term stuff and I understand...but we are talking 2~3 years since public reveal. I mean, come on.

And then when finally our complaints reach the royal management, they decide to restart the whole thing by starting another package to attempt to fix the incomplete feature. Then the cycle begins again.

Sigh

 

Well at that I disagree. Unity took big risks with SRP, DOTS, VFX Graph and etc, risks that were necessary for the future of the engine. Especially in this world where Unreal is super agressive on the rendering front; Unity had to step its game up.

Of course they could have rolled out an alternative to Enlighten sooner, or kept supporting it longer, but Enlighten is still made by a third party and a third party can always close or stop supporting its product. Furthermore, rushing an alternative is something that would give developers major headhaches so I am not convinced this was the preferable option as well. I was vocal about this feature drop because it did affect us negatively, but that does not mean we do not understand it's quite possible that this was the rationale thing to do.

While not ideal, we still have baked GI and that works really well. Of course that is a limitation in design, but would we return to the look and feel of the built-in redendering pipeline? Heck no! HDRP looks stunning, is performant and stable. DOTS and VFX graph are game changers.

It is easy to say things like "ex-EA CEO taking over to drive an IPO", but reality is that this IPO brings more cash into the Unity ecosystem and that will certainly mean more people working on the features we love. We see big progress on HDRP, VFX Graph, DOTS and the raytracer. A speed of progress of a healthy, commited company, not of one that does not care or one that just care about its shareholders.

Just go into the DOTS section of the forum and you'll see Joachim Ante answering personnaly to developers feedback. This guy is the CTO and cofounder of Unity. That is one big flag of the executives caring about their userbase.

 

alexandre-fiset and jjejj87, I hear what you're saying and I agree with a lot of it.

My point is, the CEO is responsible for steering the direction of the company. The direction that I see is more towards diversification of userbase through new features and packages rather than finishing, improving, and/or updating longstanding pain points that more developers deal with. This is just my perspective.

All that said, Unity has amazing people that do some very awesome things. There are plenty of examples of dedicated and passionate Unity staff trying their damnedest to get things done. It appears the problem lies in the leadership not keeping focus and throwing spaghetti against the wall far before it's al dente.

 

Look, before this gets weird, no one is really questioning if the people at Unity are good or bad people and there is no need to cover it every time a hard message needs to be put out.

The point is whether the product that we are paying for, and the eco system central to our businesses, ie the clients of Unity, is getting treated in a manner that is acceptable, and whether the company is doing its best to achieve it.

And we are at a point where I suspect that the company is intentionally and selectively omitting information and we are "suspecting" that the reason behind the recent deterioration is based on the management (eg ex EA CEO) or the IPO.

Now, the ex EA CEO, I'd rather say no more as it may bounce to wrong direction, but IPO, I do not understand. If IPO is imminent, then it is more the reason to run a tighter ship, afterall, it is the prospect of success that drives the IPO. Besides, I don't think IPO will bring more funding to the dev scene anyway. It is the stock holders that will make money, not necessarily the company and definitely the dev team (maybe few more projects might pop up, but then read previous post)

Also, when it comes to Joachim and DOTS, I really don't have good things to say. I like that a tech dude is running point on all this massive project, but time tells that

1) Time spent vs outcome has been really poor
2) His team is clearly doing more than they can handle (Job, Burst, ECS, Hybrid Renderer...)
3) Of which only Job system and Burst not broken but Burst needs to be updated and worked on continuously as DOTS develops
4) Job system performance is....well...sort of mediocre...
5) And lastly, major problem, is that none of this is near its end. Honestly, I expect this to go on for another few years...and until it is done, it will not be practical...

The issue with 5) is that over time paradigms change and common sense changes in this field. For example, when I suggested SSGI 2 years ago, people were convinced that SSGI will never make it to mainstream and that it is not practical. Welp, look at Unity SSGI trying to catch up to industry standard now.

The point is things change, and even things that seem like will last a good decade often last much short of it. And when a feature goes deep into development cycle for few years, usually this does not work out so well when it is done after many years. The grounds that made the development legit will have shifted.

Lastly, Unity did not take any risks going into SRP. In fact, I see it as their risk-hedging attempt. If they were taking risks and wanted to stand up to Unreal, then they should have just upgraded built in to HDRP. It was a good choice, because if they chose to go otherwise, it would have been another Unity5 disaster.

 

Making HDRP the new standard renderer would not have solved any problem. The problem is that Unity tries to support both High-End and Low-End devices and they have always been stronger on the Low-End side (especially mobile). Splitting these to segments into different render pipelines is a smart move but only if you have the resources to do so. That's what Unity lacks in (apparently).

Also I think the IPO filing was inevitible. Epic is growing fast and Unity doesn't have a Fortnite which gives them billions of dollars. I hope we see some results of that money in the near future.