What are the goals for 6.2 in terms of HDRP rendering, global illumination as Enlighten replacement and SSGI?
I will adapt my reply from over here.
Unity have in the past announced things prematurely and we want to be very careful not to do this again. Therefore, we are hesitant about sharing anything concrete until we are certain we can and will deliver. Under-promise and over-deliver.
That said, I can say that we are working towards a realtime GI solution which will work with a broad range of hardware. For example, our solution will almost certainly not require RTX cards or PS5-grade hardware. Yet, it should still give you the soft indirect lighting you expect from something like baked lightmaps, except no baking is required.
I hope we can share more details soon.
I don’t know about this approach. Not announcing and working on things silently can be misguided. At least ask us what we’d like instead of spending months on something and deprecating it because no one uses it.
I get the under promise and over deliver, but taking it to “not sharing” seems a bit too much. Like I don’t even know what is going on with 6.2
Just like I was surprised to see 6.1 roadmap in the middle of a 6.1 beta…like what is the point of the roadmap…you already decided, already worked on it, and is busy closing it up. Like how many people knew 6.1 was going to have MeshLOD. Heck I didn’t even know that 6.1 and 6.2 and so on was happening. Everywhere it was Post Unity 6… I thought Unity 6 was meant to be a 2023 LTS, now it is LTS.2 LTS.3 …I don’t know and when is 7 happening…like which 6.x is the last version… XD
The official roadmap is full of stuff that is done, and are missing “what is being done now”. And some stuff on it like couple years like creating mesh reflection probes…like no info at all…
How many years is it going to be “Under Consideration” while “In Progress” is almost always empty and one day a bunch of things pop up in “Released”…
Come on…get rid of the roadmap if you are not using it.
Lastly, aiming to for good performance is good.
I get the hardware agnostic approach.
But if you are trying to make a Realtime GI that doesn’t require RTX or even PS5 grade hardware…that sounds very bad mate…
Sounds like we are going to make a Toyota priced porsche…haven’t we reached a good solve with APV? maybe start playing the catching up game with latest tech…please?
I agree. I think we need to find a balance where we don’t release (or god forbid, promise) anything too early but where we also avoid getting feedback too late. From my personal POV this is not an easy thing to do in the complicated machine that is the Unity organization. We want to get this early feedback without affecting the general user experience with anything unfinished or low-quality. We hope to share more soon on how we are going to do this.
We chose this based on data that suggest this is what most of our users want. I appreciate this direction may not be what you want personally, but we believe this is how we can deliver value for the majority of our users, many of whom want to target a broad range of devices.
Are you suggesting you cannot make something nice without requiring RTX? I do not agree with this. Look at examples like Tiny Glade, H-Trace, and Godot’s recent realtime GI probe system. All of them seem to deliver great value and they work on a broad range of devices. Sure they may not be of the highest possible fidelity but that is an intentional compromise because platform reach was deemed more important.
As for the general comments about roadmapping, I am not the right person to answer but I’ll pass on your feedback.
Well, good, but maybe pay attention to the minority users, as they are minority for a reason. It is a chicken-and-egg game, what comes first. Do users join first and services improve, or services improve and users increase. So you know take it with a grain of salt. But you’d get same poll results based on the current user base.
Great value, but nothing that trumps APV. All within the similar sector. Maybe H-Trace is different, and as an user of H-Trace that for it to match RTX quality, it sort of requires RTX level cards anyway. So, if you somehow magically pull of once in a decade move, great, but in case it is a variation of an existing tech at the non-RTX level, then it probably will not perform as great as APV, and look slightly better or vice versa. My point is subdividing an already saturated sector (non RTX GI) seems like a bad move. Isn’t there some work going on for semi realtime APV in the works anyway, where we can update the probes procedurally?
If this doesn’t touch you closer to your heart, think of the relationship of DLSS to FSR/STP. Now to GI, RTX to APV. What you are saying is you want to work on something in between, which is ok, but it will probably never catch up to RTX, I mean look at the research funds they are pouring into it. Hardware scales very well, and rasterization is definitely slowing down in terms of generational scaling anyway.
My point is that, if someone says they want to make a paper plane that can go to space, then I won’t say it can’t be done. My only 2 cents is that why not just make a rocket…
I can see a lot of shitstorm coming my way for saying this, and I’ve been trying/making realtime GI for a few years now, VXGI, SSGI, Octree Conetracing, Probe tracing, SDF GI etc. I even tried to make a decal based GI which, actually turned out to be pretty cool. The point is, it doesn’t match lightmap quality, which doesn’t scale well in large worlds (I think a realistic size is about 500 x 500 meter for a medium size team) and Probe lacking detail and procedural application (and still doesn’t scale well in my testing, mybe 2km x 2km) and both require a lot of elbow grease to make it work. And a non practical time to iterate. A non-RTX GI solution (and eventually AO and Reflection) that has good performance, is RTX exactly, it already exists, a hardware accelerated solution. Btw, we say RTX, but it is just raytracing. But realistically you do need RTX cards for raytracing so…
Back to the topic, I wish you do pull it off, but I feel like it just won’t have a good place by the time you have it ready for release. Unless your release window is in the next 6 months. It is already late, to be honest, but will still serve a good portion for few years. There is another asset dev working on SDF based GI, they already have AO done, so look into it, I believe they are mostly done with the GI system that uses the same SDF system. I think it was called Erebus. I could be wrong. Its the guys working on Nanite, I can’t recall what their Unity version name was. Nanomesh?
I actually think that Unity has a very good foundation for raytracing. But you guys keep culling the feature to improve performance, which I have mixed feelings. Unity handles memory and CPU way better than Unreal, and probably will be a lot more flexible adapting AI into rendering. Its like how ARM cpus were designed for excellent power consumption in the past now scales better today in parallel. Unity painstakingly optimized the CPU side for mobile all its life. I just can’t imagine what will happen if these same crazy devs start eating up the rendering side. And I foresee a similar impact that ARM Cpus have against the x86 Cpus.
@rasmusn, could you improve SSGI for 6.2 to be more performant with more lighting bounce effect?
I trust that you guys are cooking up something good but to this specific point it’s less about feedback and more about helping developers plan. By all means be conservative with dates/releases but it would be great to have a line in the sand that by x point in time some set of functionality will be available. All to the good if it’s available early / in preview. People will dump a lot of effort into working with deprecated GI and stuff like APV scenario blending that they might be able to avoid if the GI replacement is on a public roadmap they can plan against.
Probaby not possible for it to use more light bouncing and be more performant XD
Given that the poor real time GI (or GI alternatives) is the number one issue with Unity (at least in my opinion) it seems like this should get higher priority. In the mean time, maybe just buy out NGSS and make it a bundled part of unity, or something.
Just saying, everything you’ve done in the 6.x releases so far ranks lower than improving lighting/shadows from their current state.
A lack of proper Realtime GI for entities graphics is also my biggest pain point right now. I’d love to use HDRP raytracing, even when the performance isn’t that good and it has some rendering issues (artifacts, light / shadow points that aren’t blurred correctly).
Everything is better than no GI.
But not having it available at all in entities graphics is a real pain for me.
Having it available and in a solid, good looking way would be a dream.
If Unity had better lighting, and better asset imports that can be shared among a team, it would be suitable for AA and AAA on PC and Console. Right now, it still could be done, but at a higher cost than in other engines. I wish Unity would be suitable for all kinds of games, small ones, big ones. Right now, it is recommended to switch to Unreal Engine (or Cry Engine) if your Team gets bigger than 10 people because the Engine will get in your way (with imports, mostly, maybe also with git conflicts).
Better lighting would be a step towards AA sized Console/PC games. I think this is a market worth chasing for Unity. You don’t want a team to get successful with your Engine and then having to switch to another one because the team grew, like the team behind Subnautica for example.
Oh I thought your name was familiar, followed you on Twitter years ago after seeing some cool videos on realtime gi. Could it be the same project, or something better?
Yeah that sounds like me 
.
If you want the latest news on realtime GI in Unity you may head over here: Feedback Request: Changes to Unity's Dynamic GI Roadmap. We haven’t yet shared specific details but we are working on a new system, and we hope to soon be able to invite everybody to test it.