Download

Upcoming Push for dynamic GI after Neo & Scorpio Announcements?

The majority of games are really only 3 types: Open World, Map-based and Corridor.

There are several games now with some type of dynamic GI that the excuse can’t fly anymore.

There’s a lot more to it than that, graphical style, gameplay, etc. all can have an effect on how you might do graphical features.

Well, not every game needs DGI. So saying that engine needs DGI that works for thousand varieties of projects is incorrect.

SVOTI in Cryengine cost a fraction of VXGI in UE4. :frowning:

Epic wants features that most games can take advantage of, if dynamic GI can only work in the high-end games then it might not be important enough for them to want to put the resources into it. Ideally it would replace the current lighting system entirely and they could just go with that for everything.

Well, why not? That was the plan with Fable Legends and the reason why LPVs in the engine exist in the first place. LPVs, DFAO, Distance Field Shadows, DFGI, Heightfield GI, all of these things are already settings that can generally be set by the user through the options menu either directly or indirectly. I would argue that having the option there for the users that can run it wouldn’t negatively impact the experience of the users who couldn’t, assuming realtime GI isn’t actually a requirement of the gameplay.

We’re already doing exactly what you’re describing whenever we include a scalability setting in a game, I’m not sure I see what the difference is. All that happens when you don’t have the option to begin with is you’re taking options away from the end user for really no reason assuming the integration doesn’t really take that much time to integrate (such as UE4 LPVs which is quite literally just a checkbox and all the work is done). It’s hardly an all or nothing scenario, this is easily something that can be transitioned to over time, and the sooner work begins on something like this the more developed and faster it will be when the average user has the necessary hardware.

“global illumination = indirect illumination from light source interacting with other surrounding objects.” so what happens when those “surrounding objects” go out of your screen space? They stop contributing to the lighting. So it isn’t global :wink:

that’s why screenspace methods aren’t good. same thing happens with SSR :wink:

do you understand that even with one object that it is still global illumination?
it has nothing to do with screen space. global = indirect light, local = direct light

What I mean is, it’s not like you can mix between Lightmass or DGI. LPV is insanely cheap anyway so it’s not really a similar comparison. I mean sure - you can do it, but the game is going to look significantly different without it. I wouldn’t personally put something as critical to the appearance of the game as lighting, only for the top-end of users.

Besides if that is what you’re going to do, you may as well use VXGI and scale it down.


I’m all for a DGI solution, but at the same time it has to be recognised that Epic has other priorities too and a small (albeit insanely talented) team of rendering programmers. Sequencer, Niagra, Vulkan, better VR performance etc are all part of that list and in some cases more urgent than DGI. Epic is more than aware that they need a DGI solution of their own eventually I’m sure.

Either way, I still don’t understand the crazy urgency for it. VXGI is really good and can be used in a production pipeline. It’s as scalable as anything else.

To some extent you can. If you’re building something with a mix of interiors and exteriors, you can bake down the lights that don’t change and keep VXGI for example on the lights that do. It’s a totally viable path that makes dynamic GI a slightly less painful thing to add to the game.

Unfortunately Epic isn’t in a position where it’s just by itself in the market right now, and their response to a growing trend in the industry can’t just be “We’ll ignore it and work on other things, hopefully it’ll work itself out.” Yes, they have a small team, but with the initial UE4 Elemental Demo they showed that they -can- pull it off if they choose to. We’re already at the peak of lightmap technology and where it can take us between things like stationary lights or what Enlighten has been doing, until work starts to get done on a more realtime lighting system it’ll just stagnate where it currently is.

Direct support from Epic is the main push, I think. There’s a lot of uncertainty behind VXGI, such as how it’ll run on a wider range of AMD cards, proper support on consoles and other OSes for multiplatform releases, and whether or not Nvidia will even be updating the branch for it a year from now. There’s also of course the large consumer backlash among the hardcore crowd against Gameworks as a whole at the time being, and it could very quickly turn into accusations against the developers of the games if the game doesn’t perform well because of it. People don’t always understand that sometimes the rendering technique is just really demanding, and not necessarily slow because of the company that made it.

Would Epic put real time GI into priority list if…
Community did a crowdfunding and raised like $1 million for it? (stupid idea)
Or if we get a thread with like 1000 community member posts in it expressing they need it?
Or anything else, would anything affect it or it’ll stay backlogged until the “unknown” ?

If everyone who uses the engine voted on it then that could make a difference. Doing a forum poll only gets people who read the thread and care to vote

What’s the most effective way to get everyone to have a say on it?

Edit: Would an online petition do the job?

Would Epic approve it?

Except that it’s not multi-platform and it’s prob slower than 5yo SVOGI…

I would definitely sign it :slight_smile:

SVOGI is memory efficient but slower to trace. VXGI has hardware optimized voxelixation and fast tracing thanks to uniform grid.

Meanwhile lighmaps was more cheap that gi there are no reason to use gi. Better hardware means more trees, more monsters there are not reason for lost render time in static objects.

Btw i was listen “lighmaps are death” from 20 years.

Nobody’s saying to abandon lightmaps entirely, just that they don’t work well for every single type of game that one might want to throw at the engine. Unreal 4 in general is amazingly well put together, and it can be used for almost anything. However, recent trends in gaming have pushed towards a more procedural or user driven world that changes over time, and that’s just not something that lightmaps play nicely with. Even things like the Battlefield series, which arguably don’t really need those sorts of features, have seen things move towards a more dynamic world with all of the destruction that the maps offer to the player. In addition to that, when you have either larger worlds (such as in Epic’s Kite demo) or a lot of dynamic lighting, lightmaps aren’t as useful there either.

Lightmaps aren’t death, and lightmaps are definitely not dying. Nothing here is getting replaced, only expanded on to give developers more options.

How about this: http://graphics.cs.williams.edu/papers/DeepGBuffer16/