lumen lighting is very good in interiors in robocop and fortnite
Digital Foundry Article Technical Discussion [2025]
- Thread starter Shifty Geezer
- Start date
arandomguy
Veteran
This is a bit of an interesting subject because I remember there was similar concerns when games transitioned from sprites to polygons. People were used to sprites games visually. But in the long term 3D polygonal games established a new paradigm of what games could be that sprite based games could not, as such the transition went beyond just the visual aspect.
arandomguy
Veteran
I'm just going to guess but i don't think hardware lumen is really going to work for current gen consoles given the hardware constraints. It's likely the performance budget is best used elsewhere.
Early cross gen games (or games that stick with a "legacy" pipeline) for next might be when.
Circling back a bit one thing I speculate about is what the actual future of RT is going to be given the now full parallel developments of ML based graphics techniques. It'll be interesting if ML based rendering displaces RT based rendering as we know it before that can even fully displace raster based rendering.
Yeah I think it's just practically too shiny for the ray budgets in use there. I'm sure HWRT would help somewhat but most of the issues there seem to primarily come from pushing specular/reflections too hard.
As DavidGraham noted, this would be VSMs (direct shadowing) not Lumen (indirect diffuse/specular). VSMs support all of the local light types and even approximate area lighting. Whether a light casts shadows is fully under the control of the developers though, so presumably they turned some of these off as a performance compromise/priority decision.
For cutscenes specifically - somewhat like film lighting - it's basically all hackery. The more people/time you have the more special case tweaked it can be. Even with GI/raytracing systems in place cutscenes will often set up tons of fake lights with carefully tuned shadows, fake light blockers and diffusers and so on. Think of these sequences as being lit more like a film shot than like a realistic scene, because that is what people expect to see. Film would look weird to people if they didn't largely fake the lighting as well. Thus while fancier lighting tech can sometimes help push the upper limit of these visuals, a lot can be done even with relatively modest tech and a lot of careful tweaking. The fancier, more physically accurate and general lighting stuff helps more in gameplay and sequences with unpredictable camera/assets/etc.
Right and this is the real big one, because it affects so many decisions that gamers may not be aware of directly, but can make the game better regardless. In those old games you simply could not move lights, so the sequences were designed around that. Obviously we did a good job within those constraints at the time but believe me the writers/designers still felt the pain of that constraint on the story-telling.
But even if you don't need flashy lights moving around everywhere, there are other ways that things can be impacted. I was at the GDC talk for how they did the Star Wars Outlaws planetary transitions and there were several points where they made it clear that it basically would not have been possible to do it with baked lighting. The issue is that in a lot of these places to make it seamless they have to do things like block some lighting in one area, then teleport the whole scene to a new level, then reveal that new lighting and similar. Stuff like that simply isn't possible in a baked world in any sort of practical way.
Some of that improvement you will only be able to feel when you can make a game that targets RT exclusively. We've only just started to see games that require RT hardware, and even in those cases the use is fairly light on the consoles. There's no question there are production benefits to heavy RT use in the long run, but in the period where you have to manage both RT and raster across a potentially wide range of platforms and performance levels, it's harder to realize those production gains.
You can absolutely do this? This is what a tech demo traditionally was. You don’t need to create a full length game to research new 3D rendering features, do you think all those academic papers on RT were attached to full fledged commercially released games?
I’m not talking about online, like I said I’m excluding technical forum people (who generally love this sort of stuff). I’m talking about people I speak to in real life about this. The most common sentiment is “yeah it’s cool that it’s all real time but it halves FPS and you need aggressive upscaling to make it work”. These are mostly somewhat-technical friends of mine, but I’ve also asked people who don’t really keep up with the technical side and they mostly can barely tell the difference.
I can’t say I’ve ever noticed this.
DavidGraham
Veteran
It is there, it is the consequence of lighting those characters by cubemaps or by coarse global illumination probes, where the characters pick up the color of the sky even in interior areas and under shadowed/indirectly lit areas.
Transferring techniques and approaches detailed in academic papers to game development, with entire game designs and content pipelines and specific hardware and specific performance requirements, is a huge undertaking.
Whitepapers and even tech demos are lightyears distant from finding out what works and doesn't for your game, and games are often in development while key technologies are being developed and refined. Look at the radical changes that UE5 has seen, and how those technologies have iterated and performance has changed.
Image starting development on a game 4 or 5 or 6 years ago and trying to land that on the moving target that is UE. If your buildings are leaky when using software lumen, or your choices of materials are not ideal for it, or performance isn't where you want it, you have limited options at the time you ship your game. And as technologies get more complex and development becomes longer and more involved, managing issues like this is only going to get more difficult.
You can't look at a white paper or tech demo and be certain of where your UE5 based game will be years in the future. One of the advantages that in-house engines probably have over UE5 is the tight integration of engine and the game's development, and the ability to direct the entire engine towards the needs of one (or a handful) of games. But only if you've got the talent, the budget and the time ... and that's a really big 'if'.
arandomguy
Veteran
So this is something that can benefit a bit from a different perspective. Just comparing against the best games from 5-10 years ago with huge teams with unlimited budgets shouldn't just be the criteria.
For example with a game like Clair 33 and similar or smaller in scope games today, how do they compare to games of similar budget/dev size and scope from 5-10 years ago? How do games compare across the stack and with a large sample size vs 10 years ago?
The other issue is that 5 years nowadays is too short a period. I know being used to tech advances in the past can make 5 years feels like an eternity but the market/industry is much more mature now. 5 years can be the development cycle of a single game for perspective nowadays. It might be 5 years now before there is even a doubling of hardware capability from a cost and energy perspective. We are far from the era in which even 1 year would see generational advancements from both the software and hardware front.
That isnt true. 4A games has ~200 employees and their GI solution is one of the best on the market and miles better than Lumen. Inhouse engines have to be efficient and must create a return of investment.
4A games is precisely one of the developers I was referring to. They have a great in-house engine that is tailored specifically to the needs of the game they are making. Their GI solution's focus can be entirely dedicated towards the goals of that game, and their content can be made to play to the strengths of their technology roadmap which they control in its entirety.
But that does not mean that 4A's GI solution would be better than Lumen in the context of UE5, where it has to support all kinds of games with all kinds of requirements and all kinds of assets, and the flexibility to out of the box do well with a 1 v 1 fighting game or turn based strategy or a huge open world sandbox adventure, while also supporting something like Nanite.
It's fine to state "I think the tradeoff that these games/systems made on performance/IQ is better" but pretending that a system that does just relatively coarse indirect diffuse is comparable to systems that do fine-grained reflections and specular is a bit silly.
To be clear, I agree that indirect diffuse-only is a good solution for a lot of games and content, and is *far* cheaper than shiny things. But the two are in very different categories of theoretical and practical complexity... diffuse-only GI you can make a lot of assumptions and simplifications that just don't work at all for specular. Fundamentally with specular a lot of the assumptions around "things far away from the surface matter less" fall apart.
I think Lumen could probably improve quality/performance in cases where people only want indirect diffuse and that's a fair discussion to have. But it - like several other modern "GI" systems - really targets a tier up in the global lighting space. It's also why I do think we at B3D need to stop calling all of these things "GI" and at least start talking about what terms they are actually approximating. Unfortunately all of the conventional words in this space have becoming meaningless now that marketing has gotten their hands on them.
Last edited:
How many concepts actually worked in a real game? Academia finds potential, but only when it's tested in real products does it prove itself or prove to be practically too limited. We've had interesting lighting paradigms going back a decade or two like SVOGI that didn't transform games. Back before this gen released there was a lengthy discussion about whether RTRT hardware was needed in the next consoles or not, and all these academic papers and impressive YT videos were presented, and none turned into games in the intervening years. Very pretty Cornell Boxes and Stanford Bunnies don't indicate a technique will scale up for a real game.
And for production companies, they haven't the time or resources to create demos to explore if a tech is ready or not. If RT sounds good, they have to get on it and implement it in the next game. they cant' create half a game around it, find it doesn't work, and then re-implement the whole thing. That kind of approach results in Star Citizen and Dreams.
Good review! A little sad to see the TSR-related flicker on consoles. I'm curious whether 60fps is actually a big advantage to the QTEs - I've been playing at high frame rates on PC but as I've gotten further into the game you sorta realize that the audio cues are a way more reliable way to do the parries/dodges than the visual ones anyways (which often have fakeouts or long pauses). For one boss I tested by closing my eyes and was able to do the parries perfectly still
Small note - VSMs in UE5 can only be sampled with shadow-map raytracing (outside of some debug options). They have a bunch of scalability settings though, one of which will effectively put a cap on how soft shadows can get. I imagine on the performance mode this one got cut a bunch to the point where the max softness just isn't that noticeable beside the probably also lower resolution sharp shadows.
In general Clair is a surprisingly great game. Highly recommend it to anyone who has even a passing interest in the genre, or really anyone who enjoys a good story and lots of artistry in the visuals/music.
Thats fine but why does that mean every UE5 title needs to have these taxing lighting methods?
My point is I don’t think ‘rank and file’ games ought to have lighting that’s so advanced it starts to impact IQ from how far we have to push upscaling.
Yep. And some people agree with you. And some don't.
DavidGraham
Veteran
Didn't Metro Exodus ship with ray traced reflections on PC? or is that a separate thing from specular illumination?
Been a while - looking back at the DF video it does seem like they do have some (basic, possibly unlit?) RT reflections on water specifically, but not other specular surfaces. Certainly a reasonable-tradeoff but again, not exactly apples to apples with other systems.
gotta wonder if a pure UE5 rendered could get rid of transitions and loading screens when going from the surface to a cave. That'd be excellent for any future TES game.
Similar threads
- Locked
