Next gen lighting technologies - voxelised, traced, and everything else *spawn*

[London Geezer]

That's the nature of ultra settings in general, the differences between High and Ultra in most games is hard to spot without side by side comparisons, heck, even the difference between High, Medium and low is hard to spot most of the time.

The problem is it's not really that evident between ON and OFF - forget about High vs Ultra! Also the point is, if you can't really see it then what's the point of the performance hit??

 

[eloic]

We still have RTX support for Shadow of the Tomb Raider coming up. Let's see how that will work out. Apart from that, do we have any other games with RTX support to look out for?

Atomic Heart is supposed to be a true showcase. Later on, Justice and Control.

 

[DavidGraham]

The problem is it's not really that evident between ON and OFF - forget about High vs Ultra! Also the point is, if you can't really see it then what's the point of the performance hit??

You can't expect that to happen in every scene, light cheats abd static GI is very well made in the game You also don't focus on the scenes that don't work and ignore those that show a huge difference . Most outdoor areas show a significant improvement, that alone is enough to showcase the tech.

 

[London Geezer]

You can't expect that to happen in every scene, light cheats abd static GI is very well made in the game You also don't focus on the scenes that don't work and ignore those that show a huge difference . Most outdoor areas show a significant improvement, that alone is enough to showcase the tech.

I've watched the video. Literally paused and slow-mo all these moments. In my eyes/brain, the difference just is not 'significant' as you say. Sorry *shurg*

If pause+slowmo is required to see these differences (and still have a hard time recognising them), then that kind of defeats the point of it all, no?

 

[milk]

Are lightprobes completely off when RTX is ON? As far as I understand, their RT only does 1 bounce. Proper GI requires multiple. So even if they have RT first bounce in there, it would still be important to include probes for 2 bounce and onwards.
If they did not do that, their RT GI is not realy more accurate than the traditional image based one. It just has a different set of inaccuracies.

 

[DavidGraham]

If pause+slowmo is required to see these differences (and still have a hard time recognising them), then that kind of defeats the point of it all, no?

No. Because graphics suffer from the phenomenon of diminishing returns, difference between PS1 and PS2 games is large, but it's smaller between PS2 and PS3, and even smaller between PS3 and PS4, PS4 to PS5 will be even smaller than that. The closer you get to photorealism the harder you would need to look. However, once you go back to an older generation, you will miss all those little details you shrugged over. Your brain registers everything, if you played the game with RTX on, and then went back to playing it with RTX off, your brain will immediately tell the difference in out door and several indoor scenes and recognize you are running a lower quality version of the game.

 

[London Geezer]

No. Because graphics suffer from the phenomenon of diminishing returns, difference between PS1 and PS2 games is large, but it's smaller between PS2 and PS3, and even smaller between PS3 and PS4, PS4 to PS5 will be even smaller than that. The closer you get to photorealism the harder you would need to look. However, once you go back to an older generation, you will miss all those little details you shrugged over. Your brain registers everything, if you played the game with RTX on, and then went back to playing it with RTX off, your brain will immediately tell the difference in out door and several indoor scenes and recognize you are running a lower quality version of the game.

I totally agree with that. My point, and my OPINION, is that this specific example (Metro) is nowhere near a huge jump to photorealism, compared to the non-DXR version. Hence why you can't see it.

DXR has made things look a lot better than this game.

 

[Ike Turner]

RTX

Are lightprobes completely off when RTX is ON? As far as I understand, their RT only does 1 bounce. Proper GI requires multiple. So even if they have RT first bounce in there, it would still be important to include probes for 2 bounce and onwards.
If they did not do that, their RT GI is not realy more accurate than the traditional image based one. It just has a different set of inaccuracies.

RTX OFF:

Hallf Life 2 has some competition... The lighting in this game with RTX Off is ridiculous (by today's standard...).

 

[Jupiter]

GI is bad and very basic in most games.

RTX

Off


On

I can spot the differences in Metro quite easily. It appears more movie like with enebaled RTX.

 

[DavidGraham]

My point, and my OPINION, is that this specific example (Metro) is nowhere near a huge jump to photorealism, compared to the non-DXR version. Hence why you can't see it.

I respect your opinion, but the game is simply astonishing to look at in motion with RTX on, and those deep blacks help immerse the player more in the world. We should get more visual showcases for the game once it launches.

 

[Ike Turner]

GI is bad and very basic in most games.

RTX

Off
View attachment 2879

On View attachment 2880

I can spot the differences in Metro quit easily. It appears more movie like with enebaled RTX.

Because the non RTX lighting looks like utter trash compared to today's games.
"Movie Like"? "It just Works!"

 

[London Geezer]

GI is bad and very basic in most games.

RTX

Off
View attachment 2879

On View attachment 2880

I can spot the differences in Metro quit easily. It appears more movie like with enebaled RTX.

I respect your opinion, but the game is simply astonishing to look at in motion with RTX on, and those deep blacks help immerse the player more in the world. We should get more visual showcases for the game once it launches.

Yes, that is quite a nice upgrade. But you're comparing a VERY BAD non-RTX presentation, to their RTX one. No AAA game would have such a rubbish 'standard' lighting implementation, and at the same time I'd argue that the result in the better looking image could be done with non-RTX solutions.

 

[N00b]

Hallf Life 2 has some competition... The lighting in this game with RTX Off is ridiculous (by today's standard...).

BUT we have seen games that had much better lighting long before RTX. So the question here is, why did the devs not implement a proper fallback for RTX off: lack of time, lack of ability, wish to make RTX look better. Nobody knows. But we know it can be done better.

 

[Ike Turner]

BUT we have seen games that had much better lighting long before RTX. So the question here is, why did the devs not implement a proper fallback for RTX off: lack of time, lack of ability, wish to make RTX look better. Nobody knows. But we know it can be done better.

Your guess is as good as mines. Would love to see direct comparison (same exact scenes) with console versions just "in case" (I mean it's not like NVidia already did stuff like paying for the exclusivity of god rays in Far Cry 4 PC while they where perfectly fine & present on PS4/Xbox versions.)

 

[Shifty Geezer]

I've watched the video. Literally paused and slow-mo all these moments. In my eyes/brain, the difference just is not 'significant' as you say. Sorry *shurg*

If pause+slowmo is required to see these differences (and still have a hard time recognising them), then that kind of defeats the point of it all, no?

I think as well, some of us are thinking of raytracing as being photorealistic, or at least finally having correct lighting so video games move beyond the gamey look that we've had since 2005. We've all seen 'next gen lighting' and it looks clearly next-gen. RTX isn't achieving that at the moment. For those for whom mild, iterative improvements are noteworthy, they possibly see Metro as 'better', but for us, there improvement are far from what's needed to class as next-gen. Coupled with the considerable cost of entry for these mild improvements, we're unimpressed.

 

[JoeJ]

Because I think in the future we will get the the point where games uses things like physically correct refractions, translucency, reflections, real-time GI and you instantly can tell the difference between DXR/RTX on and off.

Your order of effects is interesting. I would reverse it if we begin with the most important. Also i think we do not need RTX for GI so let's ignore this for now.

For sharp reflections RTX is very good. I still see BFV as the RTX killer app.

Translucency - what do you expect here? Better skin or vegetation?

And refractions - is it about better water? Or better Glasses on a table?


Atomic Heart shows most of these effects. They are local: A soft shadow here, a reflection there, cool water and lenses, etc.
There is no dynamic GI and no unified lighting solution to make it consistent. But does this game otherwise satisfy your expections the most? What do you still miss?
(Control is no good example because it is a too clever mix of static and dynamic stuff. Will look good On and Off.)

I'm trying to figure what people want. And i tend to ignore the transparency problem - maybe too much.

 

[Jupiter]

CIG graphics programmer about Raytracing:
 

Proper Reflections -I'd call this a subtle improvement and a medium sized headache: It has to be a subtle improvement because we can't have any scenes where improper reflections look terrible, and it's a headache because we currently have a pretty optimised system for both culling and lighting objects that 100% relies on anything that needs to be seen, being visible in the viewport. So right now, objects behind you aren't lit, there aren't any shadows for them even if they were, and if you had one of those mirrors like they have in train stations, to see round a corner, you'd probably be seeing empty space. Anything that's calculated ahead of time would need to be much more conservative just in case a ray wanted to hit it later.
Proper Shadows - Better sharp shadows is actually something I think would be really easy for us to use this for, a ray per pixel per light in areas that were likely to have shadow aliasing would clean things up really well, without just cranking the shadow map resolution ever-higher. Soft shadows need quite a few rays per pixel per light, so we'd probably have to do something to restrict them only to the areas where there might be penumbra, and there would still probably be areas where a lot of the screen was covered by penumbrae.
Global Illumination - The very grandest of headaches! Diffuse GI needs a ton of rays to work well, I assume we'd have to cast them at a reduced resolution (either screen resolution or a sparse set of points in the world) and map that back to the full-res scene/geometry somehow (bear in mind that such things never work, they can just be massaged to fail to work subtly enough that things look ok), but even with that work done, we'd need a completely separate solution for non-raytracing users not to be left high and dry. Unfortunately, like with the fog I mentioned somewhere in this thread, a specific GI solution is exactly the kind of thing that influences what the lighting team can/can't do, so having two different systems with different strengths and limitations means compromising between the two, or having some extra features that let them change the lighting to fit the GI tech (possibly doubling their workload). I'm a big flag-waver for GI on our team, but I really think we'd be better doing something that's available to all users first, maybe with one eye open for how it could be accelerated or enhanced using ray-tracing later on.

It's more likely to be a case of
"We've got a test-bed scene for a ray-based environment lighting technique, it works for one type of light so far, here's a list of things you can no longer safely have in a room without it breaking, with rough probabilities of how likely we are to work out a solution"  
The added headache is that whatever we offered would have to be an "as well" feature rather than an "instead", developing a feature for a single manufacturer's top-end cards means also maintaining feature parity for everyone else's hardware.

Personally I'd like to use it to make the shadows crispier.

Sorry yes, I should have been clearer. By "as well" I meant we'd have to maintain support for two modes:
1) A mode that uses raytracing (though probably not exclusively) in some way to get its results
2) A mode that doesn't use raytracing, but can achieve a similar look without changing the art assets or lighting setup.

As an example, when we ported over the voxel fog, we intended for it to be an "as well" feature because it's expensive to run, but eventually ran with it as an "instead" feature because it interprets the fog parameters totally differently, and drastically changes what does/doesn't work in terms of light placement.

Indeed not, that's one of the "improvements in looks" that I'd expect to get out of it. The "headache" part is writing all the management code that will be needed, eg. to calculate lighting/shadowing on objects that aren't on screen.

https://www.robertsspaceindustries.com/spectrum/community/SC/forum/50259/thread/rtx/1915059

Yes, that is quite a nice upgrade. But you're comparing a VERY BAD non-RTX presentation, to their RTX one. No AAA game would have such a rubbish 'standard' lighting implementation, and at the same time I'd argue that the result in the better looking image could be done with non-RTX solutions.

I am not ure about their standard GI solution yet but it does not look good. Witcher III had no GI at all.

 

[Shifty Geezer]

No. Because graphics suffer from the phenomenon of diminishing returns.

To iterate my response to LB above, next-gen lighting isn't a case of diminishing returns. When RT 'gets real', games will lose their incohesive lighting and gamey look. Things like no contact shadows on sleeves and faces that aren't lit properly and whatnot that still plague these RTX games. We've seen examples of next-gen lighting in various demos over the past years and are well aware of what to expect. nVidia has also promised that in their RTX promotional materials:

Ray tracing has been used for years to pre-render lifelike worlds in movies. But until now, it has been too computationally demanding to be practical for real-time, interactive gaming, which requires fast frame rates and low latency. NVIDIA RTX overcomes those limitations.

With these capabilities, developers can create realistic, high-quality reflections that capture the scene around it and achieve physically accurate lighting and shadows. Making these capabilities available on an industry-standard platform like Microsoft DXR means every PC game developer will have access to levels of realism never before possible.​

...so after raising everyone's expectations, it shouldn't be unexpected when responses to real-world gains are muted. Metro is not physically accurate lighting and shadows or particularly lifelike. Maybe an RTX game will achieve life-like lighting? That'll be the game that impresses those of us looking for next-gen lighting. These marginal improvements currently on show aren't at all exciting.

 

[Shifty Geezer]

I can spot the differences in Metro quite easily. It appears more movie like with enebaled RTX.

Great comparison, but the RTX off shot is terrible by modern game standards. Far better is possible, making the gains from RTX versus other games muted.

Also, RTX isn't solving a whole host of issues, as raytracing is supposed to. The chair's floating 6 inches off the ground according to its shadow, while the glass/plastic/alien material bottles look like holes in the scenery. This isn't the raytracing we've been shown in the Star Wars demo or Pica Pica; not even close.

 

[Jupiter]

The Metro games have never been technical pioneers for me but it's RTX GI is better than the GI of any other videogame. Other video games also have many floating objects. There is also a lot of adjustment thing. For example, the VXAO was too strong in Rise of the Tomb Raider.