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

only Minecraft beats it.

I think Q2 has the better tech, but another likes minecraft more. Everyone has opinions, they differ very much apperently.

I dont think we can say Q2 had a 10 year dev time, cant compare it to some aaa studio. As rtx came last autumn, i see it more as a one man project with a small budget.

Edit: what happened to ttc? I played it for free some years ago but never heard of it again untill today.
 
I would easily classify this scene as wrong and unrealistic, as well as ugly:
Agree, but with RTX on an a single bounce it is still much to dark, and AO makes this even worse. Overall surely RTX looks MUCH better. But it does not make it realisic, just less bad to me. In such a bright scene with snow everything should be pretty bright.
To fix this, one could use a grid of cheap 2 band spherical harmonic probes over the ground, and the result would endup better, because you can precompute infinte bounces. No need for RTX.
So the GI is not really that revolutionary here. It supports dynamic geometry, but can not utilize this advantage. Dynamic GI is interesting mostly for interiors and moving light sources, etc. But for outdoor you could just precompute, even for open world. (Farcry does something like this in the background.)
Personally i like most how AO fixes the missing occlusion probe reflections problem, which i consider the worst artifact of current games.
It also gives more depth, which is great.

You really cant compare two different titles, they are going for totally different styles.
I compare them figure out DXRs best application. Even if GI is most important for me personally, it is not impressive to regular people i think.
BFV has never been seen before, but Metro seems too subtle.
 
Again. If the RTX ON or RTX OFF weren’t there (I had to check) then I wouldn’t know which is which, except one is darker and one is brighter, which can be easily done in other ways. Meh.
I guess without getting too vested here like yourself, I would assume that you’re mind would recognize it at being more correct.
 
Last edited:
Voxel cone tracing TTC. A game with the complexity of minecraft/fortnite, in a voxel space. Game apperently discontinued 2017.

Without

TC%207.jpg


With

TC%206.jpg



Just brighter, can be done other ways. Meh.
 
Last edited:
I think Q2 has the better tech, but another likes minecraft more. Everyone has opinions, they differ very much apperently.
It's not a matter of opinion. Quake has only one bounce, Minecraft has infinite bounces.

However, i have an idea: Quake can show all mirrors scene, which is int the code with a #define, and in DF video i saw it can be turned on/off with console command.
The number of bounces is also set with a #define!
It is thus likely you can increase bounces with console command too! Anyone tried?
 
Just brighter, can be done other ways. Meh.
... same we complain for Metro. GI is really not grateful to sell ;)

But TTC can do this indoors as well, with moving area lights. And it has reflections for proper metal. It is is not restricted to voxels as you can see. It only uses them for GI. (which has issues, but skylight only is a much larger issue than some leaks!)
That's what i wanted to point it out. I leave it your imagination how TTC would scale to a modern high end GPU.
Maybe somehow like this, which can compete here as well:
 
...they say the woods here look so good. Can't tell myself, because i can't stop to focus just that horse head :D
 
... same we complain for Metro. GI is really not grateful to sell ;)

But TTC can do this indoors as well, with moving area lights. And it has reflections for proper metal. It is is not restricted to voxels as you can see. It only uses them for GI. (which has issues, but skylight only is a much larger issue than some leaks!)

Exactly, any game can be faulted, thing is that TTC doesnt even come close to Metro's complexity. Its in a voxel space too, cutbacks where made for performance.

Kingdom cone looks rather nice yes, will look into it.

 
That's what i wanted to point it out. I leave it your imagination how TTC would scale to a modern high end GPU.

Yes a modern high end gpu utilizing both hw acceleration and compute, with some form of reconstruction and perhaps vrs to scale performance, killer apps are possible. In special if RT isnt bolted on as an afterthought.
 
killer apps are possible. In special if RT isnt bolted on as an afterthought.
Killer apps are always possible, independent of hardware progress.
It will take 5 years until RT can be targeted exclusively, because BC, and assuming other vendors follow. RTX will be better then i hope. If so it's also thanks to some people trying to spot its limits as early as possible.
 
Other vendors will follow, amd probably has something in the works, successor to navi perhaps. Intel isnt that dumb either, they researched RT many years ago?

And yes i agree, we need more flexible solutions that scales well with performance. RT needs to be in lower end hw too, hopefully amd helps us there. Will take couple of years but install base has grown by then.

But just now, we need hardware acceleration.
 
Last edited:
TTC was designed with realtime GI in mind, Metro was not, it's just an extra effect.

The people saying they should have baked probes are missing the point. With ray tracing you no longer have to. All that complexity is suddenly gone. It's a big win.
 
The people saying they should have baked probes are missing the point. With ray tracing you no longer have to. All that complexity is suddenly gone. It's a big win.

When you can shoot many thousand rays per pixel you can do away with probes/voxels and etc. We are far away from that holy Grail.
 
I found these images in the documentation for V-Ray, and I'm using them to illustrate the importance of many-bounce lighting.

The link is: https://docs.chaosgroup.com/display/VRAY3MAX/Global+Illumination+Rollout

The first image Is post shows 1 bounce of GI. That is as far as Metro is going (and probably at much worse quality, but that's not my main point). The second image shows just 3 bounces of GI and the difference is very big. On the page itself they also show 0 and 8 bounces, but the difference from 3 to 8 is not as great as from 1 to 3, because lighting becomes much fainter at each successive bounce.
Anyway, in Metro, RT is doing a much more spacially precise job than probes would do at achieving the 1 bounce exemple, while lightprobes, get you a less spacially detailed approximation, but they can aim to approximate the 3 or 8 or 1000 bounce version, it just depends on the devs willingness to let the light baking solution process that much.
So, even if they can do 1 bounce with RT, they still need probes to get the other bounces, otherwise they loose too much indirect lighting on shadows.IrradianceMapOnly.jpeg3bounces.jpeg
 
The people saying they should have baked probes are missing the point. With ray tracing you no longer have to. All that complexity is suddenly gone. It's a big win.
I say probes would give similar GI quality because i would be disappointed from Metro not looking better than previous GI approaches with lesser hardware requirements.
If i would consider to upgrade GPU, i would expect significantly better visuals than ever before, after seeing Star Wars.

It's very difficult. Current games can show very impressive trailers, so NV had to show something like Star Wars to beat this.
But the few games do not look thaaat good, so likely people decide to wait with the upgrade until more games show, or until HW becomes cheaper.
At the same time people may even decide not to buy the games, because they want to wait until they have the HW to enjoy all features.
Gamedevs so also have a difficult decision if RTX is worth it or not, even if UE4 and Unity make it easy.
It will take some time...

But the promise of dynamic lighting is not to save work on probe placement or baking time. That's really a bad argument gamers do not want to hear.
The proper argument is the opportunity for new level design, or even new options for gameplay.
And that's exactly what's not possible until everyone has the hardware.

It is also a bad argument for developers, because they save no work by adding RTX support - it's just more of it.
And they know complexity is not going to decrease while performance is finite.

NV might consider to buy a game studio... but secretly, to prevent another stupid shitstorm ;)
 
Then why does it look like ass?
That's a ridiculous thing to say. Whether you like the sparse art-style or not, the visuals were clearly 'next-gen' in approach. The lighting and shading were world's apart from every other game.
The only good thing with that game is it (was) free a while ago. And for good reasons perhaps.
This comment suggests you've no real interest in technology. TTC was a technological marvel with the most beautiful lighting of any game yet that I know of.


Q-Games really explored a new approach and it was fabulous, and they deserve the kudos for achieving something simply wonderful (even if the game stank).
 
I do not think GI in tomorrow children has glossys refleciton support, I thought it was just diffuse? The reflections I think are planar reflections with some neat filtering, I asked once on twitter to Qbert. Let me see if I can find that tweet...
 
When you can shoot many thousand rays per pixel you can do away with probes/voxels and etc. We are far away from that holy Grail.
Probes give very low detail results. DXR in Metro is already far better. Not only does it handle dynamic objects, it also supports occlusion. Indirect shadows, oh yeah. They could have even pushed for more bounces and effects but the spoiled-brat, 4k-60fps crowd would have whined about the game not being "optimized".

I hate those bastards.

DF Analysis, some of the comparisons in their video are simply astonishing:


https://www.eurogamer.net/articles/digitalfoundry-2019-metro-exodus-tech-analysis
The good stuff starts at 14:00. It's light night and day.

I say probes would give similar GI quality because i would be disappointed from Metro not looking better than previous GI approaches with lesser hardware requirements.
If i would consider to upgrade GPU, i would expect significantly better visuals than ever before, after seeing Star Wars.

It's very difficult. Current games can show very impressive trailers, so NV had to show something like Star Wars to beat this.
But the few games do not look thaaat good, so likely people decide to wait with the upgrade until more games show, or until HW becomes cheaper.
At the same time people may even decide not to buy the games, because they want to wait until they have the HW to enjoy all features.
Gamedevs so also have a difficult decision if RTX is worth it or not, even if UE4 and Unity make it easy.
It will take some time...

But the promise of dynamic lighting is not to save work on probe placement or baking time. That's really a bad argument gamers do not want to hear.
The proper argument is the opportunity for new level design, or even new options for gameplay.
And that's exactly what's not possible until everyone has the hardware.

It is also a bad argument for developers, because they save no work by adding RTX support - it's just more of it.
And they know complexity is not going to decrease while performance is finite.

NV might consider to buy a game studio... but secretly, to prevent another stupid shitstorm ;)
Tech demos and actual games are different things. Obviously the first will look better than the second. And of course different arguments will be made to different audiences. Some people were saying how devs could fake the effect so I addressed it from a developer's point of view. Previously I've also made the argument you're making also, about the new experiences it brings to consumers.

Probes suck and should only be used when performance is critical (or in open fields).

Right now you have the effects shown in the demos split among different games: reflections on BFV, GI in Metro, shadows in Tomb Raider. It's just a matter of time before a game mixes them all. And with UE4 adding support for DXR in its upcoming version this will happen sooner rather than later.
 
I do not think GI in tomorrow children has glossys refleciton support, I thought it was just diffuse? The reflections I think are planar reflections with some neat filtering, I asked once on twitter to Qbert. Let me see if I can find that tweet...
The ground reflections are definitely plannar. But I think everything else still relies on their voxel representation of the scene.
 
Back
Top