Digital Foundry Article Technical Discussion [2021]

Status
Not open for further replies.
RT will also present some issues. If your game is aiming for a particular lighting aesthetic then game directors, much like lighting directors in movies, will be fighting realistic light using diffusers and colours that mute reflections.
If we are able to simulate light, wouldn't they be able to just use diffusers and filters to light a scenes like a movie set? I'm not sure this would be more work than manually placing every light in a scene already is. In fact, it's probably still easier.
 
me 2. Lol. Gonna rely on the consoles for heavy lifting. Not sure if I can afford a new GPU right now

I'm hoping these crypto crippled Ampere's that are incoming will bring the prices back in line with MSRP's in the nearish future so that I can afford either a 3060Ti or if I'm feeling flush a 3070. I'm holding off the big money for the 4xxx generation!

They also mention in a FAQ that the FidelityFX Super Resolution isn't compatible with their rendering technique: https://www.metrothegame.com/news/the-metro-exodus-pc-enhanced-edition-arrives-may-6th/

Some pretty interesting insights there about FidelityFX Super Resolution. For one they appear to have access to it suggesting that it's real and in an advanced stage of development (this may have been common knowledge already, I'm a bit out of the loop on this) and secondly they claim their own Temporal upscaling implementation is at least as good. That may suggest FFX SR is just a (fairly common now) temporal upscaling solution and in quality terms falls quite short of DLSS as per Alex's video. Still great to widen the reach of upscaling tech though even if the quality isn't quite on par with a ML based approach.
 
Some pretty interesting insights there about FidelityFX Super Resolution. For one they appear to have access to it suggesting that it's real and in an advanced stage of development (this may have been common knowledge already, I'm a bit out of the loop on this) and secondly they claim their own Temporal upscaling implementation is at least as good. That may suggest FFX SR is just a (fairly common now) temporal upscaling solution and in quality terms falls quite short of DLSS as per Alex's video. Still great to widen the reach of upscaling tech though even if the quality isn't quite on par with a ML based approach.
A large number of implications from that FAQ here.
a) Firstly the most important being that it exists for sure and is close enough to completion that they could have integrated it, or have previewed the technology enough to determine whether it could be integrated.
b) it being incompatible suggesting that the technique requires some pipeline customization to support, in the same vein, if it was similar to DLSS (in the way it is slotted) that may have worked, but we know that doesn't work.
c) it's still not entirely clear what FSR does yet, or what the base technology of FSR is quite yet; but it's entirely possible that FSR is not something a developer could replicate in house necessarily and thus would need to extend support for FSR.
 
What is that white speaker-like thing? :p

Spotlight, obviously - it's a requirement for every small office space. I need it when I lose my pen. :p

I remember playing Doom 3 for the first time and being amazed at the stark shadowing, but yeah, it's not actually how light illuminates spaces. The bounce lighting in Metro Exodus looks like it's a first step in changing how games can be developed. Would be interesting to see a game designed around this type of GI and using it as a function within the game space.
 
Last edited by a moderator:
Spotlight, obviously - it's a requirement for every small office space. I need it when I lose my pen. :p

I remember playing Doom 3 for the first time and being amazed at the stark shadowing, but yeah, it's not actually how light illuminates spaces. The bounce lighting in Metro Exodus looks like it's a first step is changing how games can be developed. Would be interesting to see a game designed around this type of GI and using it as a function within the game space.
ah, I was going for heater, but I saw heaters on the wall, so I wasn't so sure.

Yea, games that move to this model can have much more dynamic environments and story telling methods. Shifting planes, falling into caverns, blowing light holes through caves, or tunnels, pipes etc. You can definitely play a lot more in the 'dark' now with realtime GI that does this.
 
ah, I was going for heater, but I saw heaters on the wall, so I wasn't so sure.

Was going for some kind of speaker or alternatively infrared light. A heater wouldnt or shouldnt really be needed seeing the section-style water radiator under the wall :) Looks like a nice cozy office space.
Anyway, the drawers do remind me of Spitfire wings :p
 
If we are able to simulate light, wouldn't they be able to just use diffusers and filters to light a scenes like a movie set? I'm not sure this would be more work than manually placing every light in a scene already is. In fact, it's probably still easier.

With a lot more control than any movie set, too. You could set how light bounces per material, without affecting that material visually. You could use a diffuser or blocker that was invisible to everything except light propagation. Etc. Definitely easier for the artists, although there may be a learning/process/tools curve.
 
If we are able to simulate light, wouldn't they be able to just use diffusers and filters to light a scenes like a movie set? I'm not sure this would be more work than manually placing every light in a scene already is. In fact, it's probably still easier.
Once you toss in RT with PBR, you're having to 'con' the lighting system to get the desired look.

This is what the lighting director is responsible for, they achieve impossible lighting aesthetics; bright (unrealistic) lights here, gloomy (unrealistic) lights here, using alternative prop materials to over or under reflect ambient light. This only works because the scene is contrived is such a way that the camera cannot glimpse all these tricks but in most games the player has full agency over the camera.
 
D4WrTa4.png


8yBBJuK.png


Gh8ywh4.png

It's interesting looking at the top two comparison images.

The 2019 screenshot has very dramatic looking lighting that instantly reminds me of Cinematography lighting that is used extensively in Movies. The faces of each character for example, are well lit so that the viewer can easily see their expressions which are often key to a Films conveyance of meaning in enhancing any dialog, or even if there isn't any dialog.

The 2021 screenshot has much more natural looking lighting. Again, focusing on the faces, they are dimly lit with key facial features less easy to see. Nuanced facial expressions (if used) would be difficult to see and would be easily missed. Basically something you'll almost never see in dimly lit scenes in movies.

It'll be interesting going forward. After the novelty of RT wears off, will developers be faced with a choice of choosing not to use RT if they want dramatic cinematographic lighting without it adversely affecting the rest of the scene if the player moves the camera? Will techniques be found that allow for easy manipulation of RT lighting to enable dramatic cinematography lighting effects? Will we see continued use of conventional lighting techniques combined with RT lighting in order to achieve cinematographic lighting?

In the past, I've always gravitated towards a realistic presentation (60 FPS versus 30 FPS, for example) versus a "filmic" presentation. But when it comes to lighting, I suddenly find myself conflicted where before looking at those two screenshots I was absolutely certain that I would prefer the more realistic presentation. I'm extremely conflicted as to whether I would prefer the more natural RT lighting (the environment looks better) or the more "filmic" non-RT lighting (where the characters faces are much more pleasingly lit).

Hmmmm... I'm so torn.

Regards,
SB
 
Last edited:
@Silent_Buddha They can just add invisble light sources to do that kind of stuff, and if they're ray traced they'll look even better.

However, that has the potential to negatively impact the rest of the scene if it uses an RT lighting model.

Like in movie they use artifical light source.

Unlike a movie, a player moving their camera can show unintended side effects of altered RT lighting meant to light a specific feature as Dsoup noted in a previous post.

For example, the dramatic lighting used on an actors face in some horror movies where the light is below the actors face shining upwards. Works in a film because the viewer can't change the camera. But in a game? What if the player wants to look around and suddenly sees the ceiling of the room lit up?

Or, more problematically the bounce lighting from that ceiling potentially lighting up the rest of the room?

Regards,
SB
 
Just a refresher...
pyHfJQT.png


Everything but reflections on the RDNA2 consoles puts this slide into a different perspective:


00.png




Some pretty interesting insights there about FidelityFX Super Resolution. For one they appear to have access to it suggesting that it's real and in an advanced stage of development (this may have been common knowledge already, I'm a bit out of the loop on this) and secondly they claim their own Temporal upscaling implementation is at least as good. That may suggest FFX SR is just a (fairly common now) temporal upscaling solution and in quality terms falls quite short of DLSS as per Alex's video. Still great to widen the reach of upscaling tech though even if the quality isn't quite on par with a ML based approach.

FSR has been in complete secrecy from the public and there's no info other than AMD banking on it to become an "industry standard" (open source + available to consoles + effective on AMD and Nvidia hardware).
Which makes that statement about their temporal upscaling being "equal or better" than FSR a pretty odd and bullish choice of words for a public statement.
I don't think AMD is happy about that choice of words and I wonder if 4A had access to the final FSR code to make a plausible comparison.

OTOH, like all Metro titles this is a TWIMTBP/Gameworks title whose RT implementation was probably tuned and funded by nvidia's developer program (at least partially), so maybe 4A doesn't care whether or not they get at odds with AMD.
 
However, that has the potential to negatively impact the rest of the scene if it uses an RT lighting model.



Unlike a movie, a player moving their camera can show unintended side effects of altered RT lighting meant to light a specific feature as Dsoup noted in a previous post.

For example, the dramatic lighting used on an actors face in some horror movies where the light is below the actors face shining upwards. Works in a film because the viewer can't change the camera. But in a game? What if the player wants to look around and suddenly sees the ceiling of the room lit up?

Or, more problematically the bounce lighting from that ceiling potentially lighting up the rest of the room?

Regards,
SB

I'm sure they can come up with ways to have the light only affect a particular character etc.
 
I'm sure they can come up with ways to have the light only affect a particular character etc.
Ask yourself, how? How do you light a scene realistically using RT while also aiming to achieve a particular unrealistic lighting aesthetic whilst also allowing the player to freely move the camera around to see the conventional lighting tricks because suddenly hiding lighting impacts the lighting you're seeing as you pan the camera around?

If you can solve this, you could be the greatest Hollywood lighting director of all time. :yes:
 
Ask yourself, how? How do you light a scene realistically using RT while also aiming to achieve a particular unrealistic lighting aesthetic whilst also allowing the player to freely move the camera around to see the conventional lighting tricks because suddenly hiding lighting impacts the lighting you're seeing as you pan the camera around?

If you can solve this, you could be the greatest Hollywood lighting director of all time. :yes:

During cutscenes they can add, remove lights however they want and hide them just like you would in a movie. During gameplay "cutscenes" I don't see why they couldn't use invisible lights to highlight particular things like faces. Prevent the light from contributing to GI, prevent the light from lighting anything other than the particular objects they want. Sure, someone that's really looking for it might wonder where that light is coming from, but if it's tastefully done they probably wouldn't even look for the incongruity in the lighting ... except maybe @Dictator
 
During cutscenes they can add, remove lights however they want and hide them just like you would in a movie.
They can cut from camera to camera in a scene in a movie but not in a video game where the player has agency over the camera. If you're paining a camera around the lighting is changing because your looking at an invisible light source, how weird is that going to look? If there was a non-visible lightbulb in your room casting light it would look as weird as it sounds.

It'll look shit and less consistent that carefully place fake sources of GI in a scene.
 
They can cut from camera to camera in a scene in a movie but not in a video game where the player has agency over the camera. If you're paining a camera around the lighting is changing because your looking at an invisible light source, how weird is that going to look? If there was a non-visible lightbulb in your room casting light it would look as weird as it sounds.

It'll look shit and less consistent that carefully place fake sources of GI in a scene.

In cutscenes you don't normally control the camera or have very limited control of the camera.
 
Status
Not open for further replies.
Back
Top