Support NeoGAF

[Myuni]

Nah, we saw what RTGI did for Ghosts 2. Not much. There are unreal engine 5 games with software GI that have vastly better lighting.

DS2 needed a complete lighting overhaul. It needed way better geometry, materials, draw distance, volumetric effects, and foliage. Slapping RTGI on these visuals would result in slightly better AO, and maybe some orange hue in the surrounding rock areas. They need better more reflective materials when building the ground level textures, rocks, and other pebbles and smaller rocks on the ground to have RTGI bounce off lighting that gives it that next gen look.

Here are a couple of examples that i could find. Expedition 33 and Avowed arent even the top tier UE5 games, but you can easily see that the lighting doesnt appear flat. In Exp 33's world you see the volumetric fog inherit the red/pink hue of the trees and foliage. In Avowed, you see the wooden planks and ships bounce off sunlight making that entire environment pop in ways you just dont see in DS2 even with direct sunlight.

The only time DS2 impresses is when you get to the mountains, and their snow shaders properly bounce off the sunlight. The rock textures they use up on those snowy mountains are also way better than the ones used in the rest of the open world. Resulting in some of the best visuals ive seen this gen. But thats like 20% of the game. That along with that fireworks level, and character models in cutscenes that do look next gen probably means the game deserves an honorable mention, but thats it.

The Avowed gif looks great, and I get what you mean.

But Exp33?
To me that's a blurry mess (maybe 'cause I don't comprehend the art style).
The vegetation feels very artificial. Grass does not cast shadows, no AO at all. It's somewhat disconnected from the terrain, lighting is just wrong/nonexistent there.

 

[FalconPunch]

What I want to know is why F1 25 on PC with path tracing has just been forgotten from all graphical discussions. My own screenshots:

The environmental assets, crowd, etc all look bad. The car's on track and the driver look great but the game still fundamentally looks last gen with rt bolted on.

 

[analog_future]

So nothing ever looked really good prior to the introduction of RTGI ?

Sorry but I'm just tired of this false argument that just because a certain technique exists, that it should be mandatory to employ it, no matter how well other existing solutions perform or the cost of the implementation.

Global illumination methods have been around forever. Anything you can calculate real-time can definitionally be pre-calculated and stored for rapid retrieval. Generally the only thing lost through pre-baking is flexibility and precision, but again, RT GI has its own drawbacks particularly in terms of computational cost.

Insisting it should be employed in everything "just because" it can improve results is missing the whole point of the exercise: To create a convincing illusion.

There needs to be cost/benefit analysis to justify implementation, and a lot of the time the benefit isn't that great. Especially because for that precision to shine you need to have the assets and very hq materials system to showcase it.

Visuals in the past have looked appropriate for their time and hardware. Just like how textures/physics/animations/sound quality/fluid simulations/volumetric effects/etc..etc.. have advanced over time, so has lighting.


Baked lighting will never be able to touch all of the different advantages that RTGI brings. That's kind of the whole point. Yes, you can pre-calculate RT lighting on a limited basis, but you lose all flexibility and dynamism which is half the point of playing a videogame; feeling like you're actually in the environment and that you can manipulate and interact with it.

 

[SlimySnake]

So nothing ever looked really good prior to the introduction of RTGI ?

Sorry but I'm just tired of this false argument that just because a certain technique exists, that it should be mandatory to employ it, no matter how well other existing solutions perform or the cost of the implementation.

Global illumination methods have been around forever. Anything you can calculate real-time can definitionally be pre-calculated and stored for rapid retrieval. Generally the only thing lost through pre-baking is flexibility and precision, but again, RT GI has its own drawbacks particularly in terms of computational cost.

Insisting it should be employed in everything "just because" it can improve results is missing the whole point of the exercise: To create a convincing illusion.

There needs to be cost/benefit analysis to justify implementation, and a lot of the time the benefit isn't that great. Especially because for that precision to shine you need to have the assets and very hq materials system to showcase it.

Well, of course you can have great lighting with pre-baked solutions but as we've seen time and time again, the games are just too big nowadays to get every single room and area looking perfect. You have screenshots of Uncharted 4 posted up there that still look next gen, but then you dont have a single room in TLOU2 that looks that good. In fact, you have dozens if not hundreds of interiors in those large TLOU2 levels that look a generation behind that U4 prologue. This despite TLOU2 shipping at a whopping 79GB.

So what happened? Did their baking method fall apart? Did they run out of space holding all that lighting data in the textures? After all, the baking process is getting the highest quality lighting baked into the textures so why does it look so low quality? Why would they bake in low quality lighting data? They wouldnt unless they were limited by either file size, baking process, or simply poor materials that dont respond well to whatever lighting model they have.

As games get larger, it will become harder and harder to get baked lighting just right. It will work for some scenarios, and if its a first party polished title like a Sony game then it might even work for most scenarios, but as we saw in TLOU2, GOW Ragnorak, HFW, and DS2, it just cannot retain that level of quality in every area leading to some dated visuals.

I dont even want RTGI on all games. Especially on these old consoles. I completely understand the high computational cost. Just do software based realtime GI. I first saw it in Starfield and was blown away. Stick with software based lumen, and leave RTGI for PCs and Pro consoles.

 

[Gamer_By_Proxy]

AC Shadows and BF6 look damn good maxed out on PC. It makes me wanna play BF6 when I'm trying to finish other games.

 

[RPS37]

Couldn't finish watching, there's gotta been better looking games this year.

 

[nemiroff]

So nothing ever looked really good prior to the introduction of RTGI ?

Sorry but I'm just tired of this false argument that just because a certain technique exists, that it should be mandatory to employ it, no matter how well other existing solutions perform or the cost of the implementation.

Global illumination methods have been around forever. Anything you can calculate real-time can definitionally be pre-calculated and stored for rapid retrieval. Generally the only thing lost through pre-baking is flexibility and precision, but again, RT GI has its own drawbacks particularly in terms of computational cost.

Insisting it should be employed in everything "just because" it can improve results is missing the whole point of the exercise: To create a convincing illusion.

There needs to be cost/benefit analysis to justify implementation, and a lot of the time the benefit isn't that great. Especially because for that precision to shine you need to have the assets and very hq materials system to showcase it.

Yeah, great visuals for its time existed before RTGI, and no technique is mandatory. But this is not about quality or cost/benefit for static scenes,it's about viability for dynamic worlds in ~2025.

Once we introduced stuff like dynamic time, destruction, real-time animation of all kinds of assets across big maps, RTGI became a necessity, not choice. ...At least if a contemporary look and feel is a priority at all

Pre-calculating illumination becomes hard to manage when the light sources and geometry are constantly changing. Not to mention the massive storage needs and the development slowdown of having to wait days or weeks to re-bake a high fidelity map just to move a bush or something.


Anyway, kinda unrelated to your post, but cool illustration nevertheless:

Explanation from Gemini:

Spoiler

Curve	Name/Concept	Definition & Trend
Green Curve	Baked Lighting (Pre-calculated GI)	Represents the peak viability of the older technique. It was highly effective for static worlds in the PS3/PS4 era but dramatically loses viability as game worlds become dynamic and too large (data/bake time problems).
Blue Curve	Real-Time Global Illumination (RTGI/Path Tracing)	Represents the technical potential of the modern, dynamic lighting technique. Its potential has always existed, but its viability was low early on due to high computational cost. Its rise is continuous as the technique is optimized.
Orange Curve	Consumer Hardware Power	Represents the availability of consumer hardware capable of efficiently running the new techniques (primarily dedicated Ray Tracing hardware on modern consoles and PCs). It peaks as this technology becomes widespread.

The downward slope illustrates the idea that while new GPUs will always be faster, the performance ceiling for high-fidelity RTGI becomes less demanding relative to the total hardware power.

 

[FeastYoEyes]

Not much that impressed me this year. Avowed looked incredible but even after upgrading my PC, it performs like crap and stutters like crazy. I really hope next year comes out swinging in a way this year never reached.

Routine looked incredible though.