Support NeoGAF

[Soi-Fong]

ROTR is really limited you can tell by the Xbox One. There's just so much compressed textures in the whole game. I don't think any of the non-compressed shots are particularly striking as well. I feel like the roughness maps aren't being used to their full effect in the game to give some oomph to some of the materials.

I've been working with PBR for a while now and it really does constantly amaze me. The Substance packages are just glorious combined with 3DSMax.

 

[DocSeuss]

It's legit surprising to me that people don't realize most of the reason Battlefront looks any good is because of its use of photogrammetry.

 

[pottuvoi]

It's legit surprising to me that people don't realize most of the reason Battlefront looks any good is because of its use of photogrammetry.

No, it looks good because those involved in the project did excellent job on it.
As I have said earlier, you need a lot of work on most assets after the scan is done, scan is just the first pass on the object or surface.

 

[Stiler]

No, it looks good because those involved in the project did excellent job on it.
As I have said earlier, you need a lot of work on most assets after the scan is done, scan is just the first pass on the object or surface.

Photogrammetry just gets so many details though, yes it still has to be gone over so they can actually fit it into the game (The size/detail of the assets would take up way too much space on disk) but you can't honestly say it's not a huge part of why the game looks so good.


it's exactly how the Vanishing of Ethan Carter was able to look so great for an indie game with some of the most detailed textures of a game.

 

[pottuvoi]

Photogrammetry just gets so many details though, yes it still has to be gone over so they can actually fit it into the game (The size/detail of the assets would take up way too much space on disk) but you can't honestly say it's not a huge part of why the game looks so good.


it's exactly how the Vanishing of Ethan Carter was able to look so great for an indie game with some of the most detailed textures of a game.

It's a great tool, but far from fire and forget solution, especially with a PBR pipeline where you must take all the lighting out from your color map.

 

[shacklesmcgee]

The Substance packages are just glorious combined with 3DSMax.

mmmmmm Substance

This just won the Allegorithmic substance contest....it's an entirely procedural (except the letters) material on a flat plane. Just amazing, insane quality. Oh, and the best part.... it's free!
More shots/other winners here

 

[pottuvoi]

mmmmmm Substance



This just won the Allegorithmic substance contest....it's an entirely procedural (except the letters) material on a flat plane. Just amazing, insane quality. Oh, and the best part.... it's free!
More shots/other winners here

Impressive.

 

[Dictator93]

Photogrammetry just gets so many details though, yes it still has to be gone over so they can actually fit it into the game (The size/detail of the assets would take up way too much space on disk) but you can't honestly say it's not a huge part of why the game looks so good.


it's exactly how the Vanishing of Ethan Carter was able to look so great for an indie game with some of the most detailed textures of a game.

Yet you can see in EC how the image only looks completely right from one angle. Far from being anywhere near PBS-compliant as other pipelines would define it (i.e. looking physically plausible under different conditions and with different materials).

It's a great tool, but far from fire and forget solution, especially with a PBR pipeline where you must take all the lighting out from your color map.

Exactly, and that is a lot of work! Unfortunately, like the term PBR/PBS itself, photogrammetry has already become a buzz word :/

 

[brainchild]

Yet you can see in EC how the image only looks completely right from one angle. Far from being anywhere near PBS-compliant as other pipelines would define it (i.e. looking physically plausible under different conditions and with different materials).

Exactly, and that is a lot of work! Unfortunately, like the term PBR/PBS itself, photogrammetry has already become a buzz word :/

So true.

Not even perfectly sound terminology is safe from being botched by the evolution of language.

 

[veidi]

No forza yet?
Doesn´t Turn10 also use PBR in in FH2 and Forza 6?
The always talked about physical based materials for wet and dry conditions.

 

[2+2=5]

Innovations like PBR are what make graphics feel "next gen", the hardware power alone is meaningless.

TBH i will never understand why 3d didn't evolve in this direction earlier, i mean it seems natural to me to think of material qualities instead of diffuse/normal/etc, i thought of something like PBR years ago, but i'm not good with 3d.

 

[cyberheater]

The more I think about it. The more I think Star Wars Battlefront is the best looking game on any platform right now in terms of realism.

Real life

Star Wars battlefront

So impressive.

 

[CHC]

So basically what I learned from this thread is that a game's graphics can be judged by the tarps it depicts. If a blue tarp looks good, then the artists have really done it.

 

[Slavik81]

What were studios doing before PBR?

I see some others have noticed the clumpy flat lit leaf and foliage of W3. PBR on armor is great, but that quality hardly permeates in the open environments at all.

Most leaves need translucency and subsurface scattering to look right. I'm not sure what the state-of-the-art is for real-time foliage, but UE4 has been working on some neat stuff recently.

 

[brainchild]

Innovations like PBR are what make graphics feel "next gen", the hardware power alone is meaningless.

TBH i will never understand why 3d didn't evolve in this direction earlier, i mean it seems natural to me to think of material qualities instead of diffuse/normal/etc, i thought of something like PBR years ago, but i'm not good with 3d.

Old habits die hard. Even now, there's a lot of improvements that can be made in a rendering pipeline to make the following more intuitive and uniformly implemented:

- diffuse interreflection
- indirect shadows
- environment sub-surface scattering (foliage, fruit, etc.)
- gaseous particles with order independent transparency and self-shadowing
- light diffraction
- iridescence of thin-film materials that interfere with water, ice, crystals, etc.


There's a lot to be done. Now, I'm not necessarily saying these features should be completely physically accurate, but approximations of these phenomenon would go a long way towards realism in graphics rendering. Of course, it would require yet another re-working of the pipeline, and that will require a huge initiative before others in the industry will follow suit.

 

[Slavik81]

TBH i will never understand why 3d didn't evolve in this direction earlier, i mean it seems natural to me to think of material qualities instead of diffuse/normal/etc, i thought of something like PBR years ago, but i'm not good with 3d.

I'm not an expert, but from what I've seen, the shaders are significantly more computationally intensive. The overhead of switching shaders is apparently rather high, though, so being able to use the same PBR shader for all sorts of different surface types offsets the extra computation cost somewhat.

Presumably PBR's popularity stems from the hardware tradeoffs just making sense now. I think that most of the underlying math has been around for a decade or more.

 

[brainchild]

I'm not an expert, but from what I've seen, the shaders are significantly more computationally intensive. The overhead of switching shaders is apparently rather high, though, so being able to use the same PBR shader for all sorts of different surface types offsets the extra computation cost somewhat.

Presumably PBR's popularity stems from the hardware tradeoffs just making sense now. I think that most of the underlying math has been around for a decade or more.

You can make PBR computationally cheap... it just won't necessarily look realistic

The PBR pipeline can be done on any GPU with programmable shaders. The better the shader implementation, the better the materials will look (granted the values are physically accurate).

 

[Acheteedo]

Oh god the jaggies, wrecks the entire scene, hopefully that doesn't represent the final product.

 

[dark10x]

Not all AA solutions are created equal. Apply liberal amounts of FXAA and your game will look like it's smeared with Vaseline. Apply liberal amounts of SSAA and your game will turn into a sideshow. Temporal AA is nice in motion, but screenshots reveal a different picture.

MSAA is typically the most balanced AA solution, but it doesn't play nice with deferred rendering.

AA does help image quality, but not in all cases, and you really have to be careful about how you use it, as there's always a cost.

I always see this "vaseline" argument and I just don't get it. Just referring to AA in general, it's a key element to matching advanced offline rendering. Films are soft and I tend to prefer games that attempt to duplicate this look.

That said, I do agree that Nintendo's games can work well with minimal edge smoothing due to asset creation.

Let's keep in mind that Nintendo still seems to use some sort of edge smoothing in some of its games. Super Mario 3D World uses an edge smoothing algorithm that is not particularly aggressive but it helps smooth out certain edges as you can see here. There is clearly some sort of edge anti-aliasing present here.

On the flip side, Mario Kart 8 uses NO anti-aliasing at all and actually IS an aliased mess. FXAA would actually be a marked improvement here as the shimmering and aliasing does hurt image integrity. This is a more challenging scene to render, of course, but still.

 

[DOWN]

What were studios doing before PBR?


Most leaves need translucency and subsurface scattering to look right. I'm not sure what the state-of-the-art is for real-time foliage, but UE4 has been working on some neat stuff recently.

Frostbite does an exceptional job above Witcher 3 in large environments, not only in in Battlefront but in the older Dragon Age Inquisition tech as well. Shame that Witcher 3 has such clumpy 2D foliage textures and LOD.

 

[brainchild]

I always see this "vaseline" argument and I just don't get it. Just referring to AA in general, it's a key element to matching advanced offline rendering. Films are soft and I tend to prefer games that attempt to duplicate this look.

That said, I do agree that Nintendo's games can work well with minimal edge smoothing due to asset creation.

Let's keep in mind that Nintendo still seems to use some sort of edge smoothing in some of its games. Super Mario 3D World uses an edge smoothing algorithm that is not particularly aggressive but it helps smooth out certain edges as you can see here. There is clearly some sort of edge anti-aliasing present here.

On the flip side, Mario Kart 8 uses NO anti-aliasing at all and actually IS an aliased mess. FXAA would actually be a marked improvement here as the shimmering and aliasing does hurt image integrity. This is a more challenging scene to render, of course, but still.

Because that's exactly what FXAA looks like

Look at that electronic panel in the second gif. The fact that you can even see the blurring in low resolution gifs speaks volumes. FXAA doesn't just smooth out the edges, it blurs textures as well, resulting in an overall blurry image. It just doesn't look good, and you lose image detail. And like with most post-AA solutions, if an overall image has a ton of aliasing, the resulting image with FXAA applied will be even blurrier than normal.

When AA is done right (like the SSAA that you'll find in offline rendering) it can clean up an image with minimal blurring, but if I'm being honest, I'd prefer no blurring at all, as any AA solution still results in some level of detail lost. But FXAA is just too noticeably blurry for my tastes.

Anyway, notice that I never used MK8 as an example of a Nintendo game that doesn't need AA. That game IS an aliased mess, and I'd rather it have some type of AA than not. However, it has so much aliasing that I think FXAA would just murder the image detail in the game.

Ultimately, if Nintendo continues to increase their baseline resolution with each generation, their games' need for AA will eventually vanish. The real solution to superior IQ is increasing resolution, not AA, and I do believe that Nintendo will find that solution worth pursuing, because there are gameplay benefits to a game having a super high resolution, like Pikmin 5 being able to render thousands of individual Pikmin, for example.

 

[MisterCrow]

The PBR in Fallout 4 makes everything look pretty good most of the time. The visual style of that game is real inconsistent but the PBR is real nice for sure. Halo 5 looks pretty phenomenal as well, especially those outdoors areas with cliffy rock faces and such.

 

[FunkyDealer]

My main problem with FXAA is that it can (but not always) have a large negative effect on perceived resolution and text-elements.

As for PBR; I think Rise of The Tomb Raider looks wonderful on Xbone. Especially clothes. It does at least as good a job as The Witcher 3 in my eyes.

 

[HTupolev]

The fact that you can even see the blurring in low resolution gifs speaks volumes.

The gifs are 1400x1050 and 1:1 pixel-mapped to the source render.

Halo 5 looks pretty phenomenal as well, especially those outdoors areas with cliffy rock faces and such.

The thing about Halo 5 is that it regularly suffers from some this grimy, hazy rim lighting. Halo 4 was the same way but perhaps even worse, sometimes it's almost like having Reach's night-vision mode engaged.

I think it might be a combination of 343i trying to use punched-up Fresnel and not using the area specular that Bungie introduced in Halo 3.

 

[dark10x]

Because that's exactly what FXAA looks like

Look at that electronic panel in the second gif. The fact that you can even see the blurring in low resolution gifs speaks volumes. FXAA doesn't just smooth out the edges, it blurs textures as well, resulting in an overall blurry image. It just doesn't look good, and you lose image detail. And like with most post-AA solutions, if an overall image has a ton of aliasing, the resulting image with FXAA applied will be even blurrier than normal.

When AA is done right (like the SSAA that you'll find in offline rendering) it can clean up an image with minimal blurring, but if I'm being honest, I'd prefer no blurring at all, as any AA solution still results in some level of detail lost. But FXAA is just too noticeably blurry for my tastes.

Anyway, notice that I never used MK8 as an example of a Nintendo game that doesn't need AA. That game IS an aliased mess, and I'd rather it have some type of AA than not. However, it has so much aliasing that I think FXAA would just murder the image detail in the game.

Ultimately, if Nintendo continues to increase their baseline resolution with each generation, their games' need for AA will eventually vanish. The real solution to superior IQ is increasing resolution, not AA, and I do believe that Nintendo will find that solution worth pursuing, because there are gameplay benefits to a game having a super high resolution, like Pikmin 5 being able to render thousands of individual Pikmin, for example.

I understand what you're trying to show, but I know exactly what FXAA looks like and, while it's not my favorite AA method, I definitely prefer the look of it in those examples. Especially when playing on a TV where the blurring is even less of an issue but aliasing remains highly visible.

I think we'll just have to disagree on this.

 

[Nostremitus]

It's very subtle but One Piece Pirate Warriors 3 seems to have it.

Off topic, what is that abomination?


On topic, thanks everyone for all the info. I didn't know what PBR really was before this thread. I see people bring it up but never understood what exactly they were trying to convey by doing so.

 

[tuxfool]

My main problem with FXAA is that it can (but not always) have a large negative effect on perceived resolution and text-elements.

The issues with blurred text elements are purely due to injection or terrible interface layering practices. Any game natively supporting any form of post process aa should be applying it only to the viewport, and drawing the interface at native resolution of on top.

 

[Dictator93]

The issues with blurred text elements are purely due to injection or terrible interface layering practices. Any game natively supporting any form of post process aa should be applying it only to the viewport, and drawing the interface at native resolution of on top.

The question is, what does one do at those times when the interface is rendered as a part of the world like star citizen?
It should be interesting to see what they do.

 

[Crossing Eden]

Oh god the jaggies, wrecks the entire scene, hopefully that doesn't represent the final product.

That gif is quite old. No idea what the final IQ will be but it certainly won't be like that.

 

[tuxfool]

The question is, what does one do at those times when the interface is rendered as a part of the world like star citizen?
It should be interesting to see what they do.

You deal with it and design your game around such limitations. It is also why I'd wish the SC developers get on with AA sooner than later.

 

[Dictator93]

You deal with it and design your game around such limitations. It is also why I'd wish the SC developers get on with AA sooner than later.

Yeah I wish they would soon enough as well...
You can get both TAA types working in the game, but they are not perfectly integrated with some rendering elements atm obviously.

Should be interesting to see how TAA works in Alpha 2.0 which has the UI now rendered like other transparent surfaces instead how they did it before.

 

[Harlequin]

- environment sub-surface scattering (foliage, fruit, etc.)

At least this is already being done in some instances, though, isn't it? I think I remember Ryse using SSS (or an SSS approximation) for marble surfaces, for example.

 

[brainchild]

The gifs are 1400x1050 and 1:1 pixel-mapped to the source render.

LOL

I swear, posting images from my phone is going to be the bane of my existence! On my screen, I noticed a ton of color dithering (as with most gifs) and so I figured the images were low-res. Thanks for the clarification.

I understand what you're trying to show, but I know exactly what FXAA looks like and, while it's not my favorite AA method, I definitely prefer the look of it in those examples. Especially when playing on a TV where the blurring is even less of an issue but aliasing remains highly visible.

I think we'll just have to disagree on this.

Fair enough. If you find it difficult to tolerate aliasing in motion, then I could understand why you'd have a strong preference for use of AA.

Thing is, I dislike aliasing as well, but it's less of an issue for me in motion, unless there is a ton of shimmering. However, I really prefer a crisp image over a soft/blurry one.

Now what really grinds my gears is a poor/choppy/unstable frame rate. I'd prioritize high frame rates over almost anything else. If higher frame rates mean no AA, then I will gladly choose that option. If I can have both high frame rates and AA, I'd choose that option only if it doesn't result in the game looking blurry.

As an aside, we need to move to a new standard for high frame rates, especially with the advent of VR. 90fps is a good baseline, but personally, I think we should be shooting for 120fps.

At least this is already being done in some instances, though, isn't it? I think I remember Ryse using SSS (or an SSS approximation) for marble surfaces, for example.

All of those features have been approximated in some way or another, but I was talking about making them standard features in a rendering pipeline.

Of course, I'm not saying developers should be trying to do this, but just pointing out that there's room for improvement beyond PBR.

 

[Dictator93]

At least this is already being done in some instances, though, isn't it? I think I remember Ryse using SSS (or an SSS approximation) for marble surfaces, for example.

Ryse has at least 5 types of sub suface scattering shaders. One for eyes, one for hero skin (with back lighting), one for vegetation, a screen space shader for marble, wax, and other organics with slight permeabily, and one for the ocean.

Pretty crazy.

LOL
Of course, I'm not saying developers should be trying to do this, but just pointing out that there's room for improvement beyond PBR.

Totally agreed. Beyond how these materials are lit indirectly and directly,
there need to be more complex BRDFs that allow for a lot of stuff normal models like those found in most engines still cannot account for so easily/ without a significant cost.

 

[『Inaba Resident』]

Oh god the jaggies, wrecks the entire scene, hopefully that doesn't represent the final product.

that footage is super old

 

[MNC]

How do people see when something is PBR? I thought I could tell but I can't. I thought it was pbr when it had "that realistic touch "

 

[TEH-CJ]

Does fallout 4 use this graphics rendering? I might be the only person in the wold the loves fallout 4 visuals. I think they look amazing.

 

[nOoblet16]

Does fallout 4 use this graphics rendering? I might be the only person in the wold the loves fallout 4 visuals. I think they look amazing.

It does, in quite possibly one of the worst ways possible.
So much so that not much would be lost if it was completely removed, and that's saying a lot considering how much of a defining feature PBR has been this gen for that "next gen look".

You know your PBR pipeline has issues when your wood texture is glossier than a metal texture, despite the metal being wet and wood being dry.

 

[TEH-CJ]

It does, in quite possibly one of the worst ways possible.
So much so that not much would be lost if it was completely removed, and that's saying a lot considering how much of a defining feature PBR has been this gen for that "next gen look".

You know your PBR pipeline has issues when your wood texture is glossier than a metal texture, despite the metal being wet and wood being dry.

Oh

 

[HTupolev]

there need to be more complex BRDFs

Most of the stuff in the article is problematic because of factors beyond the scope of what BRDFs are.

 

[Garrett Hawke]

Oh

Spoiler

I think the visuals are really nice too

 

[Dictator93]

Most of the stuff in the article is problematic because of factors beyond the scope of what BRDFs are.

Yeah, the shaped highlights example seems really out of place... with the term BRDF. MY BAD

But cloth, brushed metals, and clear coat all fall under that appreciation still.

 

[brainchild]

Totally agreed. Beyond how these materials are lit indirectly and directly,
there need to be more complex BRDFs that allow for a lot of stuff normal models like those found in most engines still cannot account for so easily/ without a significant cost.

That blog makes an excellent case for why Forward Rendering is still relevant. Thanks for posting it!

 

[brainchild]

Most of the stuff in the article is problematic because of factors beyond the scope of what BRDFs are.

Yeah, the shaped highlights example seems really out of place... with the term BRDF. MY BAD

But cloth, brushed metals, and clear coat all fall under that appreciation still.

Most people use BRDF as an umbrella term for light transport behavior, but what they really mean is BSDF, which pretty much covers everything for light transport.

 

[Shaneus]

I've read through most of this thread, but can't see any info on what kind of load this generates on a GPU (or CPU). Does it create any extra from existing techniques? If not, why aren't we seeing exponentially more implementations of it?

 

[brainchild]

I've read through most of this thread, but can't see any info on what kind of load this generates on a GPU (or CPU). Does it create any extra from existing techniques? If not, why aren't we seeing exponentially more implementations of it?

It really depends on the game. The more realistic the materials, the more complex the shader implementation. The pipeline in and of itself isn't inherently computationally expensive.


As for why PBR is just now becoming a thing, it really has to do with being efficient with development while increasing realism in games. Using the traditional workflow, in order to get surfaces with a similar level of fidelity to realistic PBR materials, it would require a lot more custom work, and potentially could increase the computational load of the GPU.

Basically, the traditional workflow was working fine for realism until developers realised that it would be too much work to increase realism with the traditional workflow. PBR was invented solve this problem.

 

[Shaneus]

Hm, that's very interesting. Glad it's now becoming a thing, especially considering one of my favourite games currently (Driveclub) seems to utilise the hell out of this method.

 

[astonish]

It's also worth noting that the PBR style of rendering is also relatively new in the film world as well.

The UE4 guys have straight up said it was a disney/pixar presentation about the switch over to PBR for Monsters University that inspired them to make the switch over very late in the engines development (around the time when the SVOGI was being removed)

The common PBR mentality should be interesting for technique and even some content sharing between the film world and the gaming world. While film often use a more complex model (with subsurface scattering more prevalent) and are also almost exclusively ray tracing from what I understand, there is still plenty of overlap

 

[Shaneus]

Is there a chance PBR could be implemented in Unity as well? I'm thinking now that when people say something looks like it was made in Unity it might be because it doesn't have PBR.

 

[brainchild]

Is there a chance PBR could be implemented in Unity as well? I'm thinking now that when people say something looks like it was made in Unity it might be because it doesn't have PBR.

Unity 5 has a PBR pipeline. It even has real time global illumination.

Maybe those people are referring to the older Unity engines.