The Holy Grail of 3D : Solving Current Limitations or Improving Upon Believability

[Nom De Guerre]

First of all, I hope everyone agrees upon the surmise that 3D computer programming in general has but one aim -- approaching realism.

Much of what we have witnessed thus far in the field, and industry, of 3D research and implementation can pretty much be summed up by (or lumped into, depending on your choice of words) these 2 categories (the two mentioned in the thread's title).

3D R&D per se cannot possibly and absolutely solve current limitations that are inherent in programming. Until some breakthroughs happen with regards to the whole sub-system (or even tenets) of programming, 3D will be dependant upon technological improvements in the central processing unit, memory, display and the whole plethora of involved components.

At the rate that the entire programming system (involved in the evolution of the creation of pixels to displaying the same) is going, where individual components of a whole (and perhaps "rigid") system are dependant on each other, which (of the two choices in the title of this thread) would be your preference?

For example (in case you don't know what I'm on to) :

It is my opinion that shaders in themselves allow us to be "impressed with 3D graphics" instead of actually being "impressed that they make 3D graphics that much more real". I hope you get what I mean.

It is also my opinion that extremely high resolution textures (a word, "texture" that is, that hopefully will be surpassed in general meaning and definition) play a much greater role in approaching "realism" than shaders do.

These are just examples. Instead of focussing on the difference between the importance of high resolution textures versus comlex shaders, I'd appreciate thoughts on the two choices I mentioned in the title of this thread.

 

[Cypher]

*Cliffs notes of the OP* Which is more important to you? Kickass shaders or kickass textures?

Kickass textures, imo. Some of the most impressive graphics I've seen (HL2, Ysaneya's Infinity engine over at GDNet, Quake Wars) have been due to extremely good texturing, not necessarily shaders.

But then again, there are some shader-based effects that I feel can add a lot of *oomph* to a scene, such as shadow mapping and anisotropic lighting. I think that both are important, but I'd like to see better textures before better shaders.

 

[Nom De Guerre]

But then again, there are some shader-based effects that I feel can add a lot of *oomph* to a scene, such as shadow mapping and anisotropic lighting. I think that both are important, but I'd like to see better textures before better shaders.

With regards to your "oomph" word -- If you had to classify this important word into either of the two options I provided (which, hopefully, I explained thereafter) in this thread's title, which would it be? In essence, more "Wow" from you from really high resolution textures (not "kickass", a term I can ascribe to shaders but not to textures, since resolution is what a texture currently depends on for what I think is the "kickass" factor as well as realism) or from really "cool" shaders?

PS. IMO, the term "cool" can be applied as a descriptive term for shaders but never for textures (ever thought about this?). Just a sidenote because, like I said, I don't want this to be about high resolution textures vs shaders.

 

[Rodéric]

First of all, I hope everyone agrees upon the surmise that 3D computer programming in general has but one aim -- approaching realism.

No.
3D graphics are just like 2D graphics, except sofar they have been less artistic in nature. (Drawing is an art, but it seems 3D modeling is still mostly about mimicking reality instead of being art, which IMO sucks.)

Grandia 2 graphics at higher resolution is fine to me, just as Zelda Wind Waker was great...

It is my opinion that shaders in themselves allow us to be "impressed with 3D graphics" instead of actually being "impressed that they make 3D graphics that much more real". I hope you get what I mean.

Shaders add flexibility, freedom, capabilities.
You move from fixed function hardware to programmable hardware, it's prolly the biggest leap in the history of realtime 3D gfx sofar.

It is also my opinion that extremely high resolution textures (a word, "texture" that is, that hopefully will be surpassed in general meaning and definition) play a much greater role in approaching "realism" than shaders do.

No.
Clearly without the proper lighting models, subsurface scattering, atmospheric scattering, BRDF... You won't have realistic graphics, detailed surfaces yes, but not realistic.

These are just examples. Instead of focussing on the difference between the importance of high resolution textures versus comlex shaders, I'd appreciate thoughts on the two choices I mentioned in the title of this thread.

More of everything would obviously be a solution, or infinitely fast processors so you can do texture synthesis on the fly

In a World War II shooter, realism is clearly a plus to me, but it shouldn't be applied to every single game out there, and IMO that's the way to go rather than keeping on improving gfx in only one respect (looking more real).

 

[JHoxley]

I think I lean more towards Ingenu's thoughts on this :smile:

3D computer programming in general has but one aim -- approaching realism.

No, not really - the term "3D Computer Programming" is extremely broad. It could mean artistic usage (I've seen some really cool demos over the year that I'd argue border on "art") and I've recently been studying the use of 3D graphics in CAD/CAM. In the latter case it might seem that realism was key, but its not always the case - visualization to help solve the problem and allow an operator to make effective and productive use of the software is not necessarily realism.


Correct me if I'm wrong, but you seem to be splitting shaders and textures - almost "shaders vs textures"? The way I see it is that the texture is nothing more than the source data for the algorithm. Simply on its own with no algorithm/process its largely useless. Equally, an amazing algorithm is usually nothing without equally suitable source data to operate on.

I will concede that excellent textures/materials/geometry (note: you didn't include geometry in your list - why?) made Half-Life 2's technology shine. I wasn't really that impressed with the "pure" technology behind HL2 - nice, but nothing I'd not seen before. However the game had amazing visuals due to the artistic assets used. Excellent data improved average algorithms.


As for "solving current limitations"... hmm, not sure it can be done to be honest. As soon as we catch up with the limitations someone goes and figures out the next cool thing that has a whole new set of limitations. That is, it strikes me as being always 1-step ahead of what we currently have...

"Improving upon believability"... I find it difficult to see any answer to this without context. If you're looking to accurately represent the real-world then you've got some real-world targets to aim for (exceptionally high targets mind), but equally - what about some crazy alien space game...? Where you're not necessarily tied to "real world" models yet you still want to create belief/immersion for your player.

Jack

 

[Frank]

While I agree that lots of great, high-res textures offer the best visual experience at the moment, they suffer from two serious limitations: memory and edges.

To get great looking textures, you need normal maps, bump maps, specular maps, glow maps, transparency maps etc. as well. And while you can combine most of those into three textures, you need to supply the textures in the highest resolution they might be used at, like when you look up close.

And the main benefits from high res textures are not only that individual objects look better, but to be able to make them all look unique. So you need a large amount of them. And a single set of 2048*2048 pixel textures takes up 50 MB. So, while you would want to have one for each object, you can cram less than 10 (as you need a framebuffer and such as well) in 512 MB memory.

Let's say we compress all that in a format that is directly supported by the GPU and you only need 25% of that space. That still leaves you with only 40 textures at one time for fluid gameplay.

Forty unique objects on screen at any one time. And massive bandwidth needed to draw the screen.

So, shaders are a way to generate textures without needing all that memory and bandwidth. They offer "free" textures, in any resolution. And they can look just as good as you want them to, as soon as someone thinks of a way to code it.

Another point is, that you could fit all you need into one texture (color + combined specular/glow) if you use high resolution geometry. One material for each small triangle, so no need for transparency. And enough geometry not to need normal or bump maps.

Then again, draw calls are expensive with DirectX, and current GPUs have a lot more pixel processing power than vertex processing power. So, if you increase the amount of geometry too much, you're only using part of the power available and your framerates will drop.

And very small triangles will also increase memory usage a lot, while all of them have to be processed to be able to determine if they're visible.

But they remove the problem of straight edges, and when they get small enough, anti-aliasing.

And so, the next DirectX incarnation allows you to specify additional geometry by using... more textures!

 

[ERK]

Modern audio synthesizers (and software synths) have advanced to such a degree that they can do surprisingly well mimicing the timbres of of a classical violin or grand piano (not quite there, of course, but some can come very close).

But more importantly, these synths can also open up the universe of sonic textures and sounds in ways the classical instruments never scratched the surface. Synths can make waveforms that no mechanical system is capable of. The possibilities become literally endless.

And so now we have entire musical genres (trance, ambient, futurepop,...) where rhythm and melody often take second stage (and may even be repetitive and simple) while the modulation is done to the sounds themselves. The "musical" interest is in the awesomeness of the sounds themselves.

Graphics capabilities should be like that.

Of course, in the same way that most literature is not fantasy, I can see that most 3D environments will likely aspire to realism.

As for the original question, my answer is both.

ERK

 

[Tim Murray]

First of all, I hope everyone agrees upon the surmise that 3D computer programming in general has but one aim -- approaching realism.

The aim is to remove limitations (as JHoxley and ERK correctly noted) so that developers and artists can choose what they want to do with the cards.

Nice nick, Rev.

 

[digitalwanderer]

I don't think "realism" is the goal per se, more like "pseudo-realism".

Removing limits as stated above to achieve that is what I think the goal is too.

 

[Frank]

Modern audio synthesizers (and software synths) have advanced to such a degree that they can do surprisingly well mimicing the timbres of of a classical violin or grand piano (not quite there, of course, but some can come very close).

But more importantly, these synths can also open up the universe of sonic textures and sounds in ways the classical instruments never scratched the surface. Synths can make waveforms that no mechanical system is capable of. The possibilities become literally endless.

And so now we have entire musical genres (trance, ambient, futurepop,...) where rhythm and melody often take second stage (and may even be repetitive and simple) while the modulation is done to the sounds themselves. The "musical" interest is in the awesomeness of the sounds themselves.

Graphics capabilities should be like that.

Of course, in the same way that most literature is not fantasy, I can see that most 3D environments will likely aspire to realism.

As for the original question, my answer is both.

EP

And the interesting thing is of course, that they do all that by approximating it by code. Shaders. But I agree, that it will take a lot of time before we can generate the right color for a complex and lifelike surface that way.

 

[Frank]

Fractal landscapes versus compression.

You can create just about any (2D) surface by writing a fractal of only a few kilobytes, no matter the amount of detail required. And you can zoom as much as you want, and it can still look good.

But the problem with that is, that you have to create the fractal up front. Nobody has yet been able to compress such a surface (texture) into a fractal, although there are plenty of rumours.

But, does that really matter? Especially when you use it to replace textures? No.

So, it can definitely be done. It's just hard and will take a lot of time and very creative nerds.

 

[poopypoo]

Modern audio synthesizers (and software synths) have advanced to such a degree that they can do surprisingly well mimicing the timbres of of a classical violin or grand piano (not quite there, of course, but some can come very close).

But more importantly, these synths can also open up the universe of sonic textures and sounds in ways the classical instruments never scratched the surface. Synths can make waveforms that no mechanical system is capable of. The possibilities become literally endless.

And so now we have entire musical genres (trance, ambient, futurepop,...) where rhythm and melody often take second stage (and may even be repetitive and simple) while the modulation is done to the sounds themselves. The "musical" interest is in the awesomeness of the sounds themselves.

Graphics capabilities should be like that.

Of course, in the same way that most literature is not fantasy, I can see that most 3D environments will likely aspire to realism.

As for the original question, my answer is both.

EP

dude. you just made me cream in my pants. if only! i might point to psychonauts as a good example of a game doing things like this. sadly, the console arena gets far, far, far, more and better treatment in this regard.

 

[Nom De Guerre]

(note: you didn't include geometry in your list - why?)

Because my mention (comparison, if you like) of textures and shaders (psst, I didn't have a "list" in my post) are what I thought I already explained -- they're just two examples and I did mention that I had hoped this topic wouldn't be just about textures versus shaders. Like someone rightly mentioned, lighting et al has a large part to play if we're talking about approaching realism (or "just" excellent 3D graphics) but the point of this thread shouldn't really specificly and, more importantly, individually address all the components that make or break the realism of 3D graphics or even just plain ole' "great 3D graphics". Otherwise, we can just say that the most realistic-looking lighting model would be wasted if we have just flat shaded textures and then proceed to go back and forth in a catch-22 situation, right?

I know 3D graphics is progressing in stages. Someone mentioned here that the introduction of programmable hardware is the biggest leap in "realtime 3D gfx" so far. However, I don't think anyone can seriously say they don't think S3TC (again, just an example, no more), when it was first introduced, has the same kind of impact (albeit limited by memory and bandwidth, as opposed to programmable hardware which is limited by clockspeed, APIs and architecture), or that the leap to floating point precision (and whichever way you look at it, ATI's 24-bits was -- is -- clearly enough of a giant "positive" leap as opposed to NVIDIA's brave but probably marketing-driven 32-bits) plays as huge a part in the general evolution of "real time 3D gfx".

I also did state that however much 3D graphics technology can improve, it cannot be independent, that 3D graphics itself will depend on other improvements in computer technologies. As such (and I thought this was understood in my initial post), I tried to seek opinions on whether 3D graphics is about one, the other or both (i.e. the subjects of limitations and believability) in its current "conundrum" (dependability).

What I have read thus far in this thread, instead, seems to challenge my perception of just what that elusive Holy Grail of 3D might be (anyone has Orton's or Jen-Hsun's email address? What is the purpose of of better graphics from their companies, besides money?). Personally, if this isn't about games looking like Lord of The Rings -- reality, realism or virtually indistinguishable from reality or even suspension of disbelief -- then I am most perplexed and am in probably the wrong industry.

 

[_xxx_]

ERK: I don't see the relevance of your audio reference. Sound and GFX work differently.

 

[ERK]

ERK: I don't see the relevance of your audio reference. Sound and GFX work differently.

Hmmm, not sure what else I can say...

Just that GFX and audio synths are both ways of constructing simulacra of the real world, visuals (GFX) and sounds (synth). But for neither should the only goal be merely fidelity to reality. The tools of generating artificial visuals and sounds are also useful for the fantastic and the artistic.

I am sorry if that doesn't help. Might ask The Baron, poopypoo or DiGuru.

ERK

 

[Luminescent]

Although I do like ERK's analogy to some degree, I must add that with 3D visuals, at least where gaming is concerned, immersion is paramount. The goal is not just slick/artisitc visuals, but visuals that are immersive and consistent with the other dynamics of the respective game. With audio in isolation, the goal is not necessarily immersion, but a listening experience, which is more akin to viewing an art piece without a thorough and cohesive context. I guess it would be more accurate if we limit the analogy to soundtrack audio.

 

[Nom De Guerre]

Although I do like ERK's analogy to some degree,

Personally, I liked his analogy quite a bit.

I must add that with 3D visuals, at least where gaming is concerned, immersion is paramount.

I didn't mention "gaming", which should be the least important industry IHVs such as 3dfx, S3, NVIDIA, ATI et al should rightfully be concerned with. But since this is essentially a gaming forum, I hope to read more comments from folks in this forum wrt 3D graphics as it relates to "immersion", which has a not so inconsiderable relationship with "realism".

Thanks for possibly throwing others into the path/light I originally intended.

 

[digitalwanderer]

I didn't mention "gaming", which should be the least important industry IHVs such as 3dfx, S3, NVIDIA, ATI et al should rightfully be concerned with.

But they're businesses, and "gaming" is where the money is at if you're a graphic IHV right now.

 

[Nom De Guerre]

Sure. But I said "rightfully", not "financially". And since you brought this up, would you say that one of this terms is holding up the progress of 3D advancement?

 

[_xxx_]

Just that GFX and audio synths are both ways of constructing simulacra of the real world

Surely the goal is the same for both (approximating reality) but the tech and ways of doing it are not even remotely comparable. For audio, you'll use more or less fixed pipelines with just a few adjustable parameters, digital sound effects are usually just discrete transfer funtions implemented in HW. Synthesizers do the same thing essentially, or extract the frequency and use just the generated sound instead of the original. Neither of these techniques is appliable in the gfx world.