Quote:Original post by origil
Seeing how 3D games apply between 47 thousand to 5.2 million calculations on many vertices/pixels/whatevers to display better graphics with shadows, lighting, bump mapping and a few other dozen effects I have only heard the names of - I started thinking: how much of this computational power be used to change game itself?

Practically none. It all happens on the GPU within the video card.

Quote:Original post by origil
Call the video card a game mechancis card instead if it'll make my point more valid...

It makes it more valid, but it doesn't make it any more practical...

Quote:Original post by origil

Quote:Original post by Kest

Quote:Original post by origil
Seeing how 3D games apply between 47 thousand to 5.2 million calculations on many vertices/pixels/whatevers to display better graphics with shadows, lighting, bump mapping and a few other dozen effects I have only heard the names of - I started thinking: how much of this computational power be used to change game itself?

Practically none. It all happens on the GPU within the video card.

So take this none that happens on the video card and make it play with the game mechanics.
Call the video card a game mechancis card instead if it'll make my point more valid...

Let me be more specific. None of the calculations that render polygons can be used to increase the complexity of gameplay - because it all happens on a seperate processor on the video card.

By the time it takes your gaming calculations to travel from your system memory over to your video card, get computed, then travel back, your CPU could have done that work ten times.

Quote:Original post by origil
Off-topic:
Ok... I see my nice "theoretical" idea has been killed again and ran over by off topic technicalities that have little to do with my idea but were merely implying that:
Just as graphics calculations are made, so can these be made to create a richer world in the terms of relationships between objects.
I am NOT saying: Becuase of 3D calculations we are losing computer power, thus can't use these for other things.

They aren't off topic; they are informing you that you can't use the GPU for general purpose programming on the level you need. You can use the GPU for some pretty physics type stuff, but that's only for physics where you don't need to keep that data around. For your game specific programming, all you have is the CPU. Doesn't matter what the GPU is doing while your CPU is running so cutting down graphics (to that extreme level) probably won't do much for the idea. Of course you'd want to use simple 2D sprites or something (there's still some CPU overhead of a complex 3D scene what with scene-partitioning and things like that).

The idea is a cool one, but very hard to do currently. Each object you mentioned can probably be accurately modeled separately, but to do them all at 60fps is the challenge. With multi-core processors coming to light and the new quad-cores that are out, it's becoming more feasible, but there's still the issue of maintaining the separate threads and keeping all the data in sync.

Everyone wants molecularly sound game dynamics, but the fact is that computers can't handle it. The GPU won't help you here since getting data back from the GPU takes far too long compared to just doing the work on the CPU.

Quote:Original post by NickHighIQ
I know EXACTLY what you mean! I've always thought that, why not represent in code the real world? That is, vigintillions of atoms and photons interacting all over the place. There's only four forces in the real world, shouldn't be too hard except for the complexity and innumerability and infinitesimallity (made up word) of the system and its components.

Maybe they could get a super-computer or six to do it. Just letting you know I'm on a similar, if not quite the same, page as you.

Represent the real world? Please, define real world. It seems like you know everything about the "real world", that it shouldn't be complicated to represent it in code.

Well, sorry but you cannot represent the "real world" in code, because mankind does not know enough about the real world to do that. The best you could do would be an approximation, but that approximation would not represent the real world.
Look at all the physics engines out there. Do you think that they have anything to do with realism? No, because they all use newtons axioms (sort of), and those are considered disproved. Even if you were to implement something on the atomic level, it still wouldn't represent the real world, as the laws concerning atoms aren't doing any good in representing the real world either, because the concept of atoms is in itself flawed (therefore science came up with quantum physics). And yet there are problems in science to apply the laws of quantum physics to macroscopic phenomenons. It seems as if these two were two fundemantal different things, and still both, macroscopic and microscopic physics are separately used to describe the real world, although they can't work together. So please tell me, how would you try to simulate a cup of tea with just implementing some laws of the atomic level, if even those aren't considered true.

Lets assume someone was to succeed with this concept: A physics engine that would compute everything by interactions between atoms. To "realistically" simulate even just a ball bouncing or something, you would need so many atoms until you gathered the mass of the earth, unless that engine used some sort of hack to do as if that many atoms existed. You still would be better off just assuming that ball was a rigid body and just applying newtons laws on it. There is no need to simulate atoms if you just end up using them for everyday bodies, but if you do that to simulate things like water, it is highly unlikely that you end up with pleasing results (or even "realistic" ones).

And I don't see the point in the ideas discussed in the first post. Why compose a grenade of gunpowder, ignition switches and so on, if you can just hard code a grenade that applies some force to its surroindings after it explodes? Yeah right, if you had a physics engine on the atomic level you wouldn't have to hard code it, but just compose it from different elements, still you would probably end up with the same result.

I think stuff like "atomic level" calculations are unnecessary, because today's pyhsics engines do just fine. There are also libraries out there which can cope with real time deformations and fragmantations. So why ask for more? Anything more wouldn't add to the gameplay anyway. The only gametype that would benefit from such a thing would be these sand games-sort of game, and they aren't any good.