Euclideon Develops New Way of Rendering Graphics ! INFINITE DETAIL !

Trando434

Member
Watch and be amazed! They are rendering using virtual "atoms" instead of polygons! Anyone with technical background in this stuff will be extremely impressed! Gamers as well! I was blown away at the number of Nerdgasms I had watching this!


 
Last edited by a moderator:

SchizophrenicMC

Well-Known Member
Old News, my friend. This has been around for a number of years. The problem with it is it takes an insane amount of disk space and loads the CPU and RAM far more than traditional 3D resources. Furthermore, no developer has taken it seriously. The concept is really cool, and there's probably a future for procedurally-rendered resources, but right now, Infinite Detail is both a misnomer and a lifeless technology.
 
Not to mention all the tools to produce graphics are polygon based and they do not seem to have a solution for creating things specifically for the tech from nothing(they need a conversion or laser scanning and laser scanning is not cheap). Also note that nothing is moving in any of those demos nor are there any physics. I bet it breaks down really fast as anything more than a visualization program. Why are they not targeting the much more lucrative medical industry if this is such a new and revolutionary thing?
 

Jake D

Jr Member
I see a couple huge obstacles in its way. First, physics, how would it be able to calculate, in real time, a ball made up of millions of their "atoms" into a floor made up of a huge amount of the same "atoms". Polygons work for real time physics because it's calculating say, a 128 sided circle (random numbers here) into a single pane of a floor. The load on your CPU would be immense. The second problem I see it the time required to make each object. That special tree they were so proud of probably took a significant amount of time to make, imagine doing a decent number of unique trees so the foliage isn't repetitive, and the same for the thousands of objects you could encounter in a game. Sure you can lazer scan the stuff into their system, but how will you scan a tank, a jet, a horse, a house, a swimming pool?

I'm not too knowledgeable about computing & graphics, but this seems to be just a flashy idea to try and hook investors with. I hope it works, for the advancement of gaming, but for now I'll stick with the tried and true polygons.
 

Toacrabman

Well-Known Member
This will require alot of space, like one of those Infared drone scans that requie a whole special computer with a crapload of space just to read it.
 

rundown

Sr Member
I don't see how this will be faster? Point cloud date has a larger file data then low poly models? I do like that it is very detailed and all, but I don't see this happening in the next 10 years really. Our processors arn't fast enough yet. But if they were I can imagine high poly pre rendered animations being rendered ingame. Anyway I think they will keep investing in better ways to do low poly modes actually.
 
As far as space is concerned. this can be done on Blu-ray disc/games. right now single layer blu-rays do 25+ GB. dual layers are 50GB. to give you something to compare to: "Halo Reach is printed on a Dl DVD (8gb) which means that you could fit six halo reach games onto one Dl blu-ray and still have 2GB left over.
Also everyone, keep in mind that this is a new way to "render" and not a physics engine. This will be the "show" (what you see) and not the "action" ( how things react in the world). I think that because they can convert poly's into atoms for "render" the physics in games will be left up to a completely diferent engine. ( yes! they may, do, and can use more than one engine).
Truth is though, this would more than likely end up in games for the next gen consoles. ones that have been made to specifically support atom rendered games. It can be done. it's just a matter of the right technologies syncing up with one another. but you guys know as well as I do that this will not happen because every developer and manufacturer will be out to make their wallets heavier so...
 

ForgedReclaimer

Well-Known Member
This is amazing stuff there doing.


I agree with Pappa Midnight. This system is very much possible. The major drawback thus far is the lack of a Next Gen console capable of running the system whitch alone would take a few years to create. The other major drawback is the lack of capable collisions and animations the "Atom" like program is lacking and would probably lack for years. But using a technique from games today which is having your visible model running in plain sight and having an invisible collision model that reacts with players, bullets and explosions underneath it would all but fix this problem. Although to see the "Atoms" have collisions would make it 1000% for perfect. As for animation that would be a problem with not as many quick fix solutions as collisions. The only way i can think of this working is if they set "Atoms" up the same way as vertex's on traditional models and run a bone structure underneath with soft edges and try to setup each piece to animate and that would not be all that much different than today's video game animating systems.

I can't program so i cant say for certain if this is possible but i would hope, and believe that the company has or will be researching these sorts of area's as they are proven and successful tools already in use.

And since you can convert 3ds max files to these "Atom" models there's no question that you could create a sort of low def collision model for the "Atom" file to be used later in game..
 

ImaGonnaGetYou

Well-Known Member
This is amazing stuff there doing.


I agree with Pappa Midnight. This system is very much possible. The major drawback thus far is the lack of a Next Gen console capable of running the system whitch alone would take a few years to create. The other major drawback is the lack of capable collisions and animations the "Atom" like program is lacking and would probably lack for years. But using a technique from games today which is having your visible model running in plain sight and having an invisible collision model that reacts with players, bullets and explosions underneath it would all but fix this problem. Although to see the "Atoms" have collisions would make it 1000% for perfect. As for animation that would be a problem with not as many quick fix solutions as collisions. The only way i can think of this working is if they set "Atoms" up the same way as vertex's on traditional models and run a bone structure underneath with soft edges and try to setup each piece to animate and that would not be all that much different than today's video game animating systems.

I can't program so i cant say for certain if this is possible but i would hope, and believe that the company has or will be researching these sorts of area's as they are proven and successful tools already in use.

And since you can convert 3ds max files to these "Atom" models there's no question that you could create a sort of low def collision model for the "Atom" file to be used later in game..
I don't think you quite understand the mind-blowingly high amount of resources needed to render this way. Everything is rendered by searching a cloud of millions of "atoms" to find which ones need to be visible in the frame that is currently being drawn, and after ruling out all of the unnecessary atoms, a single frame is then rendered. Rinse and repeat 60 times in a second to get 60 frames per second. That is an enormous cloud to search through, and I highly doubt any processor in the next decade will be capable of pulling it off. Moving those atoms AT ALL while rendering in this fashion is far more difficult. Any animations involving such complex objects, even small ones, would take either unimaginable processing power or a VERY long time to render.

This technology isn't new and for the most part isn't feasible, at least with the hardware we have now or in the next couple of years considering the rate at which processing power has increased. Low-poly models will probably always be the norm in video games, it's infinitely easier to work with and cuts down production time compared to spending a week just to make a single piece of scenery that might be used 10 times at most in a game.

Consoles have always had crap hardware to keep prices down, so honestly I doubt any console in the next decade or so will depart from low-poly models with fancy textures. They may even stay the norm for long after unless one of the three companies decide it's time for another gimmick extravaganza.

I'm glad that not everyone is riding the low-poly train all the way to Cawadootyville, but to be honest we don't have the computing power to depart from it. Besides, it would be an enormous shift from the norm for such a small gain. Developers already devote too much time to amazing graphics instead of the actual gameplay, if they had to make something of this detail I'm sure we'd be playing Pong: Uncanny Valley Edition.
 

SchizophrenicMC

Well-Known Member
Note how everything in that video was repetition of the same 3 models. It takes too long to map out each individual atom. Then, each one takes up a few bits of information, probably 4 or 8, so a million atoms is a Megabyte. It may take hundreds of millions of atoms for a single model, making each model hundreds of MB. Then, you have hundreds of models in a game, and you have tens of GB. And that's just 3D resources. That's not the color data, that's not the animation rigging. Then you have to add the physics engine, the audio, the render engine, itself, and the back code to tie it all together. Not to mention, network engine if you want online multiplayer. (This will never have splitscreen)

That still takes up more space than a DLBL disc. Not to mention, reading from that disc would take forever, and you'd have to have roughly the same amount of RAM to load the models into memory, where the processor can burn itself out trying to find relevant atoms.

It's not viable. This sort of rendering has been used in procedural models for scientific experiments for probably a few decades. Those models are very slowly rendered, and each point created, then placed. It's useful for physics modeling, and high-detail modeling, but not real-time render.

So, low-poly with more advanced methods of lighting and higher resolution textures to make things appear more real, is about as far as we can go. We can pack more polygons in every year, and it still doesn't overload the hardware. We keep making better lighting engines, that make false shadows and highlights where they ought to be. Real-Time Rendering still has room to grow before support for Procedural Rendering even exists
 

Pack hunter

Jr Member
Old News, my friend. This has been around for a number of years. The problem with it is it takes an insane amount of disk space and loads the CPU and RAM far more than traditional 3D resources. Furthermore, no developer has taken it seriously. The concept is really cool, and there's probably a future for procedurally-rendered resources, but right now, Infinite Detail is both a misnomer and a lifeless technology.
schiz is right at this point in time we as developers we done have a powerful enough computer 2 make this work for the moment, and if the developers don't have it then neither does the general public, at this point in time only private companies such as the one that developed this system are going to have the tech, basically looks like a more advanced particle system which even a small particle fire on a match stick is enough to slow our computers down thats why most game companies use an animated texture.... as a game designer i'm blown away by this, and can't wait to get my hands on it. and its got to be easy modeling if a lil programer was able to do it...

personial i found that guy's voice to be annoying, and a little full of himself
 

Tyvern

Well-Known Member
memory space is now an affordable 1 terrabite for at home computers......WE HAVE THE TECHNOLOGY!!!

RAM? meh...., needs more dev time to catch up

I forsee this being used a lot in high-end games in the next, oh 8 or so years.
 

madmodder

Jr Member
wow now I can't wait to see a game with a unlimited poly count! no more making a low poly model and adding a bump and displacement map just model to your heart's content
 

rundown

Sr Member
I don't get wy everybody compares this with atoms. Point cloud data is still polygonal data...
 

gingersnapples

Well-Known Member
sure a disc could hold it, but its not the disc doing the calculations, rendering, temporary memory storage, ect.
I can almost guarantee (sp?) the computer that was rendered on has a dedicated graphics card worth at least $1000.
 

SchizophrenicMC

Well-Known Member
memory space is now an affordable 1 terrabite for at home computers......WE HAVE THE TECHNOLOGY!!!

RAM? meh...., needs more dev time to catch up

I forsee this being used a lot in high-end games in the next, oh 8 or so years.
A 1TB hard disk goes for $50-100, and that's just the memory device, itself.

You can pack as much RAM as you like, but current operating software cap out at 128GB of RAM, due to some integer rule I don't recall. Of course, sifting through that much RAM, in itself, becomes a chore, and you might as well skip it at that point, in favor of SSD.
I don't get wy everybody compares this with atoms. Point cloud data is still polygonal data...
I was under the impression that point cloud data didn't store polygon data, but let a sophisticated vector rendering algorithm generate polygons from the point cloud. Like a bitmap holds pixels, but a vector holds lines.
sure a disc could hold it, but its not the disc doing the calculations, rendering, temporary memory storage, ect.
I can almost guarantee (sp?) the computer that was rendered on has a dedicated graphics card worth at least $1000.
No form of Compact Disc, be it Digital Versatile or Dual-Layer Blu Ray can currently hold enough data to support this. Not according to my math. You'd have to pack it on magnetic disk or solid state memory, both of which are larger than CD memory and much much more expensive.

The whole point of the Infinite Detail point cloud system is to do away with bulky graphics cards, and it can get away with that. By shunting the processing from the GPU, where it is, to overbear the CPU and RAM. Given that the CPU must also handle physics and netcode, this is an even worse idea for gaming. High-end graphics cards cost $200-300. High-end CPUs cost $300-700 (-blam!- Intel, man. -blam!- them hard). And even those CPUs would be blown by this.

This is either ahead of its time or just a bad decision. Either way, it's not going to see any serious software development for the time being.
 
Top