Let's just wait and see where it's going before jumping in to conclusion like "the world is flat" and :rofl: like ignorants.
Printable View
Let's just wait and see where it's going before jumping in to conclusion like "the world is flat" and :rofl: like ignorants.
lol, minecraft dude says "It’s a scam!"
He also did a follow-up post: But Notch, it’s NOT a scam!
Actually, I can see some immediate application of this rendering method. A perfect type of games would be something like TrackMania Nations, where the environment is virtually 100% static structure. The only aspect they need to finish is the surface shading.
The catch is you can't really animate things the same way we're used to now with polygon-based geometry. best you can do with voxels is rotate them (or basically change their xyz+A coordinates). Can't do skeletal animation, which we've been doing since Quake III. I suppose you could go for a sprite-like approach, and have N amount of frames of animation of your 3d objects (essentially keep replacing one model with a slightly deformed version to give illusion of animation), but this is impractical - an artist would have to hand-craft every frame, and this can quickly get out of hand with high-resolution art assets.
There isn't even a real win with voxel-based geometry in terms of performance or anything. Like, you can define your geometry as tiny triangles, or as tiny cubes. The calculations to put pixels on your screen are going to be roughly the same, assuming you use the same rendering technique - that is, rendering is not really that dependent on how the geometry is defined. I could build a 3D object entirely out of donuts if i wanted to. I think by far, the biggest strength with voxel-based geometry is partial asset destruction (or deformation). You can always just remove a few voxels from a model, and the rest of the model stays intact. If you start removing triangles from a polygon model willy-nilly, it'll screw up royally - texture won't wrap properly, you'll get crazy non-euclidean geometry going on, etc.
Anyway, so the biggest roadblock is animation. But if anyone could figure out how to do it, I'm sure it's Carmack.
Voxelstein
http://www.youtube.com/watch?v=uCtgtF52nAQ
edit
Also, voxels with bullet physics: http://youtu.be/eQMBGLMtdFE
AFAIK Animating something like this is a nightmare, and point cloud data requires a monstrous amount of storage and memory (like terabytes for a full game of unique points). Don't quote me on that though, just what I've heard.Quote:
Originally Posted by el mano
Their "refute" of the "archaic" polygon system with their strawman comparisons makes me angry though.
Edit2: also I saw what was essentially this same video back in 2009 I think. I believe if there wasn't some massive inherent flaw in this type of thing it would be way more known/hyped/funded/what have you.
the problem is not pairing voxel engine with a physics engine.. but the ammount of voxels you are going to simulate... the video is quite simple... now imagine the same buldingwith a view billion voxels, all of them physics enabled... hello need for super computer...
can voxels be cut in half or broken down? just imaging something simple shattering and how it would be handled, cause then the voxels dont just need to be on the surface, but the entire model needs to be filled in.
very very not impressed. like everyone said this would probably take so much computer power to run. plus it didnt even look that great at all.
Didnt say it was awesome.
It probably won't be feasible for a while (8-10 years or more).
But it's the way to go, IMO. And a perfect way to utilise all those multi-core CPUs and GPUs.
This is definitely the case.
Polygons or not, all this geometry is a lot of data. There is no way around that (unless everything looks the same or it is generated by an algorithm on the fly).
Altering the surroundings in games can be done with polygons, like was proved with GeoMod engine the ones used in Red Faction - didn't look bad compared to this demos which look like crap.
Just a idea why this works:
They don't store every pixels data into files, only the points where geometry changes and what's wrong in notch's calculation is: He count's how many pixels can be fit into a square area. Here's the catch: Obviliously the models are hollow so only exterior needs to be saved and loaded from files.
I can just quess why they're using CPU rendering instead of GPU. Ram, that thing must need tons ram. They can't get the are that needs to rendered into GPUs memory -> lets make a easy solution for demo vid: Lets use a modern cpu(s) with tons of ram.
Just my 2 cents. I have just hunch how 3D Rendering works on the practical level.
I doubt ram is the limiting factor, the russian fellow was rendering a voxel based demo on a laptop which probably maxed out at most 8GB. from what the others are saying its probably processing power for a given size of voxel which would lend credence to why it isn't used already.
@fellix_bg: thanks!
Well, it does appear to be real.
They already got funding.
The only issue that may be around is animation. But we'll see.
That seems to run decently on "just" a laptop, would like to see a tech demo on a high end stationary computer with for example some weather animation with the same amount of detail.
Gonna watch the whole video later today after working hours.
Impressive for software engine! Now bring us animations :)
1999 game Outcast. Voxels for landscape, polygons for objects. Voxel games are not impossible to make games with.