I'm personally wating to see what this guy does with his engine.
http://www.youtube.com/user/AtomontageEngine

Its a very similar principal

Once they nail the new pretties down, and I hope devs/pubs can consistently pin down games that:
1 - Have an enjoyable, complete, and engaging storyline
2 - Do not have crippling DRM

Voxel atoms appear to be a fraud.

1. how would you animate and arm based on that? you would have to count every single voxel position, which would aqcuire HUGE AMOUNTS of CPU

2. If a scene has 250 bilions of milions of voxels, and each of them is, for example 6 bytes of memory (colour, lighting, physics etc) you would need HUGE AMOUNTS of GPU to calculate this.

I can't see this as a solution.

Notch used the second to come to that same conclusion. In his estimation it woud take several petabytes worth of atom to do a relatively modest island.

And you're right, Voxels have never been very good for animation. C&C used voxels, but the limit to the animation was changing colors and orientation and that was about it for most of it.

Yeah, it just doesn't make sense, when you have to set parameters like lights, volume, etc. for every atom present on a scene... This is why polygons are much better for gaming.

No kidding, especially if the atoms are 100% distinct graphics entities.

If you use a tool like OpenGL, it will calculate things like reflection based on the number of polygons (which is a faction of the number of vertices in a regular polygon mesh).

Now, if every vertice is a complex polygon in itself (a cube for example will have 6 surfaces to calculate reflection on), you significantly increase your CPU load.

And then of course, there is the amount of atoms you will need which will be far greater than the number of vertices you'd need with a polygon mesh.

Perhaps it will find some applications in things like animated movies (speed isn't as much as factor as long as things scale linearly, they can use server farms to generate the frames), but I think it will take a while before it finds any application in gaming where real time speed does matter.

Do you have any idea how much CPU power it would take to generate such things?!?!

Making things more shiny and muddy is easier

Yes, but in the long run I'm guessing that Ray tracing is the ultimate destination. Polygons are more or less a good enough set up, but as processor capacity increases it's going to be very hard to beat ray tracers for ultimate graphic quality.

Especially given that it's much more similar to how we see than any of the other possibilities.

Then again, his calculation is seriously flawed. He assumed a solid block (at maximum resolution), which makes absolutely no sense at all.
If you have no empty space to move in, that'd make for a really fun game to play, wouldn't it? :p
Or, in other words, a very good amount (majority) of game world is usually not 'dense' and thus doesn't have a voxel/atom in this scenario.
Not to mention that you'd usually have objects requiring less than the highest resolution.


If you want to store three cardinal directions in 6 bytes, you'd end up with a maximum of 2 bytes per direction and thus a maximum range of 0-65535. I'm not sure how you intend to squeeze 4000000 discrete locations (as used in your example) into that range. Or, in other words, I think you'd at least need 4 byte per axis.

There's no problem with that. If you take a look at the assumptions, some of them are over estimating and others are underestimating the amount. What you're neglecting is that just because there's a block there doesn't mean that you can't travel through it. A jungle island would have all sorts of vines in places and probably water.

But, the bigger issue is that even if you do scale back a lot of those areas to be more open, you're still left with the problem that he's using only 1byte per block, and you're going to need a lot more than that to do anything interesting.

Typically the way these Fermi problems go, I'd be surprised if he wasn't within a couple orders of magnitude of the correct number. Which would still place it a hojillion megabytes too large for practical application any time soon.

Yeah, you can probably use quite a few voxels over and over again, but as you do, you lose out on the graphical advantage from using voxels.

I still say that crysis 2 is good looking game.. And gets too much bashing in pc, it's not so open that I hoped but it's pretty neat game. They should use nanosuit in more games.
And (another matter) propably some have said that "it's not all about the graphics" And I'll agree.
I don't really want games that need always a new processor. Better story and interactivity, better game. I don't really care how realistically some grassbush looks like in the wind. Drawing distance is more important. =)

I guess nowadays they are so interested developing a new graphics engine that they release some garbage that looks good.

Still.. In crysis 2.. FOV is too close.

With this rate.. We will see 5 games per year.
I'm just saying that developers should make games instead imitating real life.

right back at ya

Ha-haa *says in mickey mouse-voice*

One of Mircosoft's Xbox 720 developers boasted that the next gen console would produce graphics on par with the movie Avatar. Of course this has drawn a lot of laughs from the community, but maybe this is what he was talking about.

Looks cool.

Works on a low power computer, sounds like BS to me. Something will be doing this thing's work unless he's got an actual GPU in there.