Oh man, that sounds awesome. I hope it'll work.

I really like the article and video, so thanks for pointing it out, but I have a few issues.
First of all, with the maximum stated atom count of one million per cubic inch, that would mean a minimum raw data of three million double-precision floats for the coordinates and another 4 bytes each to gain colour (~36 MB raw data) per inch³. That's quite a bit data for such a tiny amount of space, even when compression comes into play. So, to keep a realistic size for the games, you'd be forced to use far less atoms.
And I'm not fully convinced on the 'figure which points need to be rendered in real time' part, because today's graphic engines already use that technique to an extent.

But, hey, I'd be very glad to be wrong and see this in action!

There's an awful lot of people calling bullshit on this one, I suspect that we'll have to wait for an actual game to use it before we can see for sure

"You can scan in things..."

I wonder how many supercomputers we'll need to run it?

this one http://www.engadget.com/2011/06/20/fujitsu-k-supercomputer-now-ranked-fastest-in-the-world-dethron/

Yeah yeah, wake me up when entirely voxel-made games will be there for real. I fear I will have to sleep a loooong sleep...

Voxelstein 3D?

Yeah, so he's talking about voxels, big news. They're fine as long as you don't have a problem with repetitive landscape/scenery, no antialiasing and crap to impossible animation.

Here.

I said games, not shits made of pixels....

We need to reach our human limits with polygons first... Like all games to be rendered at 120fps.

Then we can keep the 120 fpses and improve the tech by switching from polygons to voxels.