However, we can detect changes much shorter than 1/25 of a second, especially with training.

http://www.100fps.com/how_many_frames_can_humans_see.htm

I don't want everything "realistic". I like my games being virtual 8((((

Judging by the video, voxels are at least five times more durable than ordinary pixels.

I prefer this one.
http://wonder-tonic.com/wolf1d/

It's the same snake oil that Carmack has been peddling since 1997. It still can't do animation, it still can't accommodate a scene in a database smaller than an institutional disk farm, it still looks like crap, and he's still trying to defraud investors into funding him.

This is not new. This is old wine in new skins.

What does Carmack have to do with "peddling" this? He talked about it on his twitter and said it wasn't as legit as it sounds.

You're right. This one is not Carmack's baby. It's a ripoff, one that Bruce Dell (Euclideon, formerly Unlimited Detail Technology) is promoting now, of Carmack's work with voxels. It's still old wine in new skins. Dell says it's not voxels, but he doesn't know enough math to tell a voxel from a hole in the ground anyway.

honestly , where will it stop :)
2.5 decades ago i played with a console the size of a shoebox using a dot surrounded with dots that meant to be a spaceship.
so if they take half the time to design them voxels this would get nasty :)

The best way to use voxel tech like that would likely be a hybrid system using polygons for animated characters/objects and voxels for the terrain, structures, etc.
Have the voxels be deformable with weapons fire / explosions / heavy vehicles (eg. this video of the Atomontage Engine).
Some games could have tools for digging or manipulating the terrain in other ways - imagine something like a 3D version of Cortex Command or Minecraft with much smaller blocks.

Like, oh I don't know, Outcast maybe? :P

But yes it's a shame that more work wasn't done with hybrid engines as clearly there was a hell of a lot of potential there.

I had a good feeling this post would draw alot of different opinions & such...me, I believe new tech when I see it.

Now if he'll only get around to finishing Fez we can move on to Trixels ;-)

Exactly! :)

Recently there seems to be quite a large focus on improving the computational performance on GPUs... the main issue with Outcast when it was released was that it only used CPU-based software rendering, which was slow even on powerful computers at the time. A GPU-accelerated voxel engine would likely look and perform much better than the one used in Outcast (and could take advantage of shader effects to get some epic visuals going - maybe an effect could be used on the voxels themselves to make them blur into each other, thus reducing visible 'voxelisation'...)

Whoa...

OpenCL could certainly make voxels a lot more viable, that's for sure.

I'm with them on that. In the long run it's much more likely that games are going to go the ray tracing route as it offers better scalability and is much more naturally programmed for multicore processors.

This whole notion of virtual atoms strikes me as way too computationally intensive for the possible benefits.
The FPS is part of the governing factor when it comes to latency. 1/120th of a second is actually ~8.3ms which is definitely within the range of what better monitors can manage. You might not notice the difference between that and a refresh that takes 16ms, but the computer could plausibly slam in a movement in that time.