Hi 3D enthusiasts,
here is a funny (slightly ironic–or just British?) video and an interesting discussion of what is possible, and what might be possible in the future, in 3D.
http://www.rockpapershotgun.com/2010/03/10/unlimited-detail-wants-to-kill-3d-cards/
http://unlimiteddetailtechnology.com/
The conversation starts with the claim that polygons are "so last century" and if we'd only cull and raycast efficiently, then we could use full-detail "3D atoms" (voxels?) instead of polygons. The video itself doesn't explain how they want to do it, and they just show some colorblind static images. ;)
The blog commentors ask themselves how this would work together with shading, reflections, or animations, to become useful for games. And how would you ever create models with that degree of detail? Even if they'd ever create a mesh->"unlimited" converter (which they haven't), it makes you wonder in which cases is it really worth it to hire an army of artists to design all these details. (Or get actors and landscapes to "3-D scan" the details from.)
Ironically, the "technology" seems to be outright incapable to render a plain flat single-colored area, everything must have detail down to the atom: So in the end, everything comes out very busy and gaudy.
Well, these guys haven't released anything yet–I'm just curious of your opinions. Is the claim true that this is something new that has never been done before? The demos look a lot like fractals.
PS: Correct me if I got something wrong – my greatest computer graphics feat was a rotating cube… XD
yes, voxel / ray tracer engines definitely work. Unlimited detail, … no.
Outcast was among the first to use it back in the days. http://en.wikipedia.org/wiki/Outcast_(video_game) In the wiki entry you’ll also find a link to the voxel technology and some technical stuff to it.
Does it work well enough to some day replace current technology? We’ll see, my guess is a definite no, not ever and ever. 
It's an interesting system, sounds like they've come up with some very powerful searching and culling algorithms.
But at face value it seems to have a lot of real-world shortcomings (mostly mentioned in the comments), and the video doesn't do anything for my confidence that they have dealt with them.
First off, it's not infinite. Whatever way you cut it, points are no more infinite than shapes (polygons). Billions is still finite. Very much so when you're comparing the points to atoms.
At some stage, you have to turn a point into something you can see - render it. How do you do that? The assumption underpinning the whole presentation seems to be that a pixel will always be bigger than a point, so all you need is it's color. But you can't have your cake and eat it. Polygons can be smaller than pixels, too.
It's stated that you only need to deal with as many points as there are pixels on your screen - so it's all a matter of efficient sort/cull. Perhaps, for static scenes. But what about animation, physics, collisions? I suppose you can say the infinite detail is in appearance, and use low res approximations for the rest. Is that viable or would it mess up some search tree? I don't know.
Content creation I suspect would be a short term problem, as if the technology was sound the tools would come.
Sounds like they might have some interesting technology there, but it also sounds like it has a lot of maturing to do and they have over hyped it.
well considering the resolution and hence size of DICOM data sets as well as the number of different interpolation/samping methods you can use, due to volume buffers rectlinear nature, granularity is never going to be an issue for volume rendering…
also volume rendering structures lend themselves very very well to parallelism especially when using ray tracing and some grid axis-aligned technique like shear warp of the viewing matrix so speed is also not an issue…
the only issue is why bother?
volume rendering is really only useful when you want to look inside something hence it's use in imaging…
there was a lot of research done about using voxels for games but for the most part it was found to be a bit pointless bar maybe terrain and modeling tools…
it's super fun though and pretty straight foward 
Note they do state it's substantially different from voxel rendering.
But of course they don't explain how, or give the same break down of their tech as they do the others. I guess we're supposed to assume it does everything well. But on the face of it, that doesn't seem likely.
Alric said:
Note they do state it's substantially different from voxel rendering.
Yes they do, but if I use Triangle Strips exclusively instead of Triangle Lists, I might also say that i do things substantially different from Triangle List ...
Also many people (including the authors) don't seem to get the mathematical and physical nature of infinite ("unlimited").
dhdd said:
Alric said:
Note they do state it's substantially different from voxel rendering.
Yes they do, but if I use Triangle Strips exclusively instead of Triangle Lists, I might also say that i do things substantially different from Triangle List ...
They SAY a lot, but does it make sense, that's the question.
I cannot tell whether they really put together something new and don't go into detail to protect their IP... Or whether they are purposefully selling off something old with a small improvement as something new, under a new hype-y name... Or whether they unwittingly reinvented "the wheel" and avoid the existing vocabulary because they don't know it.
dhdd said:
Also many people (including the authors) don't seem to get the mathematical and physical nature of infinite ("unlimited").
Engineering meets Marketing. "You promised them...WHAT?" :evil:
They probably compare their "unlimited" graphics to existing games to get the message across to the general public (games being the best known example of 3D imaging).
Even if they don't ever develop a game engine, it may become a viewer for detailed images (for scientific purposes, rendering molecules, scanned models, whatever). But if the guy had recorded an honest video saying "In 2 years time, you might be able to view certain types of boring scientific data in greater detail" then we wouldn't be watching it, I guess, and nobody would consider buying the company. ;)
Thanks for your comments, this is really interesting.
Last year there were predictions for ray-traced graphics games in 3 years. Sooner or later someone will make games using technology like this. Maybe not any kind of game, but specifically using some features of the technology which would be hard to make using polygonal 3D. Or maybe a mix of polygon graphics and this.
voxels are definitely coming back. but i would bet my money on the paths ATI, id software and others are going, with sparse voxel octrees…
Definately so - a certain MrCarmack is a big fan of sparse voxel octrees and lobbying the card manufacturers to move in this direction and away from polygons - I rember this interview with him about this from last year
http://www.pcper.com/article.php?aid=532&type=expert&pid=1
More actuall than the Outcast ^^
STRATO - Domain not available
However it is probably "only" polygon based ^^ but as it seems without using voxels and stick to polygons it seems to be nearly realtime ready.
After all a 8 core is not so much future anymore:
http://www.youtube.com/watch?v=blfxI1cVOzU
Follow up, one year later: Well, we still don’t know anything more than last year, but they are losing the target group fast: http://notch.tumblr.com/post/8386977075/its-a-scam (Unless the target group is venture capitalists…)
Hm yeah still notihng more from that, however some development at raytracing, at least that is what most 3d movies use by now to generate the images.