Facebook Plunges Again
Tumblr 101: Tips And Tricks To Get Started
Unlimited Detail claims to leave current 3D technologies in the dust
Unlimited Detail is definitely the most interesting technology demo I stumbled on today. In a nutshell: current 3D technology is based on polygons. Each 3D shape you see on the screen is made out of multiple straight facets (polygons). The more polygons (or facets), the rounder and more natural it seems. The current battle is all about polygon counts -- how many polygons can a certain graphics card render per unit of time. This metric has been rising 20% per year on average, for the past few years.
But polygons aren't the only way to display 3D information. Unlimited Detail works with dots, rather than polygons. The thinking is that if we use 3D "pixels" instead of flat shapes, objects can be far more realistic and lifelike because you don't have as many straight surfaces. Each object is composed of a mass of "dots" virtually positioned in 3D space.
This, in itself, is not a novel idea. A very similar principle exists in voxel technology. What is different about Unlimited Detail is the way they select what pixels to present. And this is where the "snake oil" bit comes in, because there is absolutely no hard proof for this one. They say they're applying "search technologies" to figure out exactly what pixels to show. In other words, while an entire 3D scene may contain billions of "points", you may only have 1024x768 (786,432) pixels on your monitor. So if there's a very fast way to figure out which of these points must be shown on each of the pixels, you get very fast 3D technology.
And this is what Unlimited Detail say they can do. There's an impressive (if quite smug) screencast after the jump, with a good explanation for novices (and potential investors, I'm guessing -- the site seems quite shoddy at the moment). Graphics experts in the audience, help me out here: Is this snake oil, or is it really the next big thing?
Comments
25
Subscribe to commentsDrewMar 11th 2010 8:34AM
This is interesting if it is really like voxel based technology. This has been used quite exstensively in the medical fields of imaging where it is definitely a mature technology.
Chris LMar 11th 2010 9:24AM
Doesn't seem like it would be able to read ambient & radiosity lighting data as well. But I'm no pro...
DiddleMar 12th 2010 3:47AM
I think what some failed to note; this was more of a proof-of-concept than a "hey, we've got a playable game around the corner". They have an algorithm which figures out the limited amount of information you need on the display of your computer instead of rendering all sorts of polygons. As you get closer to an object, the algorithm says "show more of the pixels from this object".
It makes perfect sense and I'd love to see this in action.
BrentMar 13th 2010 12:12PM
If this is supposed to be so amazing, why do all the "animations" look so dated? If your goign to wow someone with a new 3d rendering technology then at least make it look impressive.
Ian WoollardApr 10th 2010 12:20AM
All he seems to be doing is generating a voxel map from the scene (the scene is presumably generated algorithmically and/or from real world voxel scans), where there's one voxel per pixel.
The clever bit is presumably that as the camera moves around they move the voxels and then check to see which pixels are now without voxels and calculate new voxels for just those, and those pixels with two voxels, they could just pick the front one and throw the rear one away.
So they're using the voxels as a sort of cache to speed stuff up. If so, I don't see any theoretical reason they couldn't, with care, animate stuff, not only the camera would move, things could move independently, rotate and light up and stuff, it's just a question of transforming the voxels appropriately each frame.