Unlimited Detail - The death of polygons?

[Questionable]

Has anybody seen this: http://www.youtube.com/watch?v=Q-ATtrImCx4

Pretty interesting...

[Anian]

As the blind man said "We shall see."

Yeah, saw it before. It does look very interesting, but the actual content making and after that physics might have some issues.

[Questionable]

As the blind man said "We shall see."

Yeah, saw it before. It does look very interesting, but the actual content making and after that physics might have some issues.

I can see physics being a problem but I don't think the actual content creation would be difficult. In some of their other videos they've stated that you can import polygon models or you can edit on a point/pixel basis. SO you can continue to develop 3D like you typically do, then import it into whatever the heck and fine tune it on a point-by-point basis, or you can extrude/paint point-by-point from scratch.

I think interactivity and kinetics might be difficult (notice how the water doesn't move...)

I remember in a Quantum Science class we were presented with an odd idea. How do you know that what's behind you back exists? Maybe it's there in terms of matter but is it there in terms or color or space, or is it even there at all if we're not interacting with it?  You need to be high to really understand what the teacher was getting at but by rending in realtime only what needs to be seen I can understand how they are minimizing the power of GPUs, if rendering polygons isn't a big issue then, those massively powerful GPUs can focus on rendering the other parts of games or maybe focus more on parallel processing.

Interesting concepts but I agree; we shall see.

[xenogia]

I see this an extension of the old voxel engines I used to see and how fantastic they were.  If they get this off the ground it would be fantastic to see once the software matures.

[Phemar]

Yea soon we'll all be buying physics cards, not GPUs.

[monkey0506]

I don't really see why there would be physics issues because essentially the models would still be created individually with their own physics handling (including movement patterns, collision detection, etc.) just using the point-cloud data instead of polygon data.

If you watch the other videos it is explained that what is being shown in these demos is to be considered as "programmer's art" as Unlimited Detail Technology does not presently claim any professional artists as a member of their staff force.

Essentially the way technology like this works is that some company develops the core idea, usually starting with a small company with an extremely limited budget and other resources. Next this company approaches a larger company (or vice versa) with a pitch of their idea for the new technology. The larger company will either obtain rights to the technology, or they won't. In the case of Bruce Dell (of Unlimited Detail) he explains that the Unlimited Detail technology has actually been pitched to a major player (either ATI, nVidia, or Intel, but it wasn't specified which one) but it was turned down.

Big companies are often reluctant to take the risk of spending resources (time, money, etc.) on a new technology because there is the chance that the technology will never be profitable for the company. Take for example the recent HD DVD vs. Blu-Ray battle. A lot of money was spent developing the HD DVD technology, and you can still find HD DVDs on the shelves at some stores. But Blu-Ray won the battle and is clearly the predominant DVD replacement technology.

The same thing applies to this Unlimited Detail technology. It's risky. Obviously if it could be developed and could be practically applied for use by the general consumer then it would vastly overpower the polygon technology. But essentially until UDT gets corporate backing, don't hold your breath to see the technology go anywhere.

My opinion: Cool story bro. Looks nice (esp. for programmer art).

[Questionable]

Reps would be appreciated...

[Anian]

The same thing applies to this Unlimited Detail technology. It's risky. Obviously if it could be developed and could be practically applied for use by the general consumer then it would vastly overpower the polygon technology. But essentially until UDT gets corporate backing, don't hold your breath to see the technology go anywhere.

Yeah, but it just takes ATI or Nvidia to say "we'll might try something" and it'll spark a race between them, maybe produce some other technology. Plus if one of them (or the one who came up with this technology) - arouse an interest from someone lie Autodesk or Adobe or Intel (or maybe even a company like Sony), who might wanna get in on the graphic engines/cards/technolgy race...we would have first test in a year or so and affordable cards within two.
Problem might be that it's not really in their interest - good money is made by issueing new models of graphic cards each year, this would be kind of a business ender (well a drop in income at least).

As mentioned, there is a problem with data, a texture of a terrain may be flat etc, now imagine how much info does it haveto have, it may become easy to render it, but it still has a lot of it draw info from (and you'll notice that there's a lot of cloned stuff in the videos and screenshots, I know it's "programer art").

[monkey0506]

Development of new technologies does not mean that all production of existing technologies would instantly stop for that company. That's completely ludicrous. There wouldn't be a drop in income for that company provided they maintained their present levels of production/development on current technologies (as they would). If they did experience a drop in income it would bear no relevance or relation to the development of the new technology. The company may record a lower profit margin for the one fiscal year during which they are spending to develop the new technology (whilst simultaneously working to continue advancing existing technologies), but based on your estimated timeline they would start recovering those profits within the second fiscal year.

As for the cloned objects, that really doesn't bear any relevance to the amount of data being displayed. There were scenes in the videos that didn't have cloned or reused models and were still rendered as smoothly as those that did. Clearly for cloned models you would expect an optimization, but at this stage I'd say it's probably actually unlikely that such optimization has been implemented. According to the videos, the processor uses the same camera-line processing as existing 3D renderers use to determine what is considered visible. From there it is capable of focusing your CPU/GPU's powers exclusively on the clouds that need to be rendered as opposed to rendering the entire scene.

In addition to higher resolution models, this concept also provides for much larger worlds. One of the major fallbacks of the polygon rendering system is that it often renders a much larger amount of the scene that what it needs to. This needlessly bloats the workload placed on your processor, and artificially limits the size of a scene that can be loaded at one time. Using this pre-rendering cropping that this unlimited detail technology provides for should vastly improve the waste margin on what gets rendered. Effectively this gives a lower workload (at any given time) on the processor which means that larger scenes can be loaded without producing wasteful slowdowns.