Unlimited Detail Technology

[LlamaMaster]

Apparently there has been a 3D graphics system out for a while now that completely does away with the need for polygons. It uses point clouds on a per-pixel basis to render geometry with a nearly unlimited number of points. I'm not sure how true their claims are, but the tech demo certainly is interesting.

http://unlimiteddetailtechnology.com (http://www.youtube.com/watch?v=Q-ATtrImCx4&feature=player_embedded#)


http://www.youtube.com/watch?v=Q-ATtrImCx4&feature=player_embedded#

[Kornman00]

Holy crap. Brilliant. I hope a company who doesn't fear change and wants to excel the industry funds them!

[Heathen]

Pretty incredible.

[paladin]

cant wait to play with the sdk

[SMASH]

Just an FYI this was first announced in 2008 or maybe earlier. The lack of videos on their site and the lack of an executable demo points towards the idea it's a lie. As well they said 16 months... I find posts about this going back 2 year, where is it? Sorry to be a downer guys i know we'd all like to see something like this but it's not what they claim.

[Kornman00]

That or they just never got the funding

like they said, some people didn't like the idea of moving away from polygon rendering. Goes against their current money making model. I've seen/heard of stuff in gov't that could have had some major improvements applied (thus saving money and time) but (for example) since someone in a senior spot had some friends employed in jobs that would have been automated (and thus no longer needed) if they were applied, that senior quickly had the improvement ideas squashed before they reached anyone with a brain.

[Botolf]

I wouldn't be surprised if a gfx technology revolution comes in the near future, but this looks rather... suspicious. Static, really vague on how this technology performs and the requirements, etc.

[Good_Apollo]

Probably fake. No sensible company would pass up on this technology...unless it were fake or horribly flawed.

Oh and obligatory 'it's the government's fault.' I want to stay in the graces of some of our more esteemed members.

[flibitijibibo]

2 reasons why a company would pass it up:
1. Computing power necessary is probably lolhueg.
2. Unlimited detail is, humorously, much more definite than the poly system. Once there is a GPU that can run it at the standard refresh rate and maximum resolution, no other GPU needs to exist. Sure, there could be cheaper GPU's for lower resolutions, but the point remains that the profits would be logistic, not linear/exponential (like it is now).

Edit: Rebutting my own reasoning. #2 is a load of shit, because at some point the GPU would be outdated to the point where it could be combined with the normal processor, making stuff like Larrabee/Ion possible. At that point, physics would be the reason to upgrade, as well as the usual advancement in the CPU.

[Good_Apollo]

Consoles seem to have survived past the 32-bit limit as he already stated. There's always improvement to be made even when you hit a supposed wall so I don't buy that cop-out. It must be fake or flawed in some way, like those could all just be renders that took days to make for all we know...

[EDIT] Also their website seems to be pretty thrown together and amateurish for people so advanced in programming and desperately trying to sell or gain interest in their product.

[Cortexian]

I can see how it would work, and it makes sense to me but the lack of additional proof makes me skeptical that they've come to a stage where they can produce it over and over with the same results... Plus as flibit stated, once a GPU was released that could run something like this at lolhueg resolutions (2560x1600 is a good example for now, it's about the largest resolution currect PC monitors display) there would be no need to produce other GPU's except to cheap down at lower resolutions. However if this were to become a mainstream game design element, people would realize that they could play ANY GAME at said lolhueg resolution with the same performance, so why would anyone buy anything cheaper than the best? That would then lead to integration with the processor and rule out GPU's all together...

So yea, basically what Flibit said.

[Bodzilla]

Oh and obligatory 'it's the government's fault.' I want to stay in the graces of some of our more esteemed members.
what the fuck are you babbling about.

[Kornman00]

2. Unlimited detail is, humorously, much more definite than the poly system. Once there is a GPU that can run it at the standard refresh rate and maximum resolution, no other GPU needs to exist. Sure, there could be cheaper GPU's for lower resolutions, but the point remains that the profits would be logistic, not linear/exponential (like it is now).

It's like the electric car in the GPU world. Like the oil industry, these big name GPU companies wouldn't be getting the money for pumping out gfx cards like a queen of an ant hill. However, they wouldn't be needing to spend so much money in developing and continuing to support drivers and old cards. Nor would they have to employ an as-large R&D group.

However, this only covers graphics. GPUs can still improve physics and AI calculations, so there is still room to grow.

It's really all about taking risk. From a financial standpoint they see the risk (as you said) logistic, so they feel the security of the current model trumphs the benefit people would gain from this card.

So much for supply and demand huh?

E: Oh and think of how much easier game development and testing would be with a much smaller range of cards to test against? Gosh, you wouldn't be needing an as-huge income then anyway as you'd cut back on money spent on test machines

[Dwood]

I find the idea sound and don't see why it wouldn't work. My only problem is the question of what happens when you want to add physics to your 'game' ?

[CrAsHOvErRide]

I find the idea sound and don't see why it wouldn't work. My only problem is the question of what happens when you want to add physics to your 'game' ?
Huh? Adding physics would be even easier because you can treat the points like atoms.

I just have one problem with this...his pointe (NOT POINT) is the search algorithm...but it's not like such algorithms differ from now. The Point to Pixel concept is already being applied...so it's not like he reinvented the wheel. Besides, all those points have to be stored in memory...he never got into that.

[Kornman00]

Not like memory is going to be a major hurdle with 64-bit computers


Of course, when it comes to physics, they'd still probably continue using vertices and such. That or the physics sdks of today would have to reformat a lot of their frameworks.

[Limited]

It just seems something a little fishy about it all, the way he says its too, he sounds really sly and like hes hiding something/just plain fully lying.

[Dwood]

It just seems something a little fishy about it all, the way he says its too, he sounds really sly and like hes hiding something/just plain fully lying.

If they make a real sdk then all the shortfalls/shortcomings will be revealed in time.

[Limited]

Yeah, BIG if.

[CrAsHOvErRide]

Not like memory is going to be a major hurdle with 64-bit computers
The problem is not only space but access time :P Pretty sure for his search algo he has to take quite a lot of points into consideration...and that at least 30 times a second.

[Phopojijo]

It's very obvious that his entire purpose for that video is to just rhile up the uneducated gamer segment with all the analogies and such. Not how he never said the pros and cons of his technology or even referenced it in any technical way for someone who might understand what he's talking about.

P.S. -- BSP trees are sorting algorithms :p

As it stands right now, we're pretty non-limited by triangle count... except RAM (especially the consoles... some videocards have, alone, 4x the RAM as the Xbox 360... which has more than the PS3)... and RAM is limited in both situations.

The whole point of DX11 tessellation is to create more polygons with less RAM. Because that's the limit these days.

[Dwood]

Looking on it it looks to me like a fractal algo more than a gfx engine.

[Cojafoji]

Looking on it it looks to me like a fractal algo more than a gfx engine.
that's basically what he said in the video. wouldn't all of those points need some sort of prereferenced data? he says, if you plop in one position, it only renders what you'd be able to see; wouldn't that only work in real time if all points had some weird sort of angular notation?

eh nevermind, i have no idea what the fuck i'm talking about. i are confuzzled.

[CrAsHOvErRide]

that's basically what he said in the video. wouldn't all of those points need some sort of prereferenced data? he says, if you plop in one position, it only renders what you'd be able to see; wouldn't that only work in real time if all points had some weird sort of angular notation?

eh nevermind, i have no idea what the fuck i'm talking about. i are confuzzled.
No you are right and that's exactly where it fails. That system if used already now. Not only angular notation (or normal vectors) but also the position in space.

[Warsaw]

Sounds like it wants to be ray tracing or voxel rendering, but forgot to do the math that would make it one or the other thus making it a false concept.

[Dwood]

I dont think its false per se but rather over exaggerrated. Even if the demo is false i still think the idea is sound but only time will tell

[thehoodedsmack]

I would have thought that a system based on points would look all scrambly. At least with polygons you've got definitive solid surfaces.

[Rob Oplawar]

At first glance I thought it was pretty cool, but I agree that upon closer inspection it's very suspicious. I'll withhold judgment until I see some actual facts about this system rather than a vague video that contains zero technical info.

That said, it reminds me of a tech demo I saw some years ago about handling an arbitrary number of arbitrarily sized images in real time. It showed a user rapidly zooming in and out of thousands of multi megapixel images in real time- the way the software was able to do that was by cleverly figuring out which pixels from the images to store/manipulate in RAM at any given time. As was said in this 3D tech video, in theory you shouldn't need much more than what you can actually see on the screen at any given time. It therefore seems plausible to me that a technology like this could exist and work well, depending on the overhead required to search through all those points. Seems to me a clever addressing scheme could make it fairly straightforward.

I heard people attacking this for the difficulty of animating it. I want to point out that as far as I can tell this is a valid concern; if it only works because it only accesses however many points can be displayed on the screen at any given time, then animating those points causes serious issues because you'd have to access and modify all the points being animated, not just those that can be seen, if that makes sense. This could still be an excellent technology to use for mostly static environments, but I can't see it working for hundreds of separately animated enemies.


It's all very speculative, though. The video really doesn't tell us much.

[Limited]

So, how do you get a normal vector on a round object? Or would they just use the angle the player is looking at the object.

Also the fact the name is unlimited detail technology yet the example looks like ass. Making things look good, by creating it atom by atom (per say) would be very tricky.

Polygons are used to display the outlines of objects, which is how humans interpret objects and what not. This is why if you put a human in a completely white room with full ambient lighting with completely white objects in it, the human can not understand what the object is (by just looking).

[Hunter]

I like the fact how he was saying all this but he never actually wen't close to an object so that you could see it... he moved around quickly, and also I find it ironic how they said that they don't use flat planes to build there geometry, but then the camera goes over a tree in the preview and the leaf/branch was completly flat with a texture on... I don't beleive it, the guy sounds like an idiot, and if they had this great technology I would of thought they would at least a good looking logo instead of some cheap looking one which you see at the start and end :/

[Hunter]

I have never understood why the Xbox360 had such little memory in it... It only has something like 700MHz speed for the memory, acording to a website I am looking at. DDR3 supports 1200MHz, so why don't they just stick 4GB of that in...

I always experience a bit of lag on Midnight Club LA when a lot of objects are breaking (Because my amazing crashing ability). And sometimes on CoD Mw2 when airstrikes are called in (Sometimes that's connection lag though).

[Dwood]

I would of thought they would at least a good looking logo instead of some cheap looking one which you see at the start and end :/

It takes a lot of money to pay for a graphic designer. Something those guys should be spending on developing a real sdk. Also, just because they made the system doesn't mean they're good at making stuff in it. As Rob said, despite the little discrepancies I'm with-holding judgement.