Homepage: http://unlimiteddetailtechnology.com/
Youtube video: http://www.youtube.com/watch?v=Q-ATtrImCx4
The site/video looks rather typical of a project by developers rather than designers, if you get my meaning. Think, "Spring Settings Window". Anyway, it seems like sort of a combination between voxel and raytracing rendering, or something like that. Obviously my understanding of even modern polygonal rendering is very limited so form your own conclusions...
Interesting points, to me:
1) He claims to have pitched this to ATI and Nvidia, both declined to use it. Did they think it would conflict with their current business model, or did they think the technology sucked?
2) I don't see much of anything moving. Does the technology rely on objects being static?
3) I see shadows being rendered and they don't look like they're cooked onto the texture.
4) He seems to imply that all the rendering is done completely on the CPU, as it's software-based. What sort of CPU configuration does he have? Maybe this sort of thing would be justification for upgrading to the 8/12 core CPUs that are coming out in the near future.
5) He says he's trying to secure funding, etc, but is also quite confident that there will be marketable products in 16 months. 360 has 3 cores, PS3 has one "real" core and 7 floating point units (IIRC), so current consoles might be ideally suited already.
6) Many of the scenes he uses as demonstrations use huge numbers of the same object. Is this a limitation of the technique?
7) Some of the youtube comments are things like, "but what about collision detection, physics, hit-detection, etc"... These are separate systems even on a polygon-based engine. This technology he's demonstrating is just to give you pretty pictures and otherwise would have no effect on gameplay.
8) He rendered the video in 4:3? What a douche.
This whole thing is a bit elaborate to be a hoax, although there are probably more than a few exaggerations. However, as it is software based and CPUs are constantly advancing, a year from now they might not be so exaggerated. Would be nice if he released the tech demo he's using so we can all see it for ourselves, though.
Cool new rendering technology... maybe?
Moderator: Moderators
Cool new rendering technology... maybe?
Last edited by Caydr on 11 Mar 2010, 19:55, edited 3 times in total.
Re: Cool new rendering technology... maybe?
raytracing is something intel hoped to accomplish well in realtime with larabee iirc
a downside to the point cloud system, is we can only render in pinkscale
Re: Cool new rendering technology... maybe?
The "comparison" video on their site is very convincing:
http://www.mediafire.com/?0k2rm0koqjx
Their method currently has some obvious problems with shadows, especially in small areas and at the edges of the screen. The latter has an obvious workaround, the former is probably just an indication of the tech still needing some work.
http://www.mediafire.com/?0k2rm0koqjx
Their method currently has some obvious problems with shadows, especially in small areas and at the edges of the screen. The latter has an obvious workaround, the former is probably just an indication of the tech still needing some work.
Re: Cool new rendering technology... maybe?
so instead of new graphic cards players must buy tons of ram all the time?
Re: Cool new rendering technology... maybe?
Interesting
Re: Cool new rendering technology... maybe?
He says it could work on Nintendo Wii or even a cellphone. Perhaps not like in the video, mind you.
I get the impression that it's something akin to PhysX, a specialized software technology that doesn't require specific hardware if it's implemented at a low-level.
I also doubt that it's a universal panacea, you probably still need dedicated graphics hardware, but this would allow you to shift much of the load onto the CPU.
Perhaps engines could be a hybrid, using this "unlimited detail" system as a software solution for things that benefit from it, but still using traditional polygonal rendering for other things.
I get the impression that it's something akin to PhysX, a specialized software technology that doesn't require specific hardware if it's implemented at a low-level.
I also doubt that it's a universal panacea, you probably still need dedicated graphics hardware, but this would allow you to shift much of the load onto the CPU.
Perhaps engines could be a hybrid, using this "unlimited detail" system as a software solution for things that benefit from it, but still using traditional polygonal rendering for other things.
Last edited by Caydr on 11 Mar 2010, 20:37, edited 1 time in total.
Re: Cool new rendering technology... maybe?
I am quite skeptical of the veracity of his claims.
Re: Cool new rendering technology... maybe?
yes its quite interessting but voxels have been around for some time. so it is a new way to render them?
or how are these "point clouds" different to voxels?
dont really know much about this topic though.
or how are these "point clouds" different to voxels?
dont really know much about this topic though.
Re: Cool new rendering technology... maybe?
@Knorke: From what I understand, he is saying that his technology is a difference in voxel rendering technique rather than an entirely new... whatever.
Basically, rather than rendering all the geometry in a scene and then figuring out what covers what and doing all sorts of related math to figure out what the pixel's final color should be (as is done now), with his technique you do this in reverse.
You figure out the user's screen resolution, then for each pixel you determine what "atom" of what object that pixel will display, and you draw just that one "atom", rather than rendering the whole scene. Repeat for every pixel.
This immediately set off warning bells in my mind about aliasing, lighting, reflections, etc, but his demo doesn't *appear* to have problems with this. There are problems with shadows in the videos but he's clearly aware of them and doesn't consider it a problem with the technology. Reflections off water seem to work, large-scale shadows and lighting seems to work... Anti-aliasing is just an elementary post-processing effect which can be added (if it isn't already there).
Looking at some of the scenes in the comparison video for example, I see insane amounts of detail and a lot of hours must have gone into building those scenes. He must have a team working for him, and professionals like that need to be paid.
He's looking for investors, but nobody with the capital to able to donate a useful amount of money for a project like this would agree to do so until they'd seen the demo run live. ATI, Nvidia, Intel, or an engine middleware developer are the likely suspects. "Oh, hi, I saw your youtube video, here's a million dollars"... no.
So either he's a billionaire hoaxing the gaming world just for shits and giggles, or he's serious and the technology actually works. There's not really a third possibility, there's too much money involved.
Well, the third possibility is, it works but has a crippling flaw, like it is totally incapable of drawing anything good at a distance without severe aliasing, or it's simply too difficult to work with.
Basically, rather than rendering all the geometry in a scene and then figuring out what covers what and doing all sorts of related math to figure out what the pixel's final color should be (as is done now), with his technique you do this in reverse.
You figure out the user's screen resolution, then for each pixel you determine what "atom" of what object that pixel will display, and you draw just that one "atom", rather than rendering the whole scene. Repeat for every pixel.
This immediately set off warning bells in my mind about aliasing, lighting, reflections, etc, but his demo doesn't *appear* to have problems with this. There are problems with shadows in the videos but he's clearly aware of them and doesn't consider it a problem with the technology. Reflections off water seem to work, large-scale shadows and lighting seems to work... Anti-aliasing is just an elementary post-processing effect which can be added (if it isn't already there).
Specifically? Unless he took the trouble to set up a whole render farm just to make a couple of hoax videos, he can't be bullshitting to a great extent. It would take HOURS to render a single frame with a hundredth that amount of detail in Max - on my computer, anyway.Peet wrote:I am quite skeptical of the veracity of his claims.
Looking at some of the scenes in the comparison video for example, I see insane amounts of detail and a lot of hours must have gone into building those scenes. He must have a team working for him, and professionals like that need to be paid.
He's looking for investors, but nobody with the capital to able to donate a useful amount of money for a project like this would agree to do so until they'd seen the demo run live. ATI, Nvidia, Intel, or an engine middleware developer are the likely suspects. "Oh, hi, I saw your youtube video, here's a million dollars"... no.
So either he's a billionaire hoaxing the gaming world just for shits and giggles, or he's serious and the technology actually works. There's not really a third possibility, there's too much money involved.
Well, the third possibility is, it works but has a crippling flaw, like it is totally incapable of drawing anything good at a distance without severe aliasing, or it's simply too difficult to work with.
Re: Cool new rendering technology... maybe?
It's pretty much just altered voxels. Besides the fact that it's obviously very early in development (doesn't even have texture support yet, forget about complex shaders and lighting), but it is a COMPLETE shift from what the entire 3D world is centered around. EVERYTHING is made with polygons in mind, including production applications, renderers, game engines, etc. Modeling, texturing, shaders, lighting, rigging, animating, physics(rigid and softbody and cloth). He says it can import a polygon model, but until I see proof I'm skeptical, and that's just straight polygons, not taking into account any of the above things I mentioned.
And I also doubt it's as efficient he claims it is. He duplicates a 300k poly object 10k+ times, which is 3 BILLION faces or more. On a simple no shadows/textures scene(i.e. sculpting) even very good rigs will slow to a crawl at 10 million or so, that's talking quad cores and at least 2nd generation video cards, and he says he can display 3 billion on a single core with no gpu usage at a decent frame rate? Very very skeptical.
So it checks pixels rather than polygons, but it would still need to check surfaces and lighting somehow, which would mean checking polygons or something similar, which would mean either very good octree usage or he's just talking out of his ass. And if it's not checking polygons or using octrees it would have to just pick a random spot on the model to display the color of the pixel, which would lead to huge accuracy problems.
I don't think a breakthrough this big could have been made this late using technology similar to what we have now, and from what I can tell this really just similar to voxels.
ZBrush for example uses an implementation of voxels, which they call "2.5D", it doesn't use real polygons for the high-res display, which lets it double or triple the polycount when compared to conventional methods, so a mid-range computer can get a few million extra poly's without killing itself. This is a very high-profile company with a lot of smart programmers that has been around a very long time, and then suddenly some nobody programmer develops a new system using similar technology that completely blows everything else out of the water? Sounds fishy to me.
But even if it is all true, it still boils down to whether or not he will be able to seamlessly integrate it with the polygon workflow that the entire industry is built on. If that can't happen, this won't be adopted, period. Polygons are king.
Honestly I think the industry will just move towards dynamic polygon meshes, rather than some system like this, where a very high-poly model is the base, and then the polycount is dynamically lowered based on distance, there are already sculpting applications that do this, and possibly even some game engines already do, so it's definitely possible. This would allow for a much better scale of detail than LoD models do, where close views could have very very high poly counts, while far away views could be lowered tremendously. Not to mention it would save a lot of work for the artists, making LoD models and textures, the only question is how resource intensive it actually is to dynamically change polycounts on large scale scenes, seeing as a sculpting application is quite a bit different than a game engine for efficiency in things like this.
Woo long post. Somebody is bored.
And I also doubt it's as efficient he claims it is. He duplicates a 300k poly object 10k+ times, which is 3 BILLION faces or more. On a simple no shadows/textures scene(i.e. sculpting) even very good rigs will slow to a crawl at 10 million or so, that's talking quad cores and at least 2nd generation video cards, and he says he can display 3 billion on a single core with no gpu usage at a decent frame rate? Very very skeptical.
So it checks pixels rather than polygons, but it would still need to check surfaces and lighting somehow, which would mean checking polygons or something similar, which would mean either very good octree usage or he's just talking out of his ass. And if it's not checking polygons or using octrees it would have to just pick a random spot on the model to display the color of the pixel, which would lead to huge accuracy problems.
I don't think a breakthrough this big could have been made this late using technology similar to what we have now, and from what I can tell this really just similar to voxels.
ZBrush for example uses an implementation of voxels, which they call "2.5D", it doesn't use real polygons for the high-res display, which lets it double or triple the polycount when compared to conventional methods, so a mid-range computer can get a few million extra poly's without killing itself. This is a very high-profile company with a lot of smart programmers that has been around a very long time, and then suddenly some nobody programmer develops a new system using similar technology that completely blows everything else out of the water? Sounds fishy to me.
But even if it is all true, it still boils down to whether or not he will be able to seamlessly integrate it with the polygon workflow that the entire industry is built on. If that can't happen, this won't be adopted, period. Polygons are king.
Honestly I think the industry will just move towards dynamic polygon meshes, rather than some system like this, where a very high-poly model is the base, and then the polycount is dynamically lowered based on distance, there are already sculpting applications that do this, and possibly even some game engines already do, so it's definitely possible. This would allow for a much better scale of detail than LoD models do, where close views could have very very high poly counts, while far away views could be lowered tremendously. Not to mention it would save a lot of work for the artists, making LoD models and textures, the only question is how resource intensive it actually is to dynamically change polycounts on large scale scenes, seeing as a sculpting application is quite a bit different than a game engine for efficiency in things like this.
Woo long post. Somebody is bored.
Re: Cool new rendering technology... maybe?
Well, I was impressed. What they were talking about made sense to me, in terms of approach, and it's a path both Carmack, Intel and others have been trying to go down, so it's fairly plausible that there was a solution out there that required thinking around the issues in a new way.
I wrote them a fairly lengthy letter, critiquing their marketing presentation (tl:dnr version: it's missing a lot of things production artists need to see before they're going to insist on using it) and their business plan (I do not think they should try selling this to the hardware people; they should sell it to a huge corporation that doesn't have a stake in the hardware vendors and would be very happy to see the justification for a new OS every 3 years sold by a certain monopoly suddenly get a stake through its heart).
If anybody is actually interested in reading the full text, I'll probably make it available if I ever hear back from them (although I probably won't make their reply available, given the subject matter covered in my analysis).
I wrote them a fairly lengthy letter, critiquing their marketing presentation (tl:dnr version: it's missing a lot of things production artists need to see before they're going to insist on using it) and their business plan (I do not think they should try selling this to the hardware people; they should sell it to a huge corporation that doesn't have a stake in the hardware vendors and would be very happy to see the justification for a new OS every 3 years sold by a certain monopoly suddenly get a stake through its heart).
If anybody is actually interested in reading the full text, I'll probably make it available if I ever hear back from them (although I probably won't make their reply available, given the subject matter covered in my analysis).
Re: Cool new rendering technology... maybe?
What they done is pretty good - considering it runs on a single core laptop without 3d card.
But they will probably run into major obstacles when they want to support dynamic lighting, transparencies, moving objects and other effects. He implies it relies on some special way to store geometry in easily searchable way - such structures will likely be expensive to update.
They should really be more technical/specific if they want to get funding for this.
But they will probably run into major obstacles when they want to support dynamic lighting, transparencies, moving objects and other effects. He implies it relies on some special way to store geometry in easily searchable way - such structures will likely be expensive to update.
They should really be more technical/specific if they want to get funding for this.
Re: Cool new rendering technology... maybe?
Yeah, I covered that, animation systems, shaders and a few other things in my critique. Ooh, I didn't cover lighting models, though... oh well, I needed to get back to hacking up the GSoC stuff anyhow.dynamic lighting, transparencies, moving objects and other effects
-
SwiftSpear
- Classic Community Lead
- Posts: 7287
- Joined: 12 Aug 2005, 09:29
Re: Cool new rendering technology... maybe?
What university did Bruce Robert Dell and/or Greg Douglas go to? Why can I find no information about their company? Why don't they have a demo to download if their platform runs so effortlessly in software? This whole thing really really stinks...
Re: Cool new rendering technology... maybe?
Am I the only one who dislikes being talked to like a retard? The video is made for the wowing of unlimited retards.
-
SwiftSpear
- Classic Community Lead
- Posts: 7287
- Joined: 12 Aug 2005, 09:29
Re: Cool new rendering technology... maybe?
Which is why I feel like it stinks. It reeks of propaganda. Show me something that doesn't contain a patronizing British guy.Beherith wrote:Am I the only one who dislikes being talked to like a retard? The video is made for the wowing of unlimited retards.
Re: Cool new rendering technology... maybe?
I made a comment to that effect and got told off for it, apparently I want jargon to jerk off too.Beherith wrote:Am I the only one who dislikes being talked to like a retard? The video is made for the wowing of unlimited retards.
Re: Cool new rendering technology... maybe?
Their aim is probably to generate a bit of hype.. thats why the retard tone. It seems to work well.
The fact they are not searchable means nothing.. you can find your university or company by googling your name? I cannot and i'm happy for that :)
The fact they are not searchable means nothing.. you can find your university or company by googling your name? I cannot and i'm happy for that :)
Re: Cool new rendering technology... maybe?
If they had something serious to show they would.
It's not "hey,you silly devs you forgot to include more detail and a demo",It's "they are not providing any more information cause it's either bullshit or in very early steps with uncertain results".
Sending them any commentary on the video is kinda silly.
They know exactly what they are doing.
It's not "hey,you silly devs you forgot to include more detail and a demo",It's "they are not providing any more information cause it's either bullshit or in very early steps with uncertain results".
Sending them any commentary on the video is kinda silly.
They know exactly what they are doing.
-
Machete234
- Posts: 642
- Joined: 12 Feb 2010, 11:55
Re: Cool new rendering technology... maybe?
I think video hardware is not made to display this kind of voxel space in high resolutions.
So this would be software rendering if there is no special hardware for it, right?
There should be a way to use the rendering power of the gfx card, it would be useless if all that would be rendered by your cpu only.
Anyways I would like a demo to run on my pc to see if it really works that well.
So this would be software rendering if there is no special hardware for it, right?
There should be a way to use the rendering power of the gfx card, it would be useless if all that would be rendered by your cpu only.
Anyways I would like a demo to run on my pc to see if it really works that well.