Unlimited Detail wants to KILL graphics cards

[Chunk3ym4n 10]

http://www.rockpapershotgun.com/2010/03/10/unlimited-detail-wants-to-kill-3d-cards/

Has anyone seen this? Amazing, and it doesn't need a GPU, it just runs on your CPU and it can even run on a mobile phone!

[Flash Thunder 10]

looks like an interesting concept, I dont think it will work out properly though.

That and ATi and Nividia would probably go out of business.

He didnt get into much of the technicality of the unlimited detail process.

[Big Dawg KS 6]

What an odd narrator guy.

http://unlimiteddetailtechnology.com/videos.html

Second video on their sight, they actually show a bit of ArmA2 as comparison.

Lol, he kind of insulted some of BI's work too... shame on him.

Edited March 10, 2010 by Big Dawg KS

[Flash Thunder 10]

What an odd narrator guy.

http://unlimiteddetailtechnology.com/videos.html

Second video on their sight, they actually show a bit of ArmA2 as comparison.

Lol, he kind of insulted some of BI's work too... shame on him.

Thanks for the link man, this stuff is really interesting, just think how Arma 2 would look with this??!!!!!! :eek:

No LOD switching even farther draw distances 20KM

Super high polygon characters

simply put it= Epic win. 3D particles, I wonder if someone will come up with some physics too.

Spread the video around!

[Daddl 10]

Isn't that basically the same as ray tracing crossed with voxel technology? Intel's been demonstrating real-time ray tracing for years now, running actual games with their engine. These guys are trying to win over investors/customers, so it's always good to sound like they've found something 'completely new' and 'revolutionary'.

Don't get me wrong, I'm sure real-time ray tracing (whether applied to polygon- or pointcloud-based scenes) will be the future of 3d graphics (if only because computational requirements are actually lower for more complex scenes than with traditional approaches), and the best innovation is often based on a clever combination of proven concepts. I just can't see the breaking news right now. Now when they've got their SDK ready and we see the first adoptions to games and animations, then it will get interesting.

[dm 9]

Wow, that video really hurt my brain.

It was as if the narrator was just saying "Has anyone really been far even as decided to use even go want to do look more like?" over and over and over and over and over and over and over and over...

[froggyluv 2133]

Everytime I see this program I think of that Sith Lord from Star Wars

"Unliiimited Powwwer!!!"

...then I hear that narrators Lenton-like voice and see the Mario Cart world, and my dreams come crashing back to Earth. Still, the concept is hella cool and can't wait to see this thing in an actual game.

[NeMeSiS 11]

We'll see how UNLIMITED AWESOME it is once we have a techdemo running on our own PC's.

[maddogx 13]

That guy needs some better material for his Demos. In theory it could look photorealistic, but at the moment it's just cartoonish. Still, quite impressive stuff.

[Flash Thunder 10]

That guy needs some better material for his Demos. In theory it could look photorealistic, but at the moment it's just cartoonish. Still, quite impressive stuff.

theres a second video on their site, showing off Arma2 comparisons (pretty funny), maybe we could ask them to remake an Arma 2 scene using their tech that would sell me. :)

Yeah I do agree, Use a better camera to shoot/capture footage with +HD not blurry footage, give us some more technical talk so it doesnt sound like a bunch of BS.

Arma 2 scene with textures and lighting in techdemo and HD footage and Ill jump on the bandwagon or maybe something of crysis nature.

for now its interesting to watch but im a skeptic.

[max power 21]

Isn't that basically the same as ray tracing crossed with voxel technology? Intel's been demonstrating real-time ray tracing for years now, running actual games with their engine. These guys are trying to win over investors/customers, so it's always good to sound like they've found something 'completely new' and 'revolutionary'.

Don't get me wrong, I'm sure real-time ray tracing (whether applied to polygon- or pointcloud-based scenes) will be the future of 3d graphics (if only because computational requirements are actually lower for more complex scenes than with traditional approaches), and the best innovation is often based on a clever combination of proven concepts. I just can't see the breaking news right now. Now when they've got their SDK ready and we see the first adoptions to games and animations, then it will get interesting.

He explicitly says it's not ray-tracing. It's based off of a search engine type method. He also says it's not voxels, but both voxels and this method are based on point cloud data.

Maybe he's just trying to distance himself from voxel technology because he doesn't want to be associated with Duke Nukem 3d.

Edited March 10, 2010 by Max Power

[Emberwolf 0]

They admit to not being artists, but they sure as hell aren't presenters either. That was the same unprofessional condescending tone I'd expect from anyone trying to hard to be funny writing a game review or something.

It's a good concept and I hope it's real, but they're going to have a difficult time trying to sell their tech if this is how they express themselves.

On a positive note, the idea that something like this could be real makes for some happy thoughts. It would be an intense relief as an artist to not have to worry about polycount while modeling stuff. In recent projects I've had to spend most of my time thinking up ways to get an effect I want to be simple on resources, and then even more time taking a model and strategically stripping geometry off until the FPS is happy, as well as replacing former geometry with texture effects. Someday I'd like to actually stop worrying about this stuff.

[instagoat 133]

Nice to look at, however, the longer this graphics craze goes on, the less benefit I see in it.

Good graphics can´t rescue a bad game, and a good game will stand without good graphics at all. (Examples would be The Void and Pathologic, or indeed Operation Flashpoint).

Instead of people pursueing the holy grail of infinitely detailed flowerpots, how about somebody concentrating on making a capable ingame AI? That´s something that´s been sort of ignored ever since Unreal was released.

Arma 2 is about the pinnacle of it, and while not bad, I think the whole Industry would be much further ahead in this departement now if they had stopped concentrating on fancy graphics two years ago.

Unfortunately good graphics is something people want, plus, it drives the technology development. From a business point of view, nice... but as a gamer, I´ve to say that this is something I´ll gladly ignore in favour of other, more interesting developments in the field.

[-FHA-Dynamo 0]

Nice to look at, however, the longer this graphics craze goes on, the less benefit I see in it.

Good graphics can´t rescue a bad game, and a good game will stand without good graphics at all. (Examples would be The Void and Pathologic, or indeed Operation Flashpoint).

Instead of people pursueing the holy grail of infinitely detailed flowerpots, how about somebody concentrating on making a capable ingame AI? That´s something that´s been sort of ignored ever since Unreal was released.

Arma 2 is about the pinnacle of it, and while not bad, I think the whole Industry would be much further ahead in this departement now if they had stopped concentrating on fancy graphics two years ago.

Unfortunately good graphics is something people want, plus, it drives the technology development. From a business point of view, nice... but as a gamer, I´ve to say that this is something I´ll gladly ignore in favour of other, more interesting developments in the field.

well, when you free yourself from the limitations of current graphics tech, you are now free to spend more dev time on making a better game.

what if BIS had this tech before they started on ArmA2, think of how much time they could have saved from optimizing ArmA2's performance and how that time could have been spent on game dynamics and campaign. if this tech really does work, then we wouldn't need our massive and expensive video cards. instead we would have dedicated physics cards, and that would be sweet.

[-)rStrangelove 0]

Just looked at the video for the 1st time, VERY spectacular indeed.

And the guy is not trying to sell something, he basically just says that both Nvidia and ATI were following a different road (rendering of polygons) than this 'unlimited detail' project.

Rendering polygons:

- render as much polygons as you can within certain fps limits and then convert all 3D coords to 2D coords to show them on screen.

Rendering points in the 'unlimited detail' project:

- take each pixel from the 2D screen of the user resolution and start searching what pixel to show there.

Simple isnt it?

And the guy is right. Whether you have 10 objects in line or 100000000 objects, only the nearest objects will be clearly visible and all others will be partly or totally hidden. So no point to 'render' all these 100000000 objects, just find a way to search which pixel on screen shows what part of the nearest object in 3d space.

I like it. :) It's not a game yet but i like it.

[nodunit 397]

It is always great to look at engines at work and new and sometimes intuitive work arounds but as great as that truly is it is still the engine we don't know how it will perform in a game exactly.

To give them credit it looks damn good and if it can run in a game as it is stated then I would be amazed however the question I would pose is how would one create content, do we build it atom by atom?

[dm 9]

)rStrangelove;1587575']And the guy is right. Whether you have 10 objects in line or 100000000 objects' date=' only the nearest objects will be clearly visible and all others will be partly or totally hidden. So no point to 'render' all these 100000000 objects, just find a way to search which pixel on screen shows what part of the nearest object in 3d space.[/quote']

This is what realtime 3d engines do already - the cull everything that isnt going to be seen on screen. So that part of the tech is nothing new and nothing amazing.

Also, I was somewhat worried about how he said "this is running software"

Its all well and good for their tech demo, but what happens when you add in dynamic lighting, AI, physics and all the other stuff that would be fighting for CPU time?

[mr burns 131]

unlimited detail ... unlimited detail .. unlimited ... wait, what?

Way to go on a proper presentation. If they make a good game out of it i might try again, but marketing bs they can keep for themselves.

[max power 21]

To give them credit it looks damn good and if it can run in a game as it is stated then I would be amazed however the question I would pose is how would one create content, do we build it atom by atom?

They suggest building content in polygons using feature movie resolution standards, then converting it using an x,y,z,colour => voxel converter.

[Whiskey_Tango 10]

I like the concept of it and I hope it actually works. Everything takes time and I wouldnt expect a potential dev to just jump into it and make a flatout end all be all game. If physics and lighting dont kill it then it has the potential to change gaming in a big way.

[STALKERGB 6]

so if i had a resolution of 1650 x 1080 then this system would have to render/process 1782000 points? if thats so then although it could in theory display trillions of points it will only really display millions?

Edited March 11, 2010 by STALKERGB

[Bulldogs 10]

The problem they have right now with their tech demos is funding. They can't afford to hire artist to work on their software (and they probably haven't even created efficient tools, right now it'd all be created based on lines of code like Ray Tracing), so they're not going to get anything good looking.

Although that's not the point of the videos either, the point is to show that using the system you can create actualy "round" objects, which due to polygon count restrictions has been completely impossible up til now.

[Daddl 10]

He explicitly says it's not ray-tracing. It's based off of a search engine type method. He also says it's not voxels, but both voxels and this method are based on point cloud data.

That's what I was saying - they build on existing techniques and call it something new. I'm not saying it's traditional raytracing or it's traditional voxel technology. But from the description it definitely builds on *both* technologies. If you have a technology and you want publicity you *got* to distinguish yourself from the competition. "No, this is no corkscrew at all - this the all new unlimited wine opening drill - the UWOD 2000!"

[sparks50 0]

This is sweet, imagine an never ending view distance with highly detailed grass and vegetation all the way.

Hoping the best for this technology, also I'm tired of buying a new kickass GFX card every two years.

[walker 0]

Hi all

The thing they are doing that is different from ray tracing and voxels is that they only drawing the data they need.

It sounds slightly similar to the neural net methods we used in UA to calculate ballistics for mutiple rounds beyond the ~2000m range. Both problems result in exponetial increases in calculations.

They are using a search algorithm, we used a neural net. The major problem in both cases is creating the data in the first place. For us we used programs and algorithms external to the Real Virtuality Engine to create a database, and multiple overnight interations to teach the neural net. I guess they can use data gathered from scans and pre renders.

Essentialy both are abstraction methods. Another problem might be data loss and incorrect data, in this case perhaps shown in the artifacts they have in the renders. If this last is the case then their search algorithm will always suffer from it.

Kind Regards walker

Edited March 12, 2010 by walker