"Hi everyone. We've been working very hard and we hope you like what we've made. This is just our 1 year report, after which we will probably go quiet again while we finish our work. This demo only shows what was ready at the time, we have a lot of really good stuff here but we are keeping it secret for now. (Yes grumpy forum people, we do have animation, but you'll just have to be patient.)

It's been a busy year, and all is going very well; we only have 9 people working here, but will be hiring a few more soon. We also have another piece of technology that isn't graphics, but does something game related that's also pretty clever, but we'll keep that secret for now.

We get a lot of fan mail from Germany and Brazil so I thought I'd put a link to the transcript of the video underneath just in case my accent is hard to understand.

I'd like to thank you all for your emails and kind support.

Kindest Regards, Bruce Robert Dell -- CEO Euclideon."

http://www.youtube.com/watch?v=UKUuUvDSXk4

SMART!!!!

Seen some of their videos, and it just comes off a bit like a scam to be honest. It's how they compare older games that have been made to run in lowest graphics settings to their "unlimited power" engine that kind of puts me off.

And their engine runs 20 fps on software now? Nice of them not to say what kind of hardware that is done with. A cluster of Cray XT5's? Also brings us around to another problem I see with this technology. The current graphics hardware is designed and built for the type of game engines we all use right now. Switching to the engine that these guys are supposedly making, would also mean switching to hardware optimized for that type of rendering engine, wouldn't it?

Awesome! I kinda wondered what the next logical step for video games would be. I thought it would actually be something renewed: games stored in solid state memory similar Nintendo cartridges, except being able to hold terabytes of game data.

The problem with point cloud graphics is that they're not really able to animate.
Additionally, the geometry can't really be handled by the polygon centric gpus, thus it has to run in cpu heavy "software" mode.
On a sidenote, I also do think that the quite generous use of repeating elements might be a factor in making it run smoothly, as I do think they "calculate" each object once and show it on multiple positions.

And he didn't go into technical details, as the type of processor it's running on etc. So it might as well be a "target shot" that's prerendered and not realtime.

Not to mention the BS numbers and contradictions: "unlimited" is a bold claim that can't be met. And the "100.000 times better" contradicts the "unlimited" claim, the proper factor would be "infinitely better".

So I remain sceptical. If it works out, yay, great times. But I don't count on it.

It is a scam, they make these videos so people fund them, and then they run away with the money.
The whole video is prerendered by the way, and in no way will it run in realtime on your PC. If they really want people to believe, they should put out a demo

Edit for the lulz: Picture is from 2003.

I don't want to dismiss their results as being a complete scam, but they have pretty much reinvented voxels once again. There is no realtime lighting and everything is static which allows for a great deal of optimization. Current GPUs are not designed to work with voxels at all so they'll probably have a hard time giving it some HW acceleration with OpenCL or similar technology.
Nice tech demo, but I don't see it being used in games anytime soon.

Personally I find their texturing method far more interesting than their unlimited 3d claims.

Plus it definitely feels like a scam. They could just laser scan a piece of street or small area of a park or some other relatively complex area and show how their system can show infinite amount of detail. But instead they use basic repetitive patterns and give no real technical information.

http://www.soundboard.com/sb/UnlimitedDetail.aspx

Because somebody could reverse-engineer parts of the source code and develop his own engine? I actually do believe that if it's a completely static scene with no real-time lighting it can run at 20 FPS on some badass CPU.

his speaking reminded me of this:


http://www.youtube.com/watch?v=cR6bCgJUcYw

Forgive me if i'm out of my depth here but...

Can anyone explain the idea of these "atoms" and unlimited detail?

I could get the idea of "unlimited detail" using vector graphics (i.e. a mathematical formula that scales well)

It just seems like they've made a ridiculously high-poly environment... impressive if realtime, but I don't think it is. Certainly not running purely in software!

EDIT - If theres one thing that Uni has taught me - it's that i'm out of my depth a lot more than I thought.

The "unlimited" detail is of course utter bullshit as you're always limited by hardware. The thing is that they don't use polygons at all, they use "volumetic pixles" or "voxels". Voxel is just like a pixel with position defined in three-dimensional environment. I could be talking complete claptrap here but I guess that all it takes is some raycasting to determine which voxels are visible from the current camera position. Polygon-based rendering is far less effective because it draws a lot of objects that in the end aren't visible. Of course if the scene was dynamic it would take a huge processing power to recalculate position of all voxels.

Quote from AstroBoy :

SMART!!!!

Derek?

My scamdar was correct it seems.

http://notch.tumblr.com/post/8386977075/its-a-scam

aw
well I'm back looking forward to blowing on cartridge contacts again

Quote from MadCatX :

Polygon-based rendering is far less effective because it draws a lot of objects that in the end aren't visible.

How does it? Occlusion culling or some other method of hidden surface determination is pretty much used in any game I know.
Voxels have been known for almost as long as polygons in computer graphics, and there's a reason they aren't used. This is not an area I'm especially clever in, but even I can figure this is likely a scam

tl;dr Voxels aren't new, and they aren't efficient. Polygons are better.

But there must be something out there better than polygons and textures/maps. Maybe it isn't this modern interpretation/reincarnation of voxels, but people have to fiddle to find solutions and improvements. Otherwise we'd still be using sprites!!

Here's what Notch thinks about it

Quote from RasmusL :

How does it? Occlusion culling or some other method of hidden surface determination is pretty much used in any game I know.

Even Occlusion culling has some extra overhead. It can't be effectively used for fully dynamic scenes (not that voxels don't have their own problems at this area) and there is still the "determine what should be drawn" and "draw it" part. Raycasting shouldn't have this kind of problem as it does the determining and drawing at the same time.

Quote :

Voxels aren't new, and they aren't efficient. Polygons are better.

Though I agree there's nothing wrong with polygons and textures, voxels are quite interesting technology that has IMHO a lot of potential.

Although we're making exponentially better video games every decade I think there is some component that is lacking. The way games are now it takes a super computer to run at highest quality and the next game that comes out a year later takes a little better computer to run at highest quality. I'm probably less knowledgeable than anyone here about technology but this is how I see it:

monochrome squares (pong)>> 2D sprites>> 3D polygons+textures>> ???>> ???>> ???>>

People saying that it can't be done because computers are programmed to run games with polygons...well then the status quo needs to get better, as it has always done.

Quote from flymike91 :

Although we're making exponentially better video games every decade I think there is some component that is lacking. The way games are now it takes a super computer to run at highest quality and the next game that comes out a year later takes a little better computer to run at highest quality. I'm probably less knowledgeable than anyone here about technology but this is how I see it:

monochrome squares (pong)>> 2D sprites>> 3D polygons+textures>> ???>> ???>> ???>>

People saying that it can't be done because computers are programmed to run games with polygons...well then the status quo needs to get better, as it has always done.

the future maybe a 3D version of Vectrex???

Quote from tristancliffe :

But there must be something out there better than polygons and textures/maps.

clearly that explains why its still used in professional cgi then
although ive heard that there are some rendering engines that do something voxel like and break everything down into spheres that can be calculated highly efficiently via raytracing... the basis is still a poly model though

The fact is that unless you have a system that can haul around petabytes of point cloud data (we don't) you're going to have to use some sort of algorithm to compress the data and it's going to be a lossy one too. Polygons are just a method of representing the most significant points in a given point cloud (the vertices) and interpolating the surface over them (the polygons).

I'd say polygon tesselation is the next quantum leap as far as surface representation goes. It still uses the discrete collection of the most significant points in the cloud (the vertices on each polygon) but instead of linearly interpolating them to create flat polygons it uses the limited point set given by the vertex data to infer where other points in the cloud should lie. It subdivides each polygon intelligently so the new vertices created by the subdivision follow the curve of the original surface. So a curve represented by a jagged angle of polygons can be interpolated into a smooth curve of geometry.

Currently interpolating normals across a polygon can achieve this effect but it breaks down at the edges of an object's silhouette - tesselation allows the silhouette to have nice curved transitions between the jagged edges too.

Quote from Shotglass :

clearly that explains why its still used in professional cgi then
although ive heard that there are some rendering engines that do something voxel like and break everything down into spheres that can be calculated highly efficiently via raytracing... the basis is still a poly model though

I don't think that means there is no 'better way'. In 200 years will all graphics be polygon based? I doubt it.