Unlimited detail in games

[groda]

Nope, I don't work there, just used google. Seems they have pulled that video, but fortunately it seems there is a guy who caught it on tape before they pulled it. http://www.youtube.com/watch?v=KRP5km8gHoo Starts at 1.15. (My guess is that they don't want to show some videos before they release a ready product. Maybe that Aero3Dpro that they plan to make available in November.)

[cybereality]

Well this technology would work well for aerial shots since they don't need the kind of things that the engine is bad at. Namely skinned animation, real-time lighting/shadowing, physics, or really any kind of dynamic objects. So in this case voxels are sufficient. I still don't think they are ready for a commercial AAA game, but maybe one day.

[groda]

A bit off-topic; which of the following technologies do you think shows most promise for the future?

Polygons + Tessellation.
Interesting examples of tessellation by Valve in 2008: http://developer.amd.com/media/gpu_asse ... _GDC08.pdf (pages 15, 27, 28)
Id Software 2008:
http://www.nvidia.com/object/tessellation.html (2nd pic)
Cryengine 3, 2012:
http://www.youtube.com/watch?v=nMwk2Zi0c3o

Online cloud rendered gaming, like Gaikai or Onlive.
Example also from 2008. Otoy founder.
http://www.youtube.com/watch?v=x5yVjHaJ ... ure=relmfu
http://www.youtube.com/watch?v=XnihU4zCXe8

Atom-based rendering, like Unlimited Detail or Atomontage.

Ray-tracing,
Nvidia, 2012:
http://www.youtube.com/watch?v=h5mRRElXy-w
iCEnhancer, 2012: http://www.cgchannel.com/2012/07/eye-ca ... irectx-11/

Any other technology I missed?

Also, when do you think we will see realistic hair in computer games, and which technology will be first?

[cybereality]

Well CryEngine3, if only because it is totally usable today for full production AAA games. The other stuff is a tech demo at best.

[groda]

The way I see it raytracing (cloud rendered and normal) and atom-based engines shows most promise for the future.

Advantages of atom-based engines: Possibility of "real" physics in games.
Difficulties: How to make all the different types of refraction, light bouncing and subsurface scattering?

Advantages of raytracing: Accurate rendering of light.
Difficulties: How to make look sharp when moving the "camera" fast.

It seems Otoy are not only making an online cloud rendering engine (Octane), but also an offline renderer for games called Brigade 2 ( http://icelaglace.wordpress.com/2012/09 ... de-engine/ ). See here for screenshots: https://www.facebook.com/media/set/?set ... 366&type=3

[groda]

A more recent clip from Euclideon about their Geoverse: http://www.youtube.com/watch?v=d7THYanat-0

[cybereality]

Interesting work. Seems like the professional market may be a better fit for this technology than gaming.

[Fredz]

Can't we just let this thread die ? This thing has been a scam from the start.

[cybereality]

Can't we just let this thread die ? This thing has been a scam from the start.

Did you watch their latest video? It does seem like they have something real. I do still think its not viable for gaming yet, but maybe other markets.

[Fredz]

Yes I did watch the video, it's pretty much similar to the one posted in september about AEROmetrex point cloud rendering, which was rendered offline in Blender. I don't see anything new in there, apart for the technology used to scan 3D scenes which is interesting, but is not theirs.

[groda]

Fredz, how can you be so sure it is a scam? Judging from what I can find on the internet, at least two companies have already bought and used the Geoverse technology, AEROmetrex and Schenkler Mapping.

And concerning gaming, I still hope that Euclideon is on the right track. According interviews earlier this year their goal was to first publish the Geoverse tech, and then concentrate more on finishing their engine for gaming. Read somewhere that their plan was to have it finished enough to present somewhere around the end of this year.

[Fredz]

Fredz, how can you be so sure it is a scam?

Because that's basically the recipe for a scam : announce breakthrough technologies and deliver nothing.

They've been at it for more than 10 years and haven't presented anything remotely verifiable or even believable. And no, posting YouTube videos doesn't count as a proof in any form, anyone can publish an offline render and claim it's realtime.

Judging from what I can find on the internet, at least two companies have already bought and used the Geoverse technology, AEROmetrex and Schenkler Mapping.

The video posted by AEROmetrex was not realtime as they posted in the comments, that's quite far from the original claim.

That last video also isn't professional at all, it's extremely light on technical details side and it's full of buzz words that make no sense. Also it doesn't seem to be related to their "unlimited detail" concept anymore, now they only show a typical cloud point rendering of a not so great quality.

Their summary in 3 points about "The Advantage of Euclideons Technology" is also pretty much laughable and makes no sense :
- Unlimited power to visualization : they claim to be able to render graphics of any size without point decimation. That's just plain stupid, how can anyone claim to be able to render in realtime any number of points ?
- Instantanly Loads Data Of Any Size : seriously, how can anyone claim something like this ? Same as the first point, there are finite possibilities for data transmissions, even with high speed fiber-optic communications. What will they claim next, that they have invented the perpetual motion ?
- Run Over a Network Or The Internet : oh, like it has never been done before ? They've never heard about people that have been doing this for decades like say, Google with Google Earth ?

They may have an interesting technology for point-cloud rendering over network connections, but they are overselling it as unlimited, which can clearly not be the case. They also didn't publish any technical information about what they do, so nobody can verify their claims. So yes, it's basically a scam.

[groda]

Yes, AEROmetrex have been using non-realtime renderers to produce videos, and they have only shown a very limited glimpse of how they have used the Geoverse technology. Still, it appears from their comments on one of their videos that they have indeed been using it.

"Hi there. This video has been created using our generated 3D models and rendered using commercial 3D rendering softwares. We are using Euclideon technology to ingest and interact in real-time with the very large amount of 3D data we acquire. You will see a short sequence of Geoverse in action at 0:17s.
Aerometrex" from http://www.youtube.com/watch?v=ABdIwKpBLQw

Therefore I guess they must have been "fooled" by Euclideon, something that seems highly unlikely since they have tried the technology. But, anyways, this is just a comment from a youtube clip. Let's wait and see.

[Fredz]

That technology could very well be useful to them, I really have nothing against that. What I don't like is the many false claims of Euclideon with their "unlimited" buzz words. They may have an interesting technology for point-cloud rendering, but that's not really earth-shattering nor new, and it certainly doesn't deserve the press they managed to get.

[cybereality]

Well I do agree that the word "unlimited" is totally bogus. Nothing on this earth is "unlimited." But aside from that it does seem like they have something, even if they are over-selling it.

[Okta]

Well I do agree that the word "unlimited" is totally bogus. Nothing on this earth is "unlimited." But aside from that it does seem like they have something, even if they are over-selling it.

Agreed. There search algorithm is a breakthrough (if it works as advertised) but the unlimited thing is a bit annoying.

[BlackShark]

I have to disagree. If their engine works as advertised, they are right to use the word "unlimited".
Their CEO explains it in this interview (second question) http://www.youtube.com/watch?v=oCgs--hv ... ure=relmfu
The logic is not to use the word "unlimited" as the imaginary concept of something which keeps going on forever, but rather use the actual technical definition of the word unlimited :
-> if you settle a big number and say "that number" is the limit, then the number "that number plus one" exists... (I think it's also how infinite numbers are defined in math). Whatever number of elements in the scene (typically polygons in realtime 3D graphics or points in their case) you can always render this number of elements plus one without killing the framerate.

The actual limit in their technology isn't what the engine is capable of, but rather "does the data exist ?" (has it been laser scanned or has an artist created it). If that data does exist and is available (is your hard drive big enough to store it), their engine can display it in real time.
The 'it works over a network" claim is an interesting one because current 3D display techniques require huge amounts of bandwidth (mostly for textures and buffers), in order to achieve this bandwidth the hard drive is not fast enough, so the most heavy elements (typically : textures) have to be loaded from the hard drive into super-fast RAM embedded on the graphics card. Even though most of that data isn't actually visible because the render process isn't able to know which pixels will be visible until after these pixels are processed.
This claim works together with "it loads the data instantly". If it does one, the other is a given. If it can't do one, it can't do the other.
And you can't seriously compare google earth with their technology. Getting google earth to run at all forces buildings to have a ridiculously low amount of polygons and low resolution texture. Google earth is a map that works very well at far and moderate distances because it uses many copies of the photos data at various quality and swaps between them at various distances, but it doesn't do seamless transition at all, it's not designed to do that. Even if you had your computer at the google earth picture datacenter, you would still have to deal with picture quality and terrain/building model swaps.

[Okta]

I have to disagree. If their engine works as advertised, they are right to use the word "unlimited".
Their CEO explains it in this interview (second question) http://www.youtube.com/watch?v=oCgs--hv ... ure=relmfu
The logic is not to use the word "unlimited" as the imaginary concept of something which keeps going on forever, but as the actual technical definition of the word unlimited :
-> if you settle a big number and say "that number" is the limit, then the number "that number plus one" exists... and their engine still works.

The actual limit in their technology isn't what the engine is capable of, but rather "does the data exist ?" (has it been laser scanned or has an artist created it). If that data does exist and is available (is your hard drive big enough to store it), their engine can display it in real time.
The 'it works over a network" claim is an interesting one because current 3D display techniques require huge amounts of bandwidth (mostly for textures and buffers), in order to achieve this bandwidth the hard drive is not fast enough, so the most heavy elements (typically : textures) have to be loaded from the hard drive into super-fast RAM embedded on the graphics card. Even though most of that data isn't actually visible because the render process isn't able to know which pixels will be visible until after these pixels are processed.
This claim works together with "it loads the data instantly". If it does one, the other is a given. If it can't do one, it can't do the other.
And you can't seriously compare google earth with their technology. Getting google earth to run at all forces buildings to have a ridiculously low amount of polygons and low resolution texture. Google earth is a map that works very well at far and moderate distances because it uses many copies of the photos data at various quality and swaps between them at various distances, but it doesn't do seamless transition at all, it's not designed to do that. Even if you had your computer at the google earth picture datacenter, you would still have to deal with picture quality swaps.

Does the data exist? If that is the question then their engine is no more unlimited than any other. He explains that their engine rendering speed (edit , there detail is still limited by the resolution which is still XGA untill they can 'optimise') is unlimited because the magic algorithm never needs to render more data than the screens resolution at any time which is the same as the amount of 'atoms'. Perhaps the bandwidth speed is gained because there is no texture data, but instead each atom is displayed as its assigned colour, giving photo like quality as high resolutions. I still have no idea how they would achieve any of this but its mighty interesting

[BlackShark]

Does the data exist? If that is the question then their engine is no more unlimited than any other.

If there is no data, then I'm sure almost every single engine out there will be fully able to display an unlimited amount of "nothing".
But I do know many engines which aren't happy with "nothing" and require at least the artist to provide a skybox due to the lack of defalut behaviour when encountering void.

[Okta]

The point being they are touting it as unlimited detail, which is false as the detail is determined by the resolution. And as for the data itself? Being able to stream it suggests that it has a tiny footprint, but once again is limited by storage space, as efficient as it may be.

So not unlimited in any sense

[BlackShark]

The point being they are touting it as unlimited detail, which is false as the detail is determined by the resolution. And as for the data itself? Being able to stream it suggests that it has a tiny footprint, but once again is limited by storage space, as efficient as it may be.

So not unlimited in any sense

I'm not worried about being tied to the resolution of a laptop running non-optimised code entirely on the CPU. I am confident that as soon as they start using GPUs (which they claim is possible) this argument will be over.
Besides, the claim is not unlimited everything but unlimited detail. And that detail is what artists create, not what window size you use. And mo matter how big your screen is, in order to see all the detail you have to zoom in on objects or bring your character very close to them in order to see how much detail the artists have put into their games.
Large amounts of detail will always have large footprints, since the detail does not come out of thin air : it has to be stored somewhere. The thing that is lower is the bandwidth since their engine is able to only load the data it needs from the very start rather than loading everything first and sorting later.

One interesting bit will be finding ways to reduce the huge amount of data that their engine can produce to be able to distribute the games. A bit like the file size issue of Rage, I remember a Carmack interview where he explains that with the megatexture technique (which could very well be marketed as unlimited texture) where after having given super high quality standards to their artists in pre-production, the guys find out that the game they've created takes multiple terabytes to store. They did not have any problems displaying on their PCs, they just cannot ship it on DVDs.

[groda]

A new interview with Bruce Dell at Euclideon: http://www.youtube.com/watch?v=j5vbLmH8YS0

[geekmaster]

My first GPU was a PCI card from Matrox that contained a voxel engine. It was not a video card, but just a coprocessor. It was from before the 3DFX card came out. The voxel engine never became popular because video games jumped on the 3DFX bandwagon instead.

Voxel processing is lower overhead than polygon processing, but you cannot easily model realistic features like lighting and reflections. Simulated lighting and shadows are typically "painted on" to the voxels. The Euclideon engine will probably have the same limitations.

However, for applications where level of detail is more important than lighting and reflections, I certainly have faith in the Euclideon engine. Each method (voxels or polygons) has its place for delivering whichever feature set an app needs.

Perhaps my preference for voxel tech is biased by my early exposure to voxel rendering hardware from before the 3DFX (and modern polygon tesselation based GPUs) were a commercial product. I have had a fascination with voxels ever since then.

What I have been playing with over many years (since the Matrox GPU came out) is a hybrid 3D dithered MIP map approach, which works similarly to what the Euclideon videos describe. The dithering thing lets me interleave my 3D MIP maps, so I can sort them voxels geographically and then just discard rows/colums/planes by truncating lower bits off my coordinates, for distant detail. Perhaps Euclideon is doing something similar to my methods, but they are secretive about it so I can only speculate. Now that the Rift is here, it may be time for me take my old project of the (far, far) back burner and run with it... My voxel data compresses well because the voxels near each other typically have similar properties/textures, and they are stored relative to their geographic position. This also lets me index quickly to the ROI (region of interest) to be displayed. And of course, I lookup pixels to put on each display coordinate too, like Euclideon described.

Anyway, because of my past experience, I am a true believer in the Euclideon engine (for those applications that can use it). Modern GPU-based engines will still have their place, for extra realistic simulated environmental effects. For low performance hardware (like the Raspberry Pi), perhaps voxel tech can have an advantage, but even the RasPi has a modern GPU, so they only way to know for sure is to compare games or demos that use both methods.

Or... perhaps my fascination with voxel tech is misplaced because of my early experience with it, and my time would be better spent learning Unity and UDK?

[colocolo]

Footage from Spatial Sciences Symposium 2012.
This thing is becoming more and more mysterious. In October he also will speak in Essen in front of
geospatial experts.
I dont understand how they stream the data directly from the HDD. Reaction time of an HDD is already about 20ms. So in order to get all the necessary data from it in the right time they must sort of
install the data in a very special order.
http://www.mediavisionz.net/mediavision ... R_SSS_2012
and an interview with a dutch guy:
http://translate.google.com/translate?s ... -revolutie

original
http://www.gamer.nl/interview/298111/de ... -revolutie

Do you have a specific game with which you would want to use the technology of Euclideon?

I would like to say that such a game is already in production. I would have liked to show you a few screenshots. Yet it has been said to me that I cannot tell anything about it this moment. Unfortunately.

[yuriythebest]

http://www.youtube.com/watch?v=Irf-HJ4fBls

Witchcraft! imagine a flight sim with this + oculus

ALSO - WHY IS THERE NO EMOTICON FOR THE OCULUS?

I tried to find this Geoverse program for download however the official site only allows you to download the brochure and it is not yet available on TPB

[colocolo]

Euclideon signing exclusive distribution agreement for Geoverse.
http://www.meixner.com/en/news/
My dreams will come true.
Cant be a hoax anymore. What do you think?