This is the archive of the old Colorless! Go to the new Colorless →

Unlimited Detail in Real-time-rendering. (Thread)

Big claims!
So basically, a technology company named 'Euclideon' have claimed they have developed a system that can produce trillions upwards of rendered atom-points, or a point-clouds of angles within a flat polygons, as game developers are familiar with. With this, infinite details in real-time can be rendered within a programme producing the most realistic geometrical graphics. Watch the video below if you want to see more.

And of course, discuss!

If there's already a thread which I haven't seen, let me know - I'll take this down or it can be merged, idk.

As for my opinion, it's a big claim and I'd like to see real-time animation happening within this system. But I would love to believe this is true. It would be sooo awesome for the future.

Also, here's an interview with an employee of Euclideon (I forgot his title).

It's a scam
But Notch, it's NOT a scam!

If it's real, this is going to be revolutionary. Imagine the possibilities for physics...

These things are what shakes me in trusting their progress. I want to see animation and applicable physics to the demo they've done! I also want dynamics. Basically things that make a good game a game.

Also, I'm sure Notch can be trusted, but I do think he's just found someone to deter new customers from buying Minecraft, a game which isn't very graphically driven. Although I don't see why he sees a threat, he's part of a small individual games company who have made more than 3 million sales and 8 million potential buyers.

@Caarbite
Doing a little research, you'll see that what Notch is talking about are really downsides to voxel renderers, not really a plan to keep others from taking away his business.

This voxel idea won't work if it's fully rendered using Voxels. What the guys did at Euclidean is simply partial voxel rendering and repeat rendering (basically copies) of trees, rocks, dirt in the game. If one was to truly render an entire map with voxels, you would need massive processing ability from the GPU. This can't work!

The technology has supposedly been around for years now, and there have already been videos released of engines using voxels to render.

What annoys me, is how he ( the narrator ) goes on to mention they've much greater shadow, and footage at a higher FPS, yet use the crappy footage in comparison. Not the first time I've heard the near exact same words on the very exact same concept.

TL;DR

Voxels suck currently.

Fair enough. But their other extreme claim is that they're not using voxels. And they don't work with graphics processors. It's all completely software systems and runs off a CPU.
And that's where I see that they're not being truthful about it.

They're most obviously lying about it, but rather than lie blatantly, they choose to deceive everyone covertly. It's a pretty desperate hook for money.

I heard they already have a government grant of 2 million AU Dollars. And they 'claim' that they have and will deny any further investments.
But yeah, obviously way too many claims to be making.

I don't know about this, the fact people are so up in arms about it being legit is making me wonder, but from my intial impression it's very dodgy feeling, kinda like those bloke down the market shouting AUTHENTIC BRAND SHOES FOR THE PROICE OF AN APPLE, or something when really they're just flogging some knock off shit from china. This is what it feels like, sure the technology they're spouting may exist but they're making wayyyy too grand a claim here and ofc the fact they're very tight lipped about it. Until I see someone legit's and not some lame ass blogger who I barely care about, impression then I will believe them, but for now it feels too much like a ruse.

Except the lame ass blogger is using fact and reasoning to prove it wrong.

nooo I didn't mean that blogger, I'm talking about a video blogger who was invited to Eucleadiaosd's fuck I can't spell their name, studio, this one which carb posted earlier.

Obviously somebody is a studio of the original developer can't be trusted unless he has a name of being a trustworthy dood.

Well my initial thoughts were that they're Australian.. And how I hated the guys' voices.

I'm from New Zealand ;_;

I'm sad now.

Allow me to butt in here, as I know not that much, but certainly a bit, about graphics programming, both the standard rasterization methods as well as all kinds of raytracing or raymarching approaches.

First up, yes, Euclideon definitely have a neat Voxel - or "Atom", as they like to call it - engine. But that is not something new, nor something particularily amazing. Moreover, they've been in the business of trying to get money from people by claiming they are doing revolutionary things for a lot of time. Here's a picture from ~2003~, where, unsuprisingly, their shit looked pretty much just like it looks now.

http://imgur.com/g0gXt

So, what do they have? They have a voxel engine, probably based on SVOs. It doesn't really matter. What's more important is what they do not have, and that is:

  1. Geometry that is not static, and deformable geometry, no matter how much they like to claim that they'll have that Real Soon Now. Note that this isn't something that's impossible, just kind of hard. Voxel deformation has been done, just need to implement it. It's not neccesarily very efficient, though, and considering that nearly all geometry must be deformable (think trees and grass blowing in the wind)...

  2. Geometry that is not the same thing over and over again. They like to pretend that this is only because they are not artists. I somewhat doubt that. See, with voxel engines, it is very easy to instantiate the same geometry over and over and over again, as long as it is on a halfway regular grid and always axis-aligned. If they could do arbitrary rotations of their geometry, they would do them, but they do not. More likely than not, it is a limitation of their engine.

  3. Interesting lighting and shadows. What they have is pretty much all-diffuse. I've no idea why. They also don't have any global illumination, which is a bit of a big deal. Again, I don't know why, there's techniques for extracting GI from voxelizations of scenes somewhat effectively.

More importantly, though, they are super dishonest all the way. They only admitted they were doing voxel rendering when pushed to do it, they never mention the downsides of their technique, and they try to make current-gen graphics MUCH worse than they are - those polygon graphics look like they took crysis, set all detail levels to bare minimum, and went with it. This is not how honest people go about business.

tl;dr: They are full of shit, please stop giving them attention, thanks.

edit: Oh, these are just the main complaints, anyways. There's a lot more that aren't worth going into in detail, such as "needs to run on more than just PCs", "needs to actually run at 60Hz to be useful", "needs to be able to render participating media", "needs to handle translucency and diffraction, or at least fake them convincingly", "needs to handle all kinds of aliasing", and probably tons of other things I am not thinking of right now.

To illustrate that last point a bit more, here's a demonstration of the Unreal Engine 3 rendering a few environments. Note how you cannot see the polygon outlines of things, most likely thanks to techniques such as continuously variable tesselation and displacement mapping and occlusion parallax mapping, and also thanks to just using a lot of polys to begin with. It's not like a modern GPU can't handle a few million (Even in HD resolutions, that's 5 to 10 polygons for every pixel on screen), if you batch them properly. Even more if you're smart about clipping and occlusion and whatnot. And after that, you can still occlusion parallax map for even more detail.

(Lighting is more of a problem, but that problem isn't solved at all by using voxels, it's still there just the same. In fact, lighting is the actual problem in graphics - polygon counts, eh. GPUs are good enough that those have kind of taken a backseat.)

You are on the old site. New site is here:

The site has been updated on the 24th December 2011. Please go there when you are finished with the archives.

  • 481,435 posts
  • 2,075 threads
  • 23,121 users