Unlimited detail graphics

[Niode]

Moore's law doesn't really apply to this. The increase in graphics quality is not directly linked to moore's law. Moore's law is simply; the amount of transistors that can be incorporated on an integrated circuit cheaply, doubles every 2 years. It's not even directly linked to an increase in performance either, since moore's law applies to the shrinking of ICs as well.

I don't think this is a fair statement. It's not quite this simple, but smaller transistors resolve their output faster than larger transistors. So, shrinking transistors lead directly to faster processors. Graphics cards are made of the same silicon that general-purpose processors are, and they benefit similarly. There are other hardware limitations that don't really benefit from Moore's Law, but even if these limitations are considered static, processing speed and power for general-purpose and graphics processors directly benefit from it.

Yes but moore's law is not directly responsible for an increase in graphics fidelity.

[AppleQueso]

For another thing... I have seen inflated claims about what a gaming system can do in the past. Do a youtube search for "Sega 32x tech demo." Read articles about the amazing claims Sega made about the Saturn. Before the PS2 came out, Sony was saying it could render millions of leaves on millions of individual trees, and for the PS3 they showed a really dubious video for Killzone 2 making the gameplay look like bloody Final Fantasy Spirits Within in real time. What I'm saying is, I'm aware of hype tactics.

Remember when there were claims of the PS2 being able to output Toy Story quality graphics.

I remember George Lucas claiming he was going to use the PS2 for special effects in his next film

[elvis]

I'm just suggesting maybe the improvements from one graphics card to the next are growing so subtle that they'll have to think outside the box. Personally I'm predicting some hybrid voxel/polygon engine

You need to realise that current generation video cards aren't "polygon engines" per se. They're just processors that can stream data (as opposed to the fetch/decode methods of general purpose CPUs) and are really good at floating point (and thus vector) maths.

Once you cement that in your mind, you'll realise that the current generation of video cards are perfectly suited to "voxel engines". Languages like CUDA and OpenCL are doing these sorts of things right now on current generation GPUs. Once again, search YouTube and see it in action on modern video cards.

Likewise, jump on Nvidia's website, and chuck the word "voxel" into their search bar. You'll quickly see they're already all over this.

Here's a video by Nvidia straight off their site based on research into efficient sparse voxel octrees (with demo implementation in CUDA):
http://www.nvidia.com/object/nvidia_res ... b_018.html
http://www.tml.tkk.fi/~samuli/publicati ... _video.avi

Clever readers will pay attention to the lighting in the scene, and the comments at the beginning of the video stating that there are no light maps. This is 30FPS raytracing.

Speaking for myself, talking about merely storing and rendering voxels is rather boring. The next evolution in 3D won't be how to draw voxels, but rather how to calculate interactions between them. Giving voxels their own individual properties in a physics engine is far more interesting, as that will be a massive jump forward for games. Once every voxel in an environment can interact with every other voxel around it governed by physical laws of the game engine, then we'll get a whole bunch closer to something interesting (and much closer to how things work in the real world).

Bring on artificial neural network voxels!

[Anayo]

Hey, Elvis, thanks for linking to that article on NVidia. I'm looking through it at the moment and it's very interesting.

[GSZX1337]

For another thing... I have seen inflated claims about what a gaming system can do in the past. Do a youtube search for "Sega 32x tech demo." Read articles about the amazing claims Sega made about the Saturn. Before the PS2 came out, Sony was saying it could render millions of leaves on millions of individual trees, and for the PS3 they showed a really dubious video for Killzone 2 making the gameplay look like bloody Final Fantasy Spirits Within in real time. What I'm saying is, I'm aware of hype tactics.

Remember when there were claims of the PS2 being able to output Toy Story quality graphics.

I remember George Lucas claiming he was going to use the PS2 for special effects in his next film

I remember when terrorists were buying PS2s for their missiles. Oh wait...

[aaron]

anyone have any thoughts on this? basically they've attempted to create software that converts polygons to point-based atoms, allowing greater detail in 3D graphics. it looks pretty impressive but i'm wondering how it will translate to the current gen (if at all) or what it means for the next gen.

[MrPopo]

As I understand it the current implementation uses a lot of shortcuts to avoid actually modeling every individual atom. Additionally, voxels have a lot of algorithmic challenges when it comes to animation, so while they can easily make a high res static shot it's much more difficult to get high res in motion in real time.

[isiolia]

It's not a new idea per se, and has been commented on by people with more knowledge on the subject than me.

Kotaku has had follow-up posts with feedback/references here and here.

IMO, the assessment that they're looking for funding is probably right. That said, the tech as a whole does have some real potential, but you'd probably have to design hardware around that moreso than polygons (which seems to be what John Carmack implied when he commented on it not being feasible on current-gen consoles). Even so, it wouldn't be from that specific company, as there are numerous implementations around.

[J T]

It's really cool looking stuff and I think it's exciting. I am wanting to see an animated character though because if it can only do really realistic static shots, then it's pretty limited for videogame usage. Though I do wonder if you could overlay polygonal animated characters over unilmited detail backgrounds? That might be a nice work around if it's possible. Even if they are just pulling for funding, I think what they have shown already is interesting enough to warrant funding (which they've already gotten by the way).

I think people are often too quick to be naysayers and shout "impossible". Admittedly, Notch and Carmack know way more than I do though, so there may be some breaking point in Euclidian's logic that I'm just not picking up, but it doesn't seem like they've released enough info for anyone to be too sure of that.

[flamepanther]

Even if it can only do static scenery, I'm sure it could be used in combination with skeletally animated polygons to populate it.