Unlimited Detail Real-Time Rendering Technology Preview 2011

16 Responses to “Unlimited Detail Real-Time Rendering Technology Preview 2011”

1. atikkur says:

   2011/08/02 at 16:50

   when it will come to commercial use?
2. fellix says:

   2011/08/02 at 17:30

   I hope never, at least not in its current form.

   It’s good only for static environments — animation, skinning an morphing with this tech will simply hog the current selection of hardware… unless you have a mini render-farm at your disposal. Similar to other complex data structures, like octree voxels and ray-tracing, traversing such a dynamic field of millions points is very expensive and it scales really bad with the screen resolution on top of it.
3. Reavenk says:

   2011/08/02 at 18:13

   Looks significantly better than it did a year ago, the edges actually look continuous now – and the material (or at least colors), lighting, and demo assets are way better.

   Though, my main question is, “what’s its advantages over an out of core SVO?”
4. Psolord says:

   2011/08/02 at 18:22

   They say in the video that they are getting around 20fps on software rendering, but they don’t mention the resolution.

   Can anyone guess how would it fair on recent graphics cards and high resolutions?
5. zorg says:

   2011/08/02 at 20:17

   Psolord: The high resolutions would be ok.
   The real problem what fellix said. The demo is a static scene, in a game we need dynamic scene.
6. Leith Bade says:

   2011/08/02 at 23:39

   Also games companies are not going to be interested in this until the next generation of consoles arrives and they stop targeting Xbox 360 and PS3.
7. Stefan says:

   2011/08/02 at 23:54

   Developer of “Minecraft” claims “It’s a scam!”

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

   2011/08/03 at 00:56

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

   Anyway, they’ve got some better textures for the video in the last year, lol.
9. Romain says:

   2011/08/03 at 08:03

   Scam… I do not know.
   Well, if it is true that they are looking for funding, it is somehow a scam as investor would believe that this tech is going to be used right now. But yes, it has many drawback (lack of dynamic stuff as other are pointing out).
   Definitely out of video game. May worth something in other “less” realtime industry. But I do not buy the argument made by the creator of Minecraft that the “world is too small”. It would be the same as saying that in your game you don’t have object but ONE complete world with all the vertex at once. No : people are using instances of objects, patches, etc… And I do not see why instances of trees could not use the same definition and the same tree be at multiple place with multiple orientation with a well crafted tech.
10. 3Dude says:

    2011/08/03 at 14:31

    There’s no technical progress since the first videos:
    No animation (except the 8bit look a like bird) No GPU port (Sparse Voxel Octrees work well on GPU, see http://www.icare3d.org/)
    No illumination except single directional.

    And still no explantation how to fit unlimited stuff in limited memory… compression? recursion? Nothing of that is new.

    However, if Bell will change the world, i’ve another brilliant idea: Since all CPU time is eaten up by rendering static worlds… Let’s create an engine that can do game logic, physics and AI on GPU
11. efikkan says:

    2011/08/03 at 22:41

    The greatest limitation of graphics in modern games is graphics memory and effective LoD. A powerful GPU like my GTX 580 has no problems rendering 4-8M polys in full HD at 30+ FPS with texture, so if someone make an effective way to just render the required polygons and only store these in memory, you can achieve “infinite detal” when polys > pixels by a factor.

    It seems like these guys has found a way to solve the LoD problem, and that’s good, but you can see in their content that it’s highly repetitive, so they have not found a way to only cache the local data.

    But anyway, for surfaces polygons will still be just as quick or quicker to depth test and so on than just points. The fact that we are using polygons instead of something else is not the limiting factor in todays games.
12. Joseph says:

    2011/08/04 at 13:35

    @ 3Dude

    It only loads what is needed, I.E if you go close to something then it will stop rendering things you can’t see and use all it’s resources on one item.

    So It’s not unlimited in the sense that everything is there all loaded up at once, it will be loaded on the fly.

    It’s drive space that this will will be hard not V-ram.
    That’s my understanding of it anyway.

    ( If something blocks another thing the thing behind will not be rendered either.)
13. DrBalthar says:

    2011/08/04 at 14:31

    Oh not these suckers again. Point-based rendering is nothing new guys. It has been shown that with instances you can render large data sets already in 2001. Even animated datasets a few years later. They just making fools out of themselves.
14. 3Dude says:

    2011/08/04 at 22:15

    @ Joseph

    You’re for sure right, but all you describe is in use on actual and outdatet game engines. LOD, Occlusion Culling, Hidden Surface Removal, Streaming, Megatextures…

    I am impressed by the details if coming very close but a serious man wold do following things when showing his work:

    1. Zoom close enough to show when the voxels become visible instead of claiming “unlimitied” details.
    2. Concentrate on his own work instead of showing so much other “ugly” games we all know.
    3. Tell us about the memory requirements and the resolution.

    From the first video seen most people assume something like octree with shared nodes, pretty visible because of instancing and the blocky, rotation-less look. This is why it looks bad – it’s not “bad artwork” – the artwork is pretty good, isn’t it?

    The next GPU Generation will be fast enough to use tesselation everywhere, and much more important: realtime global illumination. All that in fully dynamic scenes.

    How will Bell compete against that? He needed a year and 2 millions just to create… polygon-converters???
15. fofo says:

    2011/08/09 at 04:01

    Voxels are a non-sense having DX11′s tessellation. Period.
16. 3Dude says:

    2011/08/11 at 13:45

    There’s an interview:

    http://vimeo.com/27522131

    Animation is shown, skeletal + sprites, but no skinning.
    However, one could use polymodels for that purpose.
    The question is: Could all leaves from the trees have wind animation?

    No words about octrees or instancing, but he states that he’s not using rays.I’m getting an idea…