posted on Aug, 3 2011 @ 03:24 AM
Originally posted by BIGPoJo
reply to post by Akasirus
I think most people can grasp the physical limitations of rendering a crap ton of voxels. What if the developers found a way to only track and render
the voxels that you are looking at? Also, at a great distance away you would need less voxels, just scale them up or down based on your distance.
Maybe they found a shortcut to reduce the hardware requirements. Maybe they only track a limited number of voxels and predict the rest based on
previously rendered voxels?
I personally have no experience in 3d rendering and voxels are new to me.
That's the thing though, there aren't any shortcuts when calculating the visible faces or voxels of an object. Everything the camera can see has to
be calculated and rendered. Even if you ignore all the data points that wouldn't be immediately visible, this would still increase the number of
faces drastically. The level of detail is directly proportionate to how many points on the objects surface you use to calculate the position and
lighting. What you are basically asking is if they've found a way to do the math faster, which is very improbable. Using fewer and fewer points, to
where it's equivalent in overhead to the same scene rendered as polygons would provide no noticeable benefits.
And yes, when an object is further away you would not need to compute each point. With polygon rendering this is known as tesselation.
If they only track some voxels, and predict the rest, logistically that's exactly the same as polygon rendering, where it calculates the vertices to
render the faces. So then what is the benefit to this new rendering method, if it just uses a subset of voxels and treats them as vertices?
The other problem with voxels is that once you introduce physics into the system, you can't just calculate any voxels that you can see. You don't
have to render them, but every voxel every other voxel is touching would play a part in how that objects behaves when it's interacted with.
Now don't get me wrong, I am not ruling it out as impossible, something like that would be entirely possible, and potentially incredibly detailed.
All I'm disputing is that they've found a way to render 1000s of times the detail using less resources. Polygon rendering has gotten incredibly
efficient, and the reason it's used is because it is a fast approximation a 3D object. Voxel, atoms, or other volumetric 3D objects with smaller
building blocks would almost by definition be more demanding on hardware. Voxel rendering would require amazing processing powers to look semi decent,
it isn't until levels well beyond our current consumer tech that it would begin to surpass polygon rendering. There are many shortcuts you could
take, sure, but each of those shortcuts is going to lower the detail further.