EPTOn August 13 2011 05:54 Urnhardt wrote:
its fairly simple to understand. to run a world like they show in that video, with 100% pure atoms or voxel makeup, there is no computer in the world that can stream data that fast, especially when any AI or animation is involved.
its fairly simple to understand. to run a world like they show in that video, with 100% pure atoms or voxel makeup, there is no computer in the world that can stream data that fast, especially when any AI or animation is involved.
That's pretty much the point of Sparse Voxel Octrees, you don't need to load all voxels to render your model. It needs to load only a fraction of the data, a lower resolution representation, if the object is far away from the camera and needs to be rendered with less resolution. Then it stream all the details from disk when you get closer. This kind of progressive rendering is inherent in the data structure used.
Storage capacity is more of a concern than bandwidth, but for opaque models you only need to store its surface, and even there you can trade-off quality for storage. I suspect it is possible to use some kind of wavelet based / multiresolution compression to preserve both rough and fine details.
Skeletal animation is feasible even with static octrees, here's the Youtube channel of the guy who implemented it in his Bachelor's thesis. Doesn't seem to involve much overhead.
The Atomontage engine uses particle based physics, so that is definitely possible as well, but nothing is stopping you from using a rough polygonal model instead of the voxel model for physical calculations.
