Huatim
Insider
Euclideon
- Thread starter Huatim
- Start date
MrIdontKnow
Insider
Yeah I watched about 1 year ago, it's pulled quite a few questions. I don't know much about the technology, but if it were possible, it would be great.
I think you'll receive a more detailed post from Rob or one of the Devs, since they know more than me, I just play games, and post of Forums.
I think you'll receive a more detailed post from Rob or one of the Devs, since they know more than me, I just play games, and post of Forums.
Huatim
Insider
COme one devs
its euclideon viable?
calithlin
Insider
Currently, previous attempts at voxel based technology that they are describing is not very easily animatable or made dynamic. Obviously the people at Euclideon are attempting to change that, but the limitations with that technology tend to make it excellent for static, unchanging items/environments. You'll probably notice that there's actually very few objects on screen, and they're all duplicated billions of times in order to simulate an unlimited number of imported models. However, the system still has the same limitations of normal memory technology in that the less items loaded on screen mean lower memory/faster processing time. Their workaround is to have a few items, but instance them tons of times in a way that takes no more memory or processing overhead (roughly).
However as I understand it, their items are generally stuck where they are and cannot deform neatly or cleanly without above average processing power behind such movement.
What they're doing to be able to render all of it is sparse voxel octree partitioning (if they are smart), which is a science that is very helpful in dividing up chunks of data via a search algorithm, and in conjunction with raycasting techniques renders to screen on a per pixel basis. The idea is if you zoom out far enough you can see the entire Earth represented as a single pixel.
However as I understand it, their items are generally stuck where they are and cannot deform neatly or cleanly without above average processing power behind such movement.
What they're doing to be able to render all of it is sparse voxel octree partitioning (if they are smart), which is a science that is very helpful in dividing up chunks of data via a search algorithm, and in conjunction with raycasting techniques renders to screen on a per pixel basis. The idea is if you zoom out far enough you can see the entire Earth represented as a single pixel.
Rob
Moderator
Yes, that was my conclusion also.
Sounds both great and unfeasible, for different reasons. There's not much that can be said about it at this stage, given that they're not disclosing what they're actually doing.
Voxelisation has been used for many decades - there's nothing special about it, it's used for various purposes, particularly in multi-resolution shape analysis, and for array indexing for efficient fast search algorithms. Ultimately, in this context, it's just a different representation of the same data. If you tried to create the same level of detail with polygons, graphics cards simply wouldn't be able to cope. Basically, it takes graphics processing away from being a bottleneck when wanting to get ridiculous detail. Sounds great, doesn't it?!
The problem is that you've still got to:
It would be interesting to know what they're doing to overcome these problems. Surely they will have to have a limited number of unique "structures", with game worlds being defined by a set "arrangement" of such structure objects... note how in this context "structures" and "arrangement" feels a little bit like a generalisation of "textures" and "polygons"...
Maybe fast large storage volumes could form part of their game-plan, when they're talking about the technology being potentially feasible in the near future...
As for animations... that would require a bit more thought. I certainly wouldn't say that it's impossible... just that the solution isn't immediately obvious.
Sounds both great and unfeasible, for different reasons. There's not much that can be said about it at this stage, given that they're not disclosing what they're actually doing.
Voxelisation has been used for many decades - there's nothing special about it, it's used for various purposes, particularly in multi-resolution shape analysis, and for array indexing for efficient fast search algorithms. Ultimately, in this context, it's just a different representation of the same data. If you tried to create the same level of detail with polygons, graphics cards simply wouldn't be able to cope. Basically, it takes graphics processing away from being a bottleneck when wanting to get ridiculous detail. Sounds great, doesn't it?!
The problem is that you've still got to:
- draw it - imagine the amount of time artists would have to spend to map out a world in that level of detail! The only way you could get around it is by having an awful lot of re-used stuff.
- store it - rather than processing, I can imagine that storage would become the real bottleneck! Imagine having to store all of those data particles! For a whole game world, that sounds completely unfeasible. Again, the only way to do it would be to be reusing stuff over and over again. That's probably why their demos look so repetitive, being created from only few unique objects.
It would be interesting to know what they're doing to overcome these problems. Surely they will have to have a limited number of unique "structures", with game worlds being defined by a set "arrangement" of such structure objects... note how in this context "structures" and "arrangement" feels a little bit like a generalisation of "textures" and "polygons"...
Maybe fast large storage volumes could form part of their game-plan, when they're talking about the technology being potentially feasible in the near future...
As for animations... that would require a bit more thought. I certainly wouldn't say that it's impossible... just that the solution isn't immediately obvious.
SergeDavid
Insider
Seeing that before the biggest thing about it is that most of the stuff is reused (and I'm not just talking about rocks). It looks nice at close range however when you zoom out a bit you'll start seeing the tiling. Personally I don't care how detailed a rock is if the map can't be filled with unique geometry.
