Straight out of ModDB's homeland Australia the software developer Euclideon has slowly been refining a new technology which they claim can support unlimited graphics. UNLIMITED PEOPLE! By droping polygons for a brand new system involving atoms (similar to real life no doubt) you apparently can create models with unlimited detail which will not have any impact on the game itself.
Don't believe me? I wouldn't think so but have a look see at this informative video below.
I have never seen such detailed grains of sand before. While they have said they still have a lot of work to do on the software before developers can get their hands on it, it would be nice to see if modders or indie developers will be able to use it in the games they create, if it wont cost your first born child that is.
I just died of awesomeness. This is absolutely amazing! I can't wait to see this implemented in the future.
With standard comment technology your comment is flat and boring but because we are so awesome we made a comment converter that will make your comment a global news media event in and of itself. Tune in next year for the release of our comment converter converter that'll make your comment so outrageously popular and sexy you'll want to take it home and shag it.
You called my comment boring. *cries*
there there... it doesnt matter if your post is boring or not... the video sure as hell isnt :D
The only question I have is, Why haven't you seen anything "Moving" yet? all they have shown is static environments. both times. The theory would be cool if only they would show some animation and physics demonstrations.
Wasn't that exactly one of the contra arguments i stumbled upon last year? I mean besides the lighting stuff - which they say they have improved but then theres the thing again: It all seems static at the moment . . . But yeah it's nice to see that they are still working on it (and i do hope that they will be able to continue that, since i dont know of any other voxel-based project having games as a goal that are running today) . . . Its certainly something which will be interesting to watch once it is presented at a "playable" form but until then i guess we can just forget about it (or reserve a bit of capacity in our memory for it, so that we recognize it when it reappears) and let them do their thing with as less disturbances as possible (i really dont want to see flame-wars to rise by people getting their hopes up and be rather displeased by not seeing an update after one or two weeks starting to say that they are all false and what the hell know i)
But yeah its nice to see, that they are still working on it, nearly forgot about them (although i have to admit that i cannot really see that much new material, they showed those pyramids already last year so . . . meh whatever)
oh and was it only me who was reminded about minecraft, looking at these square-based objects that build the ground of the techdemoisland?
They've shown animation before, it was a humming bird.
It's pretty old though, and it was a bit dodgy. I'm sure it's been improved upon tho.
The thing I'm more concerned about is being able to see individual voxels at just the *wrong* distances. Most of all I'm concerned about shaders and the like.
This animation that your referencing is done frame by frame, IE the point cloud data is stored for every frame of the animation so no ragdolls, physics or anything of the sort at least so far.
Ha, Everyone should go read Notches latest post on his blog. Pretty Interesting.
It is actually quite old. They called it Point-Clouds instead of polys. I read this in a article years ago and looks like they have that same pyramid structure in the video. Very cool though I hope they get this so we can adopt it asap.
We looked at this in Interlopers and concluded it's a useless system that's pointless to implement.
this is absolutly amazing! the tech we have here is terrifyingly incredible, any type of psysics placed with this tech will redefine the industry. I feel honored to be a Game Dev here in Australia.
Oh not these guys again.
An interesting link for people curious about this technology (please research it though and don't just interpret it as "unlimited detail", as it's not):
Voxelstein3d.sourceforge.net
This project is using essentially the same technique (voxels -- En.wikipedia.org)
Actually that Voxel technology is sooo old it was what was used to make the First Delta Force game. circa 1990's
Oh and to add to that, video cards do not understand voxels. The demos are software rendered (which _is_ somewhat impressive in itself). My understanding is that they do offload some work to the GPU in the form of GPGPU instructions, but it's by no means hardware rendered (GPGPU instructions have to traverse the PCI-E bus as they still use _some_ CPU and system memory, as opposed to video memory).
I am curious to see how they implement any sort of movement and/or dynamic lighting, as this would probably prove to be problematic with software rendering and voxels. From what I saw in the previews (this one and September 2010), the worlds are fully lit and all static (immobile) objects.
Some of us are also confused about "atoms per square cm" (as opposed to per cubic centimeter), the number of polygons rendered in a whole level - which most likely has nothing to do with the number of "atoms" they're rendering, and this term "floating point cloud" that they use.
</rant>
-Keith
soo what you are saying is that, what these guys are saying is in fact a lie? Or maybe not a lie but a hoax? A set up for some sorts?
Or is it a true technology, but the video cards that we use can't render it properly?
Sorry I think you misunderstand sorry.
It's not a hoax or a lie, it's a different way of doing things. This tech has been around for many years (since the DOS days), but not many groups (that I know of) have put much effort into developing voxel graphics engines. My statements are more to clarify that there are a lot of hurdles to overcome, and the statement "unlimited detail" is not quite correct.
Video cards render what you tell them to render, but they only really understand 3D polygons and 2D pixels (and text modes), so they have to calculate it in software then either convert it to 3D polygons or 2D pixel arrays to draw.
I may very well be wrong though, this is just my understanding of the technologies.
-K
yeah i assumed its just plain old voxel mapping.. This is not that amazing, considering a number of engines already use voxel mapping. Voxel's have soo many problems its not funny and there for commercial companies dont want to take up the task of using them, The unlimited detail they kept saying through out the whole video is just a way in which to confuse you. Really nothing special about this from what i saw all it is, is a heavily focused voxel mapping technique no different from any other engine which uses its own technique, The problems with using voxel would be allot worse
Check out the C4 engine it uses Voxels for its maps this is exactly what they are doing but they seem to be using Voxels for everything which is quite stupid. its OK if u want plain objects with no animation and grass that does not sway its not ideal for a real game engine but its ideal for a scenery that does not move or have any animation what so ever.
If no one focuses on it, it will never improve. What the hell do you think these guys are doing?
This could possibly push Intel to further develop their Larrabee Project (http://en.wikipedia.org/wiki/Larrabee_(microarchitecture)) which is a GPGPU that could really help with the kind of calculations needed for Voxel rendering, as well as solve the lighting issues with realtime raycasting.
Fun times :D
But as for the technology presented, it has currently no use unless you want an incredible looking static scene without any Ambient Occlusion, Lighting effects and movable objects...
But they are paving the way for new render engines that will be available in the upcoming years.
Looks cool.
the hardware companys will gladly outsupport your idea
Unfortunately I don't think it's quite that easy, but it would be awesome if there was a unified way of expressing voxels and graphics cards understood it!
There are many reasons that voxels are only software implemented though, and I guess the best we can do is hope that this at least pushes a big change in software rendering, if not hardware accelerated rendering and the quality of games.
Well to do that would mean waiting for faster graphics cards, or highly optimized and efficient programming. Because all those trillions of atoms still need to be rendered, how brilliant they may be. But yeah I do see potential in this.
But to get things straight. If they do make this software more reliable, would it be possible to create a simple game with it, using state of the art graphics cards like AMD's 6990 or a Nvidia Quadro 6000 disigners card?
You're on the right trac, but what I mean by "software rendered" is that it's the CPU doing most of the drawing. They only utilise the graphics card to do mathematical calculations using GPGPU (http://en.wikipedia.org/wiki/GPGPU), so faster graphics cards will help a bit, but more powerful CPU's will help more (so will an APU).
Another major issue may be memory, as voxel maps contain a LOT more data than polygon maps. There could be ways around this though, eg: curves stored as formulas, solid objects stored as point arrays, etc.
Oh and I was wrong (as was aihtdihk who mentioned it to me), they did say cubic cm/mm (not square).
okay, thank you for explaining that!
But if it's a CPU and Memory problem couldn't that be remedied with a quantum computer and advanced SSD? I heard D-Wave is going to sell commercial quantum computers designed for specific tasks? So maybe that future isn't that far away as we thought.
Wrong sort of processing. Oh and budget constraints ;) (quantum computers are _FAR_ from being affordable).
You'd also need massive amounts of RAID'ed SSD's :(
The only technical problem this always brings up is animation. They have yet to Answer it, I'm sure it's possible but it will be a resource heavy system, and John Carmack even tweeted that this is great but it's ahead of our time still, You won't be seeing this in our current generation of gaming.
if you remember they said they had a polygon to point converter, I'm sure this will play a part in something like this
Not quite, that's for importing pre-existing 3D objects.
Animation with polygons wouldn't be applicable to voxels, so a new technique will have to be introduced.
And the other issue is any sort of lighting besides 100% bright ;)
-K
What about some sort of hybrid-engine? Voxels for static geometry, polygons for animated objects.
I mean to be honest, the software has unlimited detail because its essentially a really powerful procedural search engine, so I don't think it should matter if the points move, it should be able to work the same
haha you guys jumped on this FAST xD I saw this earlier today
Plus slashdot article Sept 2010 I believe ;)
Henley does watch all the gaming news avidly though, and write about the good stuff! Keep up the good work.
I won't be impressed until they start showing elements which are actually needed for a modern game (dynamic lights, skinned/ animated meshes). Right now it seems like there making a lot of hot air because frankly polygons are looking a grand size lot nicer than their atoms because they have something called art direction.
Bingo
Free and Open Source Please ;D
Then how are they going to make their muneh?
The whole "we're going to show an interesting technology, not answer any questions, then disappear for a year" attitude isn't terribly reassuring. It looks neat, but I’ll be sitting skeptically from the sidelines until we get a better idea of it all.
In Sept 2010 they said "12 to 16 months"...
-Keith
(buried)
ooooooooooooooooold
I've been pondering the number of polygons mentioned in the video - 21,062,352,435,000 - and I'm not sure what to make of it. They are talking about polygons though, not 'atoms', so I can't tell how that would translate to the 'point cloud' form that they generate.
So I'm working with the only clear numbers they did give us - 64 'atoms' per cubic mm, and the level is 1 square km.
Even assuming that the level is only 1 atom thick (and that the atoms per cubic mm are arranged as a 4x4x4 cube), that gives 16 atoms per square mm of surface.
We have 1,000,000 (mm per km) * 1,000,000 square mm, times 16 atoms.
Am I on the right track? Cause if so, that's 16 trillion atoms.
Assuming only 1 byte per atom (i.e. 256 colour) that's ~16 terabytes.
Full (24-bit) colour? That's ~48 terabytes. For a simple coloured surface heightmap.
I hope they've worked out some very very clever ways to work around this, cause I'm not ready to buy a petabyte RAID just to install a game made with this tech.
Ow snap - referring to the math, my ankle
I realise my comment above looks very negative, so I'd just like to add:
I'm a huge fan of voxels, and I would love it if somebody (like Euclideon) were to make it feasible for use in games. As they clearly show, beautiful graphics are possible.
I'm not trying to be a wet blanket, but until they give some hint that they're working around the problems that have plagued voxels forever (amount of data, animated shapes, physics), all I'm seeing here is pretty pictures and potential disappointment.
So if you're reading this, Euclideans, please do your best, and prove me wrong!
I am somewhat being a wet blanket, but that's more my pessimistic mood today bought on by not being able to walk.
But yes, please do your best and prove us negative folks wrong :)
I think that your calculation is off because you take into account every cube in the map. All objects in the scene are hollow, probably using Sparse Voxel Octree which greatly reduces the size of the data storage, by conserving only the data of the objects that can be seen.
Or so I think, could be very wrong...
Yeah, my calculations are horribly over-simplified.
But I am talking about just a 2D surface, over the whole level - every square mm, but not every cubic mm.
The reality is probably better, and worse.
Better: I'm ignoring instancing and repeated sections of the terrain (which they clearly use).
Worse: I'm also ignoring the fact that every atom will have more properties than simply colour - because they're not stored at every point in a fixed array, they at least need to somehow record what their position is.
I love you, guys! Real tree finally!:) I hope this technology will continue.
+ a mix with cryengine and it will complement itself awesome :D