Euclideon Island Demo 2011 - Games

notch is talking bullshit, he "forgets" like everything they presented.

They can make BILLIONS selling this to companies, aswell as change the game industry in an AMAZING way. I wish to see this in future games. This could be the breakthrough to reality look-alikes games.

[QUOTE=Str4fe;31488570]Just because notch said so doesnt mean "yep its bullshit notch told me" He could be wrong, He is only one indie game developer. Not a God. [/QUOTE] Uh, no, it's bullshit because the reasons Notch gave are accurate. The biggest problem is memory, the second is animation. Voxel rendering (which this is, even if they try to hide it by calling them 'atoms') has been around a long, long time- these guys are trying to dress it up as something new, and present it in a manner that glosses over the HUGE problems it has.

I don't know what to think now. I think a demo is needed before I'm convinced. The fact they've been going at this for 10 years with essentially the exact same thing to show for it is now making me very skeptical. I don't see why it would be so hard to package it up and release it if it really is as they say. They should do this.

[QUOTE=Karmah;31491475]Hey guys, I don't understand how it works so I don't believe it, despite not even understanding how a traditional game engine works. :downs:[/QUOTE] Hey guys, I'm a game designer and understanding how stuff like this works is my job, and it's bullshit.

[QUOTE=catbarf;31509530]Uh, no, it's bullshit because the reasons Notch gave are accurate. The biggest problem is memory, the second is animation. Voxel rendering (which this is, even if they try to hide it by calling them 'atoms') has been around a long, long time- these guys are trying to dress it up as something new, and present it in a manner that glosses over the HUGE problems it has.[/QUOTE]Let me correct you. Memory is not a problem, If you even bothered to read any good posts in the thread you would know that they are using a type of culling (the name of which I forget, it's not exactly culling but close to it) that enables only the voxels you can see to be rendered. And they already have animation.

[QUOTE=Meatpuppet;31509697]Let me correct you. Memory is not a problem, If you even bothered to read any good posts in the thread you would know that they are using a type of culling (the name of which I forget, it's not exactly culling but close to it) that enables only the voxels you can see to be rendered. And they already have animation.[/QUOTE] The voxels that aren't on screen still need to be stored in memory or the renderer has absolutely no way of knowing where they are and whether or not they should be rendered. Every single voxel needs to have its own memory allocated so that it can be kept track of. This can be aided somewhat by using pre-made objects and structures that can keep track of their constituent voxels' aggregate position, only actually assigning them memory when visible and saving the fine calculation for edges of the screen, but it's still a huge memory sink- not to mention having the voxels not actually being stored in memory becomes a huge problem if you're doing any sort of physics processing on off-screen objects. As for animation, just because it's been done doesn't make it practical. Notice that the render test that's been linked to was at ~35 FPS with just one character moving on screen. The animation is extremely CPU-intensive and not useful for realtime rendering of dynamic environments.

[QUOTE=Meatpuppet;31509697]Let me correct you. Memory is not a problem, If you even bothered to read any good posts in the thread you would know that they are using a type of culling (the name of which I forget, it's not exactly culling but close to it) that enables only the voxels you can see to be rendered. And they already have animation.[/QUOTE] To add onto what catbarf is saying, culling only cuts down things you have to draw, not always what's in memory. Plus, sorting through what you can and can't see over trillions and trillions of voxels would take longer than just rendering every single one. There are some streaming methods to pull things in and out of memory in addition to culling it out, but with the massive amount of unique data you'd have to pull from trillions of voxels, streaming it in and out at real-time speeds would be completely impossible with current-gen tech.

If the geometry is unique, it has to be stored, this goes for polygons and voxels. The only way to make additional detail is by adding it procedurally at run-time. We do this already with polygons. It's the reason why tesselation is such a big deal, it creates additional detail in the reference mesh. The only content actually being stored is the reference mesh.

[QUOTE=OpethRockr55;31510396]To add onto what catbarf is saying, culling only cuts down things you have to draw, not always what's in memory. Plus, sorting through what you can and can't see over trillions and trillions of voxels would take longer than just rendering every single one. There are some streaming methods to pull things in and out of memory in addition to culling it out, but with the massive amount of unique data you'd have to pull from trillions of voxels, streaming it in and out at real-time speeds would be completely impossible with current-gen tech.[/QUOTE] Exactly. The fact that they have learned how to optimize a rendering engine doesn't suddenly make it practical. I love how I'm getting dumbspammed by FPers who think that because I'm agreeing with Notch I have no idea what I'm talking about. He's not the ultimate authority on the subject but the general points of his argument are valid- this just isn't a practical technology due to limitations of voxels themselves. You can talk to any games-industry graphics programmer and they'll tell you the same thing. [QUOTE=Hostel;31510786]If the geometry is unique, it has to be stored, this goes for polygons and voxels. The only way to make additional detail is by adding it procedurally at run-time. We do this already with polygons. It's the reason why tesselation is such a big deal, it creates additional detail in the reference mesh. The only content actually being stored is the reference mesh.[/QUOTE] The problem with voxels is that you still need all the voxels present to do things like collision checking and animation- with polygons, the extra detail is added on to existing shapes, and the original polygon structure can be used for calculations even with culling. If you start taking away voxels from a voxel object to cut down on render time, you run into problems.

[QUOTE=Milkyway M16;31495821]The way it works is by displaying only the information you actually need. This means that it searches for only the information that you are currently looking at and processes only that data. The rest is left untouched until it is needed. That's what makes this system work so well. It is a basically a search engine for point cloud data.[/QUOTE] If you turn around super fast does it have trouble loading fast enough/lags horrendously?

Why do they have to show the games running on lowest possible settings to say that polygons don't allow for detail? It's total bullshit. I have played those games and when I am up close like that it isn't that low poly as they as they are trying to show, infact they are probably config modding to make it look terrible.

[QUOTE=Clementine;31511556]If you turn around super fast does it have trouble loading fast enough/lags horrendously?[/QUOTE] No because the amount of polygons loaded at any time is a lot less than most games today. Since the amount of data onscreen at any given time is equal to your screen resolution, (lets say, 1280 x 1024) that means that at any given time there are exactly 1,310,720 polygons on your screen. Games today use millions of polygons more in their levels. Current processors have no problem handling polygon counts like these which is why this system runs so well in their engine. This is my understanding of how the system works at least.

[QUOTE=catbarf;31511513]Exactly. The fact that they have learned how to optimize a rendering engine doesn't suddenly make it practical. I love how I'm getting dumbspammed by FPers who think that because I'm agreeing with Notch I have no idea what I'm talking about. He's not the ultimate authority on the subject but the general points of his argument are valid- this just isn't a practical technology due to limitations of voxels themselves. You can talk to any games-industry graphics programmer and they'll tell you the same thing. The problem with voxels is that you still need all the voxels present to do things like collision checking and animation- with polygons, the extra detail is added on to existing shapes, and the original polygon structure can be used for calculations even with culling. If you start taking away voxels from a voxel object to cut down on render time, you run into problems.[/QUOTE] What makes notch a bit wrong is that I doubt they would be storing each voxel as 1 byte, that's the first mistake he made, made an assumption and continued with it to make a point, without backstracking and saying it's just theory based on his one assumption. I mean that's like saying JPG is impossible, because you have to store every pixel per byte. You don't need all voxels present to do collision checking, just a few of them, and you can always group them. I mean really, did you really said that every game has perfect collision system with polygons? No clipping ever occurs? This is unlimited detail tech, which is like tessellation on steroids, it doesn't mean it will use atoms for every occasion, such as physics, animation or everything else. As some people said, it could be used for terrain that is static. I mean really, why people suddenly think if they said it's made of atoms, then it's a perfect simulation of atoms and everything associated with atoms.

this also isn't voxels

[QUOTE=Milkyway M16;31517470]No because the amount of polygons loaded at any time is a lot less than most games today. Since the amount of data onscreen at any given time is equal to your screen resolution, (lets say, 1280 x 1024) that means that at any given time there are exactly 1,310,720 polygons on your screen. Games today use millions of polygons more in their levels. Current processors have no problem handling polygon counts like these which is why this system runs so well in their engine. This is my understanding of how the system works at least.[/QUOTE] It still needs to calculate which ones to display, which means that all the ones within the field of view need to be loaded and processed. [QUOTE=maqzek;31518152]What makes notch a bit wrong is that I doubt they would be storing each voxel as 1 byte, that's the first mistake he made, made an assumption and continued with it to make a point, without backstracking and saying it's just theory based on his one assumption. I mean that's like saying JPG is impossible, because you have to store every pixel per byte. You don't need all voxels present to do collision checking, just a few of them, and you can always group them. I mean really, did you really said that every game has perfect collision system with polygons? No clipping ever occurs? This is unlimited detail tech, which is like tessellation on steroids, it doesn't mean it will use atoms for every occasion, such as physics, animation or everything else. As some people said, it could be used for terrain that is static. I mean really, why people suddenly think if they said it's made of atoms, then it's a perfect simulation of atoms and everything associated with atoms.[/QUOTE] All true, but the memory costs are still horrendous, and the only way it's actually renderable in the demos is by having large areas of identical, repeated clusters of voxels. [QUOTE=Meatpuppet;31518160]this also isn't voxels[/QUOTE] Yes it is, regardless of what buzzword they use.

[QUOTE=catbarf;31518331]All true, but the memory costs are still horrendous, and the only way it's actually renderable in the demos is by having large areas of identical, repeated clusters of voxels.[/QUOTE] Yea but game devs obviously aren't going to make a game as detailed as that tech demo there would be no use. Besides Not to mention its not like games already don't have the same tree a million times.

[QUOTE=catbarf;31518331]It still needs to calculate which ones to display, which means that all the ones within the field of view need to be loaded and processed. All true, but the memory costs are still horrendous, and the only way it's actually renderable in the demos is by having large areas of identical, repeated clusters of voxels. Yes it is, regardless of what buzzword they use.[/QUOTE] They can always use procedurally generated content. I mean, look at Outerra engine, in theory that would consume shitload of memory as well, but it doesn't and runs smooth on 8800GTX. There's always a way around stuff. Not saying this guy's works is the best way or perfect, but I just don't want to discard it incase it actually might be something cool. I'm all for progress. Even if this is half a scam, the amount of publicity and attention it's getting again is a good thing in my book. People will keep asking if it's really possible to do and if it is, in any way shape or form, then why is it not done yet, etc. A bit of minecraft efffect I hope.

[QUOTE=imasillypiggy;31518867]Yea but game devs obviously aren't going to make a game as detailed as that tech demo there would be no use. Besides Not to mention its not like games already don't have the same tree a million times.[/QUOTE] But it's not really 'unlimited detail' if it's 'a little bit of detail copy-pasted a million times'. [QUOTE=maqzek;31520478]They can always use procedurally generated content. I mean, look at Outerra engine, in theory that would consume shitload of memory as well, but it doesn't and runs smooth on 8800GTX. There's always a way around stuff. Not saying this guy's works is the best way or perfect, but I just don't want to discard it incase it actually might be something cool. I'm all for progress. Even if this is half a scam, the amount of publicity and attention it's getting again is a good thing in my book. People will keep asking if it's really possible to do and if it is, in any way shape or form, then why is it not done yet, etc. A bit of minecraft efffect I hope.[/QUOTE] I dunno. This guy's been at it for close to ten years now with little to show beyond extremely tightly controlled tech demos that deliberately hide all the problems he's apparently no closer to solving. In a lot of ways voxels seem like a step backwards, when technologies like bump mapping and tessellation can achieve similar levels of detail using conventional rendering and are much more resource-friendly. I'm all for experimental technologies- raytracing, for example, isn't quite good enough to be used in realtime but shows incredible promise. It's just that this voxel rendering has issues that mean it's not going to be practical within even the foreseeable future.

[QUOTE=catbarf;31526827]But it's not really 'unlimited detail' if it's 'a little bit of detail copy-pasted a million times'. I dunno. This guy's been at it for close to ten years now with little to show beyond extremely tightly controlled tech demos that deliberately hide all the problems he's apparently no closer to solving. In a lot of ways voxels seem like a step backwards, when technologies like bump mapping and tessellation can achieve similar levels of detail using conventional rendering and are much more resource-friendly. I'm all for experimental technologies- raytracing, for example, isn't quite good enough to be used in realtime but shows incredible promise. It's just that this voxel rendering has issues that mean it's not going to be practical within even the foreseeable future.[/QUOTE] So you're saying that real life is not detailed? I mean, everything is made out of atoms and atoms of the same kind are exactly the same.

Holy duck tits, imagine the possibilities....

[QUOTE=joost1120;31526942]So you're saying that real life is not detailed? I mean, everything is made out of atoms and atoms of the same kind are exactly the same.[/QUOTE] The problem is not that the atoms themselves are the same, it's that there are exactly two options with voxel rendering: 1. Represent every single voxel in memory or 2. Create groups of voxels in memory and copy-paste the group by reference Those are the only two possibilities. Because there isn't nearly enough memory on any machine to represent every voxel individually, all the demos he's shown have had large numbers of duplicated objects. That's not at all practical for any game, and it's the number one reason why this technology is extremely limited in what it can do. It's not 'unlimited detail' if you're limited to just small objects of detail, and forced to clone them identically.

[QUOTE=catbarf;31531753]That's not at all practical for any game,[/QUOTE] Really? Coz games tend to re-use the same object a lot as is...

[QUOTE=Ranik;31531823]Really? Coz games tend to re-use the same object a lot as is...[/QUOTE] This is beyond just re-using a crate here and there. A crate made out of polygons is six squares plus some bump mapping. A crate made out of voxels is thousands upon thousands- granted, there are compression techniques that could reduce it, but with things like landscapes it's not so simple. What I'm getting at is that the only way these tech demos work at all is by having only a very few unique objects, and just repeating them over and over and over again, because voxel objects take up so much memory that you just can't afford to have many of them. In the ten years it will take for the memory to be even remotely practical, polygon graphics will have progressed to the point where the tech demos boasting 'unlimited detail' aren't impressive anymore. And that's assuming the guy actually does anything in ten years, because from what I can tell he hasn't done much in the last ten.

[QUOTE=catbarf;31526827]But it's not really 'unlimited detail' if it's 'a little bit of detail copy-pasted a million times'.[/QUOTE] But the reason it was copy pasted is because it was so detailed. I bet it wouldn't have to be if we didn't give every grain of dirt its own atoms. I like how there private zoglow is rating every post that agrees with him winner and every post that disagrees with boxes.

[QUOTE=catbarf;31532609]This is beyond just re-using a crate here and there. A crate made out of polygons is six squares plus some bump mapping. A crate made out of voxels is thousands upon thousands- granted, there are compression techniques that could reduce it, but with things like landscapes it's not so simple. What I'm getting at is that the only way these tech demos work at all is by having only a very few unique objects, and just repeating them over and over and over again, because voxel objects take up so much memory that you just can't afford to have many of them. In the ten years it will take for the memory to be even remotely practical, polygon graphics will have progressed to the point where the tech demos boasting 'unlimited detail' aren't impressive anymore. And that's assuming the guy actually does anything in ten years, because from what I can tell he hasn't done much in the last ten.[/QUOTE] I'm not talking about crates, games re-use A LOT of models. A lot more than you think probably. Not even just modern games either, try going thru Half-Life 2 again sometime and you'll notice they reused entire buildings/rooms/vehicles/people/etc multiple times. I'm not sure why I said they re-use a lot of the same object and you pick the smallest most dumbest object that wouldn't matter no matter what. Games like Bullet-storm are even worse. Not only do they re-use the same objects (somewhat creatively, granted) in that game but they cut corners in the way every other polygonal game does except they do it VERY obviously. Pop in, 2D backgrounds, 2D skybox, nothing on one side of an object, etc. And if you're gonna go "but those aren't open world" you could always look at how many models are re-used in Morrowind/Oblivion/World of Warcraft (it's a shit ton) My point is, yes, they are cutting corners with this engine - game developers cut corners with polygonal engines as well. Consoles aren't strong enough to play a game at even 30 FPS without those cut corners, and most PCs playing games out there probably aren't even mid ranged PCs so they can't play games like that either.

[QUOTE=catbarf;31532609]This is beyond just re-using a crate here and there. A crate made out of polygons is six squares plus some bump mapping. A crate made out of voxels is thousands upon thousands- granted, there are compression techniques that could reduce it, but with things like landscapes it's not so simple. What I'm getting at is that the only way these tech demos work at all is by having only a very few unique objects, and just repeating them over and over and over again, because voxel objects take up so much memory that you just can't afford to have many of them. In the ten years it will take for the memory to be even remotely practical, polygon graphics will have progressed to the point where the tech demos boasting 'unlimited detail' aren't impressive anymore. And that's assuming the guy actually does anything in ten years, because from what I can tell he hasn't done much in the last ten.[/QUOTE] But you don't need to store each voxel or group of voxels, you can just use LoD to extrapolate from vector mesh or something. Unlimited detail doesn't imply realistic, it just means you can zoom in to atoms and it wouldn't be blocky. Especially for terrain, vector type mesh could work pretty fine. Pretty sure someone else could come up with much better compression or storing method if this tech is really needed.