Their demo is running on a single-core laptop processor.
In addition to that, their search algorithm makes it so only the points facing you are visible. No other points are rendered, and that is why they can do this.... making the number of points per scene consistent and predictable depending on LoD settings that change point complexity as an object gets further away.
I think people are thinking too much in terms of traditional voxels where a model is filled with points.... and unlimited detail doesn't work like that. The technology might be based on voxels, but then goes to solve many issues with voxels and real time rendering.
It isn't that complex when you look at the actual tech.
Just because the point isn't viewed on screen, that doesn't mean it's gone or never will be displayed. All of those points have to be stored some place, if its not sitting in the CPU then its in RAM or on a disc or on a hard drive. Even if you come up with quick way to sort all of those points and only display a handful you still need to have a giant storage area to house it all.
Then there is the other issue that they only show groups of points facing the same direction. This isn't a perfectly unique world with trillions and trillions of unique points, like you would expect to see in an "unlimited detail" scenario. This is the same handful of objects replicated over and over again. Is that the secret? Do all of the games need to work that way?
Then there is the issue of the very flat, very generic lighting, no animations and no other systems in place a game would need.
If there is an answer to Bruce Dell's claims, its "not ready yet, we'll get back to you when its closer to being fully cooked"
You don't even need to store the inside points for a model. Just the surface points.... just like with traditional geometry.
Similar to procedural texture generation, procedural algorithms can be made to create surface detail made up of point cloud data for objects to save on memory as well.
That tech has gotten quite good. This bathroom image was created entirely with procedural textures.
Compression could be used for all the file types in the engine too. The type of compression that is able to run in hardware, still compressed... like DXT1 or 3Dc.
Similar to procedural texture generation, procedural algorithms can be made to create surface detail made up of point cloud data for objects to save on memory as well.
Procedural stuff still needs to be generated at one point, and stored in memory after that.
And then the data has to be "pre-sorted" to make the voxel/atom search go optimally fast.
But hey, it's Bruce Dell, he can make _anything_ happen, just got to leave it in that magical oven long enough.
You don't even need to store the inside points for a model. Just the surface points.... just like with traditional geometry.
Similar to procedural texture generation, procedural algorithms can be made to create surface detail made up of point cloud data for objects to save on memory as well.
That tech has gotten quite good. This bathroom image was created entirely with procedural textures.
Compression could be used for all the file types in the engine too. The type of compression that is able to run in hardware, still compressed... like DXT1 or 3Dc.
A. Procedural texturing has pretty limited used, and wouldn't ever be used in a situation where you want a really specific detail or effect. It certainly is not viable to use for an entire game. Some tiling concrete wall textures? Sure, but not as the basis for all the textures in your game.
B. Procedural texturing is only a means for generating the content, it still has to be loaded into vmem on the video card, so just because it may take less space on the HD, doesn't actually mean its "free" or anything.
C. These procedural examples always look the same, sure the look nice if you consider it was done all with procedurals, but compared hand made artist content, its not very good. The example you posted for instance, has a very "pernil noise" look to all the textures and just doesn't look very good.
Yeah, only case where you want proceduralism is water, fire, smoke, dust - but those need a lot of processing power to have enough complexity and iterations.
For now water is procedural (and tessellated and displaced) but all the rest work better with bitmaps textured onto transparent quads...
That's assuming that all of the material work is done on the high poly. The tools aren't quite there yet. Generally the high poly is used for just surface height detail and possibly some rough diffuse work. The tools to check materials in game don't really sync up with the viewports and shaders in the sculpting apps.
You're spec could look perfect in Zbrush but once in game it looks different. Forcing you to jump in and out of several apps just to view the final material. That long lag time between iterations is a real drag and roadblock a few very dedicated and talented people have been working very hard to remove.
I don't really disagree with anything you're saying. In fact, you're right on the money. But, I do feel that there could be some ways to get around that issue; like storing multiple color layers for a single mesh. This way, artists could create Spec, Gloss or Emissive in a way similar to how we're doing it now.
Again, you're right about the tools. If they (or anyone else) would like to make a dent in the industry they would have to work together with companies like Pixologic or Autodesk. While Zbrush has a rather unique way of rendering that wouldn't translate well to games, Mudbox (with some work) could be used to quickly iterate materials. Mudbox has some nice 3d painting tools, but they simply aren't up there with Photoshop yet. But again, we might be talking about a timespan of few to several years before something like this could be used to develop games and sculpting apps are getting better and better.
In any case, if tech like this ever becomes feasible, or we get to the point where game are able to run highpoly meshes, then makers of sculpting apps would have to meet the demands of the industry.
Even if this tech takes off I doubt it will be limitless. That's part of being an experienced game artist. Knowing your boundaries and working to the best that you can, within the restricted space.
That's right too. Unlimited is just a buzzword here; things like that will always be hindered by the hardware no matter what optimizations are used, and as many pointed out disk space consumption is a huge problem right now.
I'm not dissing people's ability to deal with these limitations; in fact I do agree that it's an essential part of being a game artist. I'm not bitter about it, but I think it's a shame that we have to spend time and effort taking limitations in the account. Yes, games industry had always been about fighting against them and in some way that's what makes it fun. But with the skyrocketing costs of creating a current gen game, cutting down the production time could greatly improve the state of the industry. Of course, that wouldn't solve all the issues, but it could help.
A good game artist won't design some amazing piece and then get slapped down by restrictions or hobbled by limitations.
A good game artist knows how much sand is in the sand box and which corner the cat hid its prizes.
That's true, but what I'm getting at is that hardware still dictates what can be designed; games today still favor simple shapes. I myself am a fan of "less is more" and I prefer simplistic designs, but it would be nice if one day, artists weren't forced into making that choice. I would love to see some more complex, organic scenes in games.
Tech always out paces tools and user friendly workflows. Right now tessellation is still in the distant future its not like the 360 will support it any time soon. The PS3 could possibly do it? But I don't think it was designed to. It will take a new round of hardware for the tech to take hold and with the way Sony and MS are dragging their feet that will be a while =/ It hasn't even take hold in the PC market which is light years a head of the consoles.
Whenever tessellation takes hold it will probably be a smoother addition to the pipeline than normal maps where. That shit was hideous, buggy, slow, full of incomplete programer-esk tools packed full of errors and glitches. We're just now getting to the point where the tools have caught up to the tech...
I too doubt we'll see tesselation used on current gen hardware. I'm not sure if PS3 is capable of it, but even if it is it would probably be a waste of resources to add a bunch of heightmaps on a system that already suffers from low video memory
Speaking of heightmaps. What I was getting at is, that while tessellation is a great piece of tech, it adds another layer of complexity to the workflow. For the time being it's meant to be used in conjunction with normal maps and as such, it treats symptoms rather than the core problem of not having enough geometry to fully represent the object.
But, perhaps somewhere down the road tessellation will become so powerful that it will make normal mapping obsolete. I hope I don't sound like normal maps had killed my family, but setting them up is often a dull, time consuming task even if you know how to avoid the common mistakes.
They pretty much want to invent a parallel industry that competes or replaces the current industry. It's not just a re-imagining of how we create art (that would probably be the easiest transition). Its a change to the way the hardware is developed, the way games operate on a core level and how they are delivered.
It's not like epic, id or anyone else with current 3D game tech, could write in support for this not so new method. They way Bruce Dell is talking, everyone would have to throw everything out and start over, on all of it.
Again, I can't disagree. They would probably have to start cooperating with game studios and make some compromises if they are serious about wanting to challenge the ways of the industry. Their mention of making a polygon converter is a step in good direction, but it's certainly not enough to convert people. I guess that unless they really have something revolutionary, that blows everything out of the water, they would need backing from key players like AMD and NVIDIA to make the industry adapt (not to mention guys like Autodesk, etc).
A few pages ago some mentioned the idea of a hybrid approach that could use both polygonal and voxel data and I too think that could be a great way to introduce such tech to the industry. Again, I'm not really trying to advocate Euclideon here, nor defend their tech but I do think there is some great potential in voxels/point cloud data/atoms/godly ambrosia/whatever. Especially if someone could solve the issues attributed to the way we do things now.
It could take years of course.
This reply came out awfully long and my English is probably horrible but it's almost 1AM here now, so please forgive me that I didn't fix it In any case, I've learned tons of stuff from this thread so keep posting guys :thumbup:
A high res mesh in itself is always going to be very inefficient, if you do the extra work you can cut down the runtime memory requirements and processing times a LOT. Like, you may have 10 characters from the budget of a single one.
Because runtime resources are always going to be limited for games, it's always going to provide a competitive advantage if you process the source data into a more efficient format. So these conversion-optimization workflows may become more simple, or more automated, but don't expect them to disappear completely.
A high res mesh in itself is always going to be very inefficient, if you do the extra work you can cut down the runtime memory requirements and processing times a LOT. Like, you may have 10 characters from the budget of a single one.
Because runtime resources are always going to be limited for games, it's always going to provide a competitive advantage if you process the source data into a more efficient format. So these conversion-optimization workflows may become more simple, or more automated, but don't expect them to disappear completely.
That's true. But won't we eventually get to the point where those optimizations will not have such a huge impact on performance anymore and could be simply ignored? I'm talking of course about the distant future, of perhaps 10 or 15 years. I'm not really expecting artists to throw 20mil tris on a hydrant, but I hope they will be able to use just a many as they need.
And speaking of optimization; you're right. As artists start getting more resources to spend some optimizations are no longer necessary. Manual retopo will probably get replaced by an automatic way of doing it, once the poly budgets are less picky. Then again, yes, there will always be instances where manual optimizations are necessary; but I hope that one day it will be done primarily to make the mesh cleaner and easier to work with, rather than due to performance reasons.
imho right now the problem with game art isn't the resources, it's the talent.
if anyone thinks that throwing polies and textures at an object will make it look any better then i have bad news: if you can't make it look good with a 512 texture and 500 tris you won't make it look any better if you have unlimited resources.
compare the shitty artists who made the unlimited etail art demos, to say, crysis.
imho right now the problem with game art isn't the resources, it's the talent.
if anyone thinks that throwing polies and textures at an object will make it look any better then i have bad news: if you can't make it look good with a 512 texture and 500 tris you won't make it look any better if you have unlimited resources.
compare the shitty artists who made the unlimited etail art demos, to say, crysis.
If that demo had good lighting/shading it would look incredible. grid like sure but the quality of the scans and the level of fine detail would shine given better conditions.
I guess the idea of this crazy tech is cool... BUT really all I want is more ram in my consoles and for my maps to so being compressed in engine. Baby steps!
what a jackass. Misrepresenting notch's quotes and spouting the same dishonest bullshit is not going to win him any more followers. I cannot imagine notch would disagree with carmack's statement that some kind of all voxel solution may be practical 5-10 years down the line.
Also, you guys here are supposed to be intelligent folks, working in game development with 3D technology.
So tell me, how the hell can you believe that a thousand times more detail won't require at least a hundred times more disk space? You've all seen what happens to an image if you try to compress it to 1% of its original size. If you want to keep the information, you need the storage space. Now what's the plan, they'll just magically fit it onto a DVD?
It's no surprise to see gamers get hoodwinked, but actual game devs should be smarter than that.
Read my post above and stop thinking this is just traditional octrees.
I've already posted many things that would help to get around many issues, and you have to do those things I said to get around the issues, so they are likely doing just that.
Sorry, but people don't get $2 million from the government on a daily basis to finish their product. If you really think the company didn't have to go through an arduous process to get that money then you think it is too easy.
Also, this.
David Merson, CEO of Mincom for 21 years is also investing in Unlimited Detail technology. Mincom is basically the Microsoft of Australia.
Quote from David regarding Unlimited Detail: "it is real, real enough for myself to invest in it"
Plus, the latest video from Hard OCP shows animation and even object collision running in a gameplay scenario.
Next time you quote someone directly, do it properly (I can even hear his voice saying it).
Mincom... is not Microsoft (Dell is incredibly naive if he thinks it is, or unbelievably disingenuous), they're involved in mining, oil, defence contracts and the like, specifically "asset management software" according to their web site, nothing about operating systems :poly141:. So the $200 million (AU) Mincom earn up to 2011 doesn't really compare to almost $70 BILLION (US) for Microsoft in approx the same period now does it?
Interesting side-note: Mincom were apparently bought out (in May this year?) by ABB, which might ring bells for those of you following the shenanigans in Iraq and no-bid contracts.
I've already posted many things that would help to get around many issues, and you have to do those things I said to get around the issues, so they are likely doing just that.
And they haven't got around fixing any of the known limitations since they started.
You made some quick theoretical fixes on things you don't fully understand, it isn't as easy as finding a quick fix, people have been working with this stuff for years, carmack included.
Sorry, but people don't get $2 million from the government on a daily basis to finish their product. If you really think the company didn't have to go through an arduous process to get that money then you think it is too easy.
I doubt they explained the tech and how they would get past the limitations to the government, I suspect it was a case of "look at this tree in this game, now imagine you could have unlimited detail!!!" with some paper examples of some grass polygons and then a super detailed elephant, easily impressed government officials.
Plus, the latest video from Hard OCP shows animation and even object collision running in a gameplay scenario.
The most basic of boundingbox collision detection on a plane, and as suspected on voxel-tech: low framerate keyframed animation without soft-skinning, and that was 7 years ago, had they actually had a sollution for that problem or if their tech was so wildly different from a sparse voxel octree, they would've solved it by now.
All they have done is polished the actual thing you can do good with this tech, lots of static instanced data, and as noticed, rotational issues.
Sorry, but people don't get $2 million from the government on a daily basis to finish their product. If you really think the company didn't have to go through an arduous process to get that money then you think it is too easy..
Yes they do. Nations give out grants in the tens of billions yearly.
Googling showed technology and r&d grants from one part of the australian government were 5,000 million in 2004. [1]
It proves they are a real business that intends to make profit, and that they have a real technology they can market and sell, but not that it's anything special or some amazing super breakthrough.
As for the grant Euclideon actually got, it was 33.6 million over 888 [2] projects, one of which was Euclideon.
Considering the actual budget for r&d in australia is somewhere in the billions [3] yearly
Conning governments is easy. Here we have people paying £3000 pounds for a standard office computer. No one cares where the money goes after all its not theres.
No one cares where the money goes as long as it makes a return on the investment. Governments invest in things that will be profitable and drive their economy, and a lot of the individual investments aren't necessarily going to turn out well. The technology does not have to be good in order to have the potential to sell a lot and help the australian economy.
No one cares where the money goes as long as it makes a return on the investment. Governments invest in things that will be profitable and drive their economy, and a lot of the individual investments aren't necessarily going to turn out well. The technology does not have to be good in order to have the potential to sell a lot and help the australian economy.
If the goal behind the tech is to replace or in some degree even be better than polygonal tech or previous existing voxel tech, but that doesn't stop the government from believing it will be profitable.
Money has only gone in, not out from unlimited detail, and that's since 2003.
I'm sure some of you are sick of this topic by now so sorry for the bump, but here's a new interview that i don't think has been posted here yet, and Bruce is asked straight out about the most common doubt questions.
Still no crystal clear answers but you might get something out of it. http://pc.gamespy.com/articles/118/1187338p1.html
Even if this tech was possible, this guys is pretty much blaming everyone else for everything.
Bogged down pipeline in games? The industries fault at large.
Doubted his credibility? Programmers fault for not thinking outside the box.
No final animation? Had to revamp the system.
Tiled look? We don't have uber artists.
Whats next? Project failed because Government didn't pay us enough once a few people too many told them this was a failure?
here's a new interview that i don't think has been posted here yet, and Bruce is asked straight out about the most common doubt questions.
He doesn't answer any of those questions at all, every response is evasive.
Honestly, let's get these things straight...
A single person with no experience in game development, computer graphics, or even serious programming at all, has managed to create a new rendering method that can achieve unlimited results on limited resources. (which is, by formal logic, actually impossible)
While this tech looks extremely similar to existing tech, exhibits the same limitations and lack of features that existing tech has, it's still something completely new and different; or at least it's capable of overcoming all the known and inherent drawbacks.
Also, if it's all true then every single game we see today is using technology infinitely more inefficient, that literally ten thousands of people working in computer graphics, backed by billions of dollars of investment, are completely wrong about everything; and they're not only too stupid to see it, but are desperately holding on to their inefficient methods and approaches against all common sense.
And we're supposed to believe all of the above just because this one guy says so.
Yes, the company has been accepting many interviews now just to say later that the technology doesn't really exist.... Right....
Interviews have been performed exclusively by people with little technical background to even ask any kind of relevant questions, and always on unlimited details terms.
This is just a great example on an uninformed bad comment on games art:
"I watched the RAGE trailer this week after seeing Euclideons demo on Monday, and the graphics in RAGE looked horrible in comparison."
People without the technical background to understand what is happening will see their demo and fill in the blanks, they'll assume everything will have that detail, they believe things will be unique everywhere, and that the world will be destructible and have the full physics property per atom.
I believe Bruce Dell himself in his outmost excitement believes that he will overcome all these issues that they most likely dont have a sollution yet.
He's like the tech worlds answer to Peter Molyneux, hyping something beyond its capability and end-result before he even knows it will deliver.
What they need to show to be competitive with today's realtime rendering engines in general:
- HDR gamma correct internal rendering, linear lighting (I actually doubt that Dell even knows what this is)
- high number of dynamic lights and cast shadows (preferably use deferred rendering)
- at least 15-20 fully animated characters using skeletial deformation
- interactive physics
- complex particle systems, volumetric effects for dust, fire, smoke
- dynamic water rendering
- fully featured post processing pipeline for DOF, some level of motion blur, color correction, bloom etc.
Oh and some basic stuff, like terrain that has actual elevation. Has noone noticed that the entire demo island is completely flat??
Also, this is just for rendering, but a fully functional game engine is about a lot more, needless to say.
I'd say Cryengine 3 and Unreal Engine 3 are the ones to beat here, and I'm sure Valve's at work on the next iteration of Source too.
(By the way Rage doesn't have HDR and fully dynamic lighting, but it's at least 60fps and it offers fully unique texturing instead. Id also doesn't want to license it, so they don't have to care about what a competitive engine needs to be.)
"I watched the RAGE trailer this week after seeing Euclideons demo on Monday, and the graphics in RAGE looked horrible in comparison."
People without the technical background to understand what is happening will see their demo and fill in the blanks, they'll assume everything will have that detail, they believe things will be unique everywhere, and that the world will be destructible and have the full physics property per atom.
Tech is one thing, but Rage is like in a different world. The art direction and the level of polish in the artwork of that game is quite unbelievable.
Rest assured though that a lot of the general audience can see this difference. It's mostly the uninformed hardware geeks who get more excited by the minute detail of the same palm tree, instanced a thousand times...
I'd be happy with just arbitrarily positioned and rotated objects.
I was referring to molyneux uncanny ability to get excited about things before he even knows he will be able to deliver, which has been more prominent in his studios later titles.
So doesn't that make you suspicious at all? He fails to show you stuff that's absolutely required for any AA game, he also fails to explain how he would implement it, shouldn't he be able to come up with something other than Cinderella and other fairy tale characters?
Yes, the company has been accepting many interviews now just to say later that the technology doesn't really exist.... Right....
Yes! None of what they're doing is illegal or fraudulent, just misleading. They don't need to have tech that really changes the 3d industry to make a profit, and they do need publicity to get a lot of funding.
Interviews have been performed exclusively by people with little technical background to even ask any kind of relevant questions, and always on unlimited details terms.
This is just a great example on an uninformed bad comment on games art:
"I watched the RAGE trailer this week after seeing Euclideons demo on Monday, and the graphics in RAGE looked horrible in comparison."
People without the technical background to understand what is happening will see their demo and fill in the blanks, they'll assume everything will have that detail, they believe things will be unique everywhere, and that the world will be destructible and have the full physics property per atom.
I believe Bruce Dell himself in his outmost excitement believes that he will overcome all these issues that they most likely dont have a sollution yet.
He's like the tech worlds answer to Peter Molyneux, hyping something beyond its capability and end-result before he even knows it will deliver.
John Gatt was the international technical specialist at VIA/S3 for over 10 years, I think he knows his stuff. There might be stuff he is not telling you though... John did a talk on S3TC in Sydney with Tim Wilts at the launch of quake 3 i think i have on video some ware...
He's from marketing, I should have guessed; not exactly a genuine tech guru, or news anchor from what I gather. And now that's all I can really see that "interview" as, a friendly little marketing setup.
Honestly, what I've seen in all the videos they've shown for -unlimited detail- is neither unlimited, nor is it NEW.
C'mon, we've all seen tech demo's that show off so much more than this and just make more sense from a development standpoint. Tools, development history, documentation, etc.
I could really care less about an engine that uses rgb colors or vectors as id's to load cubic chunks of instances. That all are anchored to a strict grid.
Lets see what happens when they try to actually replicate something like Crysis down to every detail, with physics, lighting, foliage, etc.
And yes, government people are easily fooled by buzz words and flashy promotions, so don't let $2million in investments fool you.
Businessmen and government bureaucrats will throw down a million or so on a small company, simply based on image and the most BS pitch you ever heard, or if you simply buy them a lap dance; they get to write it off in the yearly expenses anyway.
It sounds like one of the investors is probably more interested in trying to get a geological density viewer out of this more than anything else. though I don't exactly see that happening with how they're managing asset instance chunks at the moment.
It's like everyone is trying to make connections that aren't there as to why this isn't legit....
PolyMonster: No, this isn't old tech. This runs in real time on a single core laptop CPU in software mode. Pure voxels could never hope to achieve that.
It's like everyone is trying to make connections that aren't there as to why this isn't legit....
PolyMonster: No, this isn't old tech. This runs in real time on a single core laptop CPU in software mode. Pure voxels could never hope to achieve that.
Hmm,I was curious as to how new this was, and how rare the tech running on a CPU http://homepages.slingshot.co.nz/~tmulgrew/VoxelOctree/ Runs great on my old CPU. Honestly I'd love to be able to hack in some objects into that demo to see how it does with a heavier load of objects. Ive seen references to SVO going back to 2008, and that's just from John Olick showcasing it, not necessarily the moment it was first tested. It would be nice to know its earlier origins, I didn't find that information right off the bat.
RexM: Do you have some personal stake in this company or something? You seem to be getting quite upset by what most industry veterans can clearly see as a hyped up tech demo that is years, or decades off from usability, and some asshat who is making ridiculous claims spouting buzzwords and ignoring any real issues with actually using this stuff to make a game.
I just don't see why anyone would so obediently defend this guy.
He's from marketing, I should have guessed; not exactly a genuine tech guru, or news anchor from what I gather. And now that's all I can really see that "interview" as, a friendly little marketing setup.
Honestly, what I've seen in all the videos they've shown for -unlimited detail- is neither unlimited, nor is it NEW.
C'mon, we've all seen tech demo's that show off so much more than this and just make more sense from a development standpoint. Tools, development history, documentation, etc.
I could really care less about an engine that uses rgb colors or vectors as id's to load cubic chunks of instances. That all are anchored to a strict grid.
Lets see what happens when they try to actually replicate something like Crysis down to every detail, with physics, lighting, foliage, etc.
And yes, government people are easily fooled by buzz words and flashy promotions, so don't let $2million in investments fool you.
Businessmen and government bureaucrats will throw down a million or so on a small company, simply based on image and the most BS pitch you ever heard, or if you simply buy them a lap dance; they get to write it off in the yearly expenses anyway.
It sounds like one of the investors is probably more interested in trying to get a geological density viewer out of this more than anything else. though I don't exactly see that happening with how they're managing asset instance chunks at the moment.
Ding! Winner. I mentioned it a few pages back, but I'm glad someone has brought it back up again. The guy is a marketer and not a reporter working for a 3rd party - that's all you need to know to determine the legitimacy of the interview.
It's like everyone is trying to make connections that aren't there as to why this isn't legit....
PolyMonster: No, this isn't old tech. This runs in real time on a single core laptop CPU in software mode. Pure voxels could never hope to achieve that.
1998 called. It would like you to retract your statement. Voxels/Atoms are tried, tested, and dropped for a variety of reasons. They have applications, but this is not an example of them.
This runs in real time on a single core laptop CPU in software mode. Pure voxels could never hope to achieve that.
Don't make yourself look bad here. That's an i7 quad core CPU in a gaming laptop, it also has 8 GB RAM, which I think is kinda necessary to work with all these voxels.
This machine he has is like 10 times better than your current game console and still a lot better than the average laptop. Just to keep things in perspective.
RexM: Do you have some personal stake in this company or something? You seem to be getting quite upset by what most industry veterans can clearly see as a hyped up tech demo that is years, or decades off from usability, and some asshat who is making ridiculous claims spouting buzzwords and ignoring any real issues with actually using this stuff to make a game.
I just don't see why anyone would so obediently defend this guy.
Polygons aren't going to be around forever. Eventually we are going to switch to something else.
This could very easily be that 'something else,' even if we have to mix it with polygons for a while.
There isn't really anything unbelievable about the software, I still cannot really understand the disbelief.
They've likely found ways around many of the issues, and also there is a video that shows collision and animation from these guys, they showed it in the Hard OCP interview... so what are the reasons as to why this isn't applicable again?
I am only basing all my opinions on this with what they have shown. They've shown detailed objects that look quite real, they've shown collisions and animation.... looks pretty solid to me.
They've likely found ways around many of the issues, and also there is a video that shows collision and animation from these guys, they showed it in the Hard OCP interview... so what are the reasons as to why this isn't applicable again?
sounds like wishful thinking to me.
They've shown some static objects with static lighting. until they show working tech people wont believe it, and rightly so.
Replies
Then there is the other issue that they only show groups of points facing the same direction. This isn't a perfectly unique world with trillions and trillions of unique points, like you would expect to see in an "unlimited detail" scenario. This is the same handful of objects replicated over and over again. Is that the secret? Do all of the games need to work that way?
Then there is the issue of the very flat, very generic lighting, no animations and no other systems in place a game would need.
If there is an answer to Bruce Dell's claims, its "not ready yet, we'll get back to you when its closer to being fully cooked"
Similar to procedural texture generation, procedural algorithms can be made to create surface detail made up of point cloud data for objects to save on memory as well.
That tech has gotten quite good. This bathroom image was created entirely with procedural textures.
http://download.profxengine.com/gallery/renders/bathroom_dirty_0.jpg
http://images.bit-tech.net/content_images/Procedural_Textures_Future_Gam/02.jpg
http://www.mapzoneeditor.com/index.php?PAGE=GALLERY.RENDERS
Compression could be used for all the file types in the engine too. The type of compression that is able to run in hardware, still compressed... like DXT1 or 3Dc.
Procedural stuff still needs to be generated at one point, and stored in memory after that.
And then the data has to be "pre-sorted" to make the voxel/atom search go optimally fast.
But hey, it's Bruce Dell, he can make _anything_ happen, just got to leave it in that magical oven long enough.
In life, shit is limitless despite what we may have is limited.
A. Procedural texturing has pretty limited used, and wouldn't ever be used in a situation where you want a really specific detail or effect. It certainly is not viable to use for an entire game. Some tiling concrete wall textures? Sure, but not as the basis for all the textures in your game.
B. Procedural texturing is only a means for generating the content, it still has to be loaded into vmem on the video card, so just because it may take less space on the HD, doesn't actually mean its "free" or anything.
C. These procedural examples always look the same, sure the look nice if you consider it was done all with procedurals, but compared hand made artist content, its not very good. The example you posted for instance, has a very "pernil noise" look to all the textures and just doesn't look very good.
For now water is procedural (and tessellated and displaced) but all the rest work better with bitmaps textured onto transparent quads...
I don't really disagree with anything you're saying. In fact, you're right on the money. But, I do feel that there could be some ways to get around that issue; like storing multiple color layers for a single mesh. This way, artists could create Spec, Gloss or Emissive in a way similar to how we're doing it now.
Again, you're right about the tools. If they (or anyone else) would like to make a dent in the industry they would have to work together with companies like Pixologic or Autodesk. While Zbrush has a rather unique way of rendering that wouldn't translate well to games, Mudbox (with some work) could be used to quickly iterate materials. Mudbox has some nice 3d painting tools, but they simply aren't up there with Photoshop yet. But again, we might be talking about a timespan of few to several years before something like this could be used to develop games and sculpting apps are getting better and better.
In any case, if tech like this ever becomes feasible, or we get to the point where game are able to run highpoly meshes, then makers of sculpting apps would have to meet the demands of the industry.
That's right too. Unlimited is just a buzzword here; things like that will always be hindered by the hardware no matter what optimizations are used, and as many pointed out disk space consumption is a huge problem right now.
I'm not dissing people's ability to deal with these limitations; in fact I do agree that it's an essential part of being a game artist. I'm not bitter about it, but I think it's a shame that we have to spend time and effort taking limitations in the account. Yes, games industry had always been about fighting against them and in some way that's what makes it fun. But with the skyrocketing costs of creating a current gen game, cutting down the production time could greatly improve the state of the industry. Of course, that wouldn't solve all the issues, but it could help.
That's true, but what I'm getting at is that hardware still dictates what can be designed; games today still favor simple shapes. I myself am a fan of "less is more" and I prefer simplistic designs, but it would be nice if one day, artists weren't forced into making that choice. I would love to see some more complex, organic scenes in games.
I too doubt we'll see tesselation used on current gen hardware. I'm not sure if PS3 is capable of it, but even if it is it would probably be a waste of resources to add a bunch of heightmaps on a system that already suffers from low video memory
Speaking of heightmaps. What I was getting at is, that while tessellation is a great piece of tech, it adds another layer of complexity to the workflow. For the time being it's meant to be used in conjunction with normal maps and as such, it treats symptoms rather than the core problem of not having enough geometry to fully represent the object.
But, perhaps somewhere down the road tessellation will become so powerful that it will make normal mapping obsolete. I hope I don't sound like normal maps had killed my family, but setting them up is often a dull, time consuming task even if you know how to avoid the common mistakes.
Again, I can't disagree. They would probably have to start cooperating with game studios and make some compromises if they are serious about wanting to challenge the ways of the industry. Their mention of making a polygon converter is a step in good direction, but it's certainly not enough to convert people. I guess that unless they really have something revolutionary, that blows everything out of the water, they would need backing from key players like AMD and NVIDIA to make the industry adapt (not to mention guys like Autodesk, etc).
A few pages ago some mentioned the idea of a hybrid approach that could use both polygonal and voxel data and I too think that could be a great way to introduce such tech to the industry. Again, I'm not really trying to advocate Euclideon here, nor defend their tech but I do think there is some great potential in voxels/point cloud data/atoms/godly ambrosia/whatever. Especially if someone could solve the issues attributed to the way we do things now.
It could take years of course.
This reply came out awfully long and my English is probably horrible but it's almost 1AM here now, so please forgive me that I didn't fix it
Because runtime resources are always going to be limited for games, it's always going to provide a competitive advantage if you process the source data into a more efficient format. So these conversion-optimization workflows may become more simple, or more automated, but don't expect them to disappear completely.
That's true. But won't we eventually get to the point where those optimizations will not have such a huge impact on performance anymore and could be simply ignored? I'm talking of course about the distant future, of perhaps 10 or 15 years. I'm not really expecting artists to throw 20mil tris on a hydrant, but I hope they will be able to use just a many as they need.
And speaking of optimization; you're right. As artists start getting more resources to spend some optimizations are no longer necessary. Manual retopo will probably get replaced by an automatic way of doing it, once the poly budgets are less picky. Then again, yes, there will always be instances where manual optimizations are necessary; but I hope that one day it will be done primarily to make the mesh cleaner and easier to work with, rather than due to performance reasons.
if anyone thinks that throwing polies and textures at an object will make it look any better then i have bad news: if you can't make it look good with a 512 texture and 500 tris you won't make it look any better if you have unlimited resources.
compare the shitty artists who made the unlimited etail art demos, to say, crysis.
If that demo had good lighting/shading it would look incredible. grid like sure but the quality of the scans and the level of fine detail would shine given better conditions.
what a jackass. Misrepresenting notch's quotes and spouting the same dishonest bullshit is not going to win him any more followers. I cannot imagine notch would disagree with carmack's statement that some kind of all voxel solution may be practical 5-10 years down the line.
Read my post above and stop thinking this is just traditional octrees.
I've already posted many things that would help to get around many issues, and you have to do those things I said to get around the issues, so they are likely doing just that.
Sorry, but people don't get $2 million from the government on a daily basis to finish their product. If you really think the company didn't have to go through an arduous process to get that money then you think it is too easy.
Also, this.
David Merson, CEO of Mincom for 21 years is also investing in Unlimited Detail technology. Mincom is basically the Microsoft of Australia.
Quote from David regarding Unlimited Detail: "it is real, real enough for myself to invest in it"
Plus, the latest video from Hard OCP shows animation and even object collision running in a gameplay scenario.
Mincom... is not Microsoft (Dell is incredibly naive if he thinks it is, or unbelievably disingenuous), they're involved in mining, oil, defence contracts and the like, specifically "asset management software" according to their web site, nothing about operating systems :poly141:. So the $200 million (AU) Mincom earn up to 2011 doesn't really compare to almost $70 BILLION (US) for Microsoft in approx the same period now does it?
Interesting side-note: Mincom were apparently bought out (in May this year?) by ABB, which might ring bells for those of you following the shenanigans in Iraq and no-bid contracts.
Even bruce himself admitted to calling them atoms instead of voxels, and every sign and known limitation does point towards something similar.
And they haven't got around fixing any of the known limitations since they started.
You made some quick theoretical fixes on things you don't fully understand, it isn't as easy as finding a quick fix, people have been working with this stuff for years, carmack included.
I doubt they explained the tech and how they would get past the limitations to the government, I suspect it was a case of "look at this tree in this game, now imagine you could have unlimited detail!!!" with some paper examples of some grass polygons and then a super detailed elephant, easily impressed government officials.
Another impressed guy in a suit, who sees it in real time and goes "this is real!! it wasn't fake!!"
The most basic of boundingbox collision detection on a plane, and as suspected on voxel-tech: low framerate keyframed animation without soft-skinning, and that was 7 years ago, had they actually had a sollution for that problem or if their tech was so wildly different from a sparse voxel octree, they would've solved it by now.
All they have done is polished the actual thing you can do good with this tech, lots of static instanced data, and as noticed, rotational issues.
Yes they do. Nations give out grants in the tens of billions yearly.
Googling showed technology and r&d grants from one part of the australian government were 5,000 million in 2004. [1]
It proves they are a real business that intends to make profit, and that they have a real technology they can market and sell, but not that it's anything special or some amazing super breakthrough.
As for the grant Euclideon actually got, it was 33.6 million over 888 [2] projects, one of which was Euclideon.
Considering the actual budget for r&d in australia is somewhere in the billions [3] yearly
If the goal behind the tech is to replace or in some degree even be better than polygonal tech or previous existing voxel tech, but that doesn't stop the government from believing it will be profitable.
Money has only gone in, not out from unlimited detail, and that's since 2003.
Still no crystal clear answers but you might get something out of it.
http://pc.gamespy.com/articles/118/1187338p1.html
And a short thingy here including how they built the world which is interesting imo and can explain the lack or variation in rotations etc:
http://www.gamingface.com/2011/08/unlimited-detail-in-swtor-or-other-mmo.html
Bogged down pipeline in games? The industries fault at large.
Doubted his credibility? Programmers fault for not thinking outside the box.
No final animation? Had to revamp the system.
Tiled look? We don't have uber artists.
Whats next? Project failed because Government didn't pay us enough once a few people too many told them this was a failure?
He doesn't answer any of those questions at all, every response is evasive.
Honestly, let's get these things straight...
A single person with no experience in game development, computer graphics, or even serious programming at all, has managed to create a new rendering method that can achieve unlimited results on limited resources. (which is, by formal logic, actually impossible)
While this tech looks extremely similar to existing tech, exhibits the same limitations and lack of features that existing tech has, it's still something completely new and different; or at least it's capable of overcoming all the known and inherent drawbacks.
Also, if it's all true then every single game we see today is using technology infinitely more inefficient, that literally ten thousands of people working in computer graphics, backed by billions of dollars of investment, are completely wrong about everything; and they're not only too stupid to see it, but are desperately holding on to their inefficient methods and approaches against all common sense.
And we're supposed to believe all of the above just because this one guy says so.
Seriously, is everyone so easy to fool nowadays?
Interviews have been performed exclusively by people with little technical background to even ask any kind of relevant questions, and always on unlimited details terms.
This is just a great example on an uninformed bad comment on games art:
"I watched the RAGE trailer this week after seeing Euclideons demo on Monday, and the graphics in RAGE looked horrible in comparison."
People without the technical background to understand what is happening will see their demo and fill in the blanks, they'll assume everything will have that detail, they believe things will be unique everywhere, and that the world will be destructible and have the full physics property per atom.
I believe Bruce Dell himself in his outmost excitement believes that he will overcome all these issues that they most likely dont have a sollution yet.
He's like the tech worlds answer to Peter Molyneux, hyping something beyond its capability and end-result before he even knows it will deliver.
- HDR gamma correct internal rendering, linear lighting (I actually doubt that Dell even knows what this is)
- high number of dynamic lights and cast shadows (preferably use deferred rendering)
- at least 15-20 fully animated characters using skeletial deformation
- interactive physics
- complex particle systems, volumetric effects for dust, fire, smoke
- dynamic water rendering
- fully featured post processing pipeline for DOF, some level of motion blur, color correction, bloom etc.
Oh and some basic stuff, like terrain that has actual elevation. Has noone noticed that the entire demo island is completely flat??
Also, this is just for rendering, but a fully functional game engine is about a lot more, needless to say.
I'd say Cryengine 3 and Unreal Engine 3 are the ones to beat here, and I'm sure Valve's at work on the next iteration of Source too.
(By the way Rage doesn't have HDR and fully dynamic lighting, but it's at least 60fps and it offers fully unique texturing instead. Id also doesn't want to license it, so they don't have to care about what a competitive engine needs to be.)
Tech is one thing, but Rage is like in a different world. The art direction and the level of polish in the artwork of that game is quite unbelievable.
Rest assured though that a lot of the general audience can see this difference. It's mostly the uninformed hardware geeks who get more excited by the minute detail of the same palm tree, instanced a thousand times...
I was referring to molyneux uncanny ability to get excited about things before he even knows he will be able to deliver, which has been more prominent in his studios later titles.
Yea I said he was asked the questions, not that he answered them.
Yes! None of what they're doing is illegal or fraudulent, just misleading. They don't need to have tech that really changes the 3d industry to make a profit, and they do need publicity to get a lot of funding.
John Gatt was the international technical specialist at VIA/S3 for over 10 years, I think he knows his stuff. There might be stuff he is not telling you though... John did a talk on S3TC in Sydney with Tim Wilts at the launch of quake 3 i think i have on video some ware...
http://au.linkedin.com/pub/john-gatt/23/356/251
He's from marketing, I should have guessed; not exactly a genuine tech guru, or news anchor from what I gather. And now that's all I can really see that "interview" as, a friendly little marketing setup.
Honestly, what I've seen in all the videos they've shown for -unlimited detail- is neither unlimited, nor is it NEW.
C'mon, we've all seen tech demo's that show off so much more than this and just make more sense from a development standpoint. Tools, development history, documentation, etc.
I could really care less about an engine that uses rgb colors or vectors as id's to load cubic chunks of instances. That all are anchored to a strict grid.
Lets see what happens when they try to actually replicate something like Crysis down to every detail, with physics, lighting, foliage, etc.
And yes, government people are easily fooled by buzz words and flashy promotions, so don't let $2million in investments fool you.
Businessmen and government bureaucrats will throw down a million or so on a small company, simply based on image and the most BS pitch you ever heard, or if you simply buy them a lap dance; they get to write it off in the yearly expenses anyway.
It sounds like one of the investors is probably more interested in trying to get a geological density viewer out of this more than anything else. though I don't exactly see that happening with how they're managing asset instance chunks at the moment.
Our Brisbane Start-Up dev team would love even half of that Government grant. lame.
PolyMonster: No, this isn't old tech. This runs in real time on a single core laptop CPU in software mode. Pure voxels could never hope to achieve that.
Hmm,I was curious as to how new this was, and how rare the tech running on a CPU http://homepages.slingshot.co.nz/~tmulgrew/VoxelOctree/ Runs great on my old CPU. Honestly I'd love to be able to hack in some objects into that demo to see how it does with a heavier load of objects. Ive seen references to SVO going back to 2008, and that's just from John Olick showcasing it, not necessarily the moment it was first tested. It would be nice to know its earlier origins, I didn't find that information right off the bat.
I just don't see why anyone would so obediently defend this guy.
the best place for this tech, is the military.
Ding! Winner. I mentioned it a few pages back, but I'm glad someone has brought it back up again. The guy is a marketer and not a reporter working for a 3rd party - that's all you need to know to determine the legitimacy of the interview.
1998 called. It would like you to retract your statement. Voxels/Atoms are tried, tested, and dropped for a variety of reasons. They have applications, but this is not an example of them.
Don't make yourself look bad here. That's an i7 quad core CPU in a gaming laptop, it also has 8 GB RAM, which I think is kinda necessary to work with all these voxels.
This machine he has is like 10 times better than your current game console and still a lot better than the average laptop. Just to keep things in perspective.
The tech only needs to render one atom per pixel.
For 1920x1080, that is 2,073,600 atoms.
Just like how screen space effects tend to have consistent performance regardless of how complex the scene is.
Polygons aren't going to be around forever. Eventually we are going to switch to something else.
This could very easily be that 'something else,' even if we have to mix it with polygons for a while.
There isn't really anything unbelievable about the software, I still cannot really understand the disbelief.
They've likely found ways around many of the issues, and also there is a video that shows collision and animation from these guys, they showed it in the Hard OCP interview... so what are the reasons as to why this isn't applicable again?
I am only basing all my opinions on this with what they have shown. They've shown detailed objects that look quite real, they've shown collisions and animation.... looks pretty solid to me.
sounds like wishful thinking to me.
They've shown some static objects with static lighting. until they show working tech people wont believe it, and rightly so.