Unlimited Detail Engine defended
Euclideon CEO Bruce Dell responds to some skeptical critics of the seemingly impressive Unlimited Detail Engine.
If you follow the games industry closely enough, chances are you heard about Euclideon's Unlimited Detail Engine. The engine uses voxels rather than polygons, and has promised a level of graphical fidelity orders of magnitude greater than anything we've seen using standard current tech. It all seems a little too good to be true, and the company's silence on the issue gave cause for some skepticism.
Now CEO Bruce Dell has defended the company's bold claims by allowing journalist access for a personal look at the engine's demo, and responded to some of its harshest critics.
GameInformer's hands-on claims that the demo video shown previously was legitimate, though the company arranges the voxels in a clever way to render them. The "atoms" (i.e. voxels) are arranged to be one per pixel for maximum efficiency. Essentially, the system decides what points need to be displayed to give the desired effect.
Well-known industry figures chimed in on the engine when it was shown, some more harshly than others. John Carmack, CEO of id Software, was impressed but said the technology was several years away. "[Carmack] was unaware at the time that we were running this purely in software," Dell said. "In his particular case, he and Intel had tried to go down a similar road themselves. Intel had tried making its own system in order to do things along the field of unlimited graphics. It ended up closing that avenue as it figured it was something for the distant future when computers have more power."
Meanwhile, Markus "Notch" Persson went so far as to claim the whole thing was a scam. "On the one hand, I could be nice and say, 'yes, there are scammers out there – it's fair enough to look at something and immediately label it as a scam," Dell said. "But on the other – which people have brought up – the man compared [Unlimited Detail] to three other engines. The mistakes Notch made were so bad that, if we were less kind, we would be able to really discredit his actual understanding of a lot of stuff in general. But that's honestly not our intention: we don't want to make enemies with him. We prefer to bring him around and be nice to him."
Critics have also observed the object repetition in their demo, claiming it's an inherent weakness in the engine. Dell claims that this is simply a product of the company scanning in objects on a short time frame, and says the company's next demo makes use of ex-THQ staff artists to show off more variety.
Even so, the engine isn't perfect. Euclideon claims it runs at 20 FPS at the time, and there's still plenty of room for improvement. Plus, it's a tech company, so Dell often notes that industry artists would be able to render better results. Still, Dell says he's aiming for less than a year before the software development kit starts hitting the hands of developers, so we may not have to wait much longer for devs themselves to give feedback on the viability of the system.
-
Steve Watts posted a new article, Unlimited Detail Engine defended.
Euclideon CEO Bruce Dell responds to some skeptical critics of the seemingly impressive Unlimited Detail Engine.-
-
I don't think so. I think they are legit as far as yes they actually have the tech up and running and what they showed is real. Their problem though is that I don't think their tech is very flexible and probably a nightmare to work with. There are so many other factors when choosing an engine and making a game than how detailed your meshes are. Having a robust material shader engine is probably more important. But like I said, there are so many other things developers want than just more detailed geometry. Any compromises you have to make to get this tech to work is where this starts to look like a scam.
-
-
-
-
Not like that. There is absolutely nothing impossible about what you see in the videos and the techniques have been widely studies for a long time already, the trouble is this kind of tech is almost completely useless for games and I have a hard time thinking of other purposes. Anything the guy says to counter that are bullshit.
I think it's a scam in the sense that they are making promises to get investor money and they will keep it up as long as they can. They have been at it for a long time now (I think like 5 years ago they were showing off the same stuff) and no real progress has been made because there is no real progress possible, it's that simple a procedure.
Any graphics programmer could build this thing in a few months but they wouldn't want to wast their time.
-
-
-
-
-
This thing again? Two problems: 1) massive data sets, the storage and RAM requirements for non-repetitive "unlimited detail" will be astromical 2) animation and destruction are basically impossible unless they are precomputed, so it only provides a static or scripted world, never dynamic or responsive to player input.
It's still kinda cool tech but it's not something you would really make a modern game out of. Nobody complains that the backdrops in other engines don't have enough fine detail. It's more important to have good animation and good art direction and a solid game mechanic than it is to have the ability to zoom in on a quarter lying on a tabletop.-
Well, ok, not "nobody complains". Lots of people complain. But spending a ton of time and doubling or tripling your artist workload in order to add higher detail to non-game-essential objects is not a good investment of development resources IMO. Better to spend that effort making the game more fun.
-
-
Here's an example:
http://www.youtube.com/watch?v=tkn6ubbp1SE&feature=player_embedded?
Not really any better than current systems and quite a few more problems.
-
-
-
-
-
-
-
-
To everyone hating, research other voxel engine based games, ie, Outcast.
That game is ancient now by internet/gaming standards, but it pulled off awesome looking style and graphics. Running entirely in software render too. There are a few other voxel games out there worth checking out also.
I've thought for a while that a really high end voxel engine could do some really cool stuff, glad to see someone else agrees, eager to see it.
Just wait, if it comes out and developers make some big name games with it, everyone in this comment thread is going to be like "Oh yeah, Voxel engine, I remember when they announced it, knew it was going to rock!".-
YDKWTF. They only used voxels for certain parts of the game for good reason:
http://www.youtube.com/watch?v=gMePgitDers
I make games because I love to, but I also have to make money. Please tell me where the stable tools, engines and proficient workforce is to make a game using this tech. -
Outcast was also famous for basically being what Crysis was (and still is to some point). It ran like crap on modern PC's of the time unless you settled for a resolution of 320x200, it's maximum resolution was 512x384. It also still used polygons for everything - Except for the landscape (and water, i think).
Same with Delta Force 1 and 2 - Landscape used voxels, everything else was polygons and the games still had to run at 320x200 to be playable on most contemporary machines.
That, and the detail of both games' landscapes ultimately wasn't THAT impressive, especially in the DF games - Games like OFP and BF1942 weren't really that far away and outdid voxels easily with polygons. Outcast 2 was also going to eschew voxels in favour of polygons.
-
As far as I'm concerned as a physicist, Voxel would be much better than polygons. Imagine you could assign mass and electrostatic repulsion to every voxel and then simply plug in Einsteinian (or even just Newtonian) equations of motion approximated to much larger sizes. You could have perfect friction, air resistance, gravity etc.
-
Markus 'Notch' Persson, creator of Minecraft covered this when it was first announced. You can read his explanation of why it's a scam on his blog here... http://notch.tumblr.com/post/8423008802/but-notch-its-not-a-scam
-
Back when this was hot news, someone got an interview with a live demo on tape. In that live demo the Unlimited Detail Engine was running on a laptop and they made a point to show there was no ethernet cable, or in fact, in cables attached to it. Then they proceded with a live demo. One where they moved around the map. So, they demonstrated that a laptop, with an i7 2630QM and 8GB of DDR3@1333 ran the demo everyone was calling fake.
For all you haters here's the link: http://www.youtube.com/watch?v=JVB1ayT6Fdc
Go to 22:20, watch, then stfu, put a cone hat on ur head, and sit in the corner where you belong.-
-
I think what most are overlooking is that they aren't even utilizing the gpu in that demo. The demo is completely software based using the cpu. It's possible that the demo, and by result a game, is using up an impractically large amount of HDD space. Perhaps that's it's achilles heel. However, seeing as so many people have been wrong about this demo in the past, about it being real, about it being practical on consumer hardware from a processing standpoint, I'm willing to entertain that these people who've created something out of the box and functional will continue to seek methods to make it entirely practical.
Besides, why would they be getting millions of dollars of funding if the tech so impossible to implement based upon such an obvious and insurmountable thing such as storage.-
Where have people been so wrong about the demo in the past? Show us, where people have been wrong. Even with using the GPU it's not going to equal the performance that a traditional poly rendering system any time soon. Also, governments put millions into R&D all the time that goes nowhere. Show me where any company involved in the games market like Intel, AMD, Nvidia, etc, have invested millions in this engine. Just because he can keep getting enough cash to keep going doesn't mean a damn thing.
-
I haven't bothered looking at this topic since that video I posted. It's been the back of my mind though. As I recall, in terms of people being wrong, there were a great many people claiming it was pre-rendered and not in-engine. A very very great many of people. They were WRONG. If you were so out of touch to not remember that, so be it. If you simply feel it's not pertinent, then whatever as no doubt you'll want to talk about being wrong from a technical standpoint and from a technical standpoint I'm willing to bet the guy's who made this engine know a lot more about it than most anyone speculating about it on forums, or even, inside the graphics industry.
It's true governments pour cash into projects that can't go anywhere. However, the government isn't their only investor. There's been a lot of interest from a lot of different companies. It might not be accurate to say that their interest means that this tech is feasible to implement, but it's not that big of a stretch either.
Also, I'll just say this in regards to poly systems. While it's true this engine isn't as advanced as any traditional poly engine I'd remind you again that no poly engine can, or does, look as good as the Unlimited Detail demo by solely working in software, on a cpu. So, in terms of resolution, and performance, once they start utilizing the massive processing power of gpus you'll start to see that massive increase in performance. No doubt, that if, or when, this tech becomes standard you'll start seeing gpu's with hardware specifically designed for this type of engine like they are for polygons.
-
And yet you continue to ignore people in this thread like daVinci who know infinitely more than you and easily as much as this Dell character. You continue to bring up a point that isn't even part of this thread. No one here said it was fake.
Show me who their big million dollar investors are and if there is one well known tech company I'll be impressed.
Lastly, CPUs are no slouches these days, so I'm not sure how you come to the conclusion that moving this render engine to the GPU is somehow going to make a night and day difference. GPUs are great for certain tasks and they might in fact be suited to accelerating this voxel based system, however that still doesn't address other issues like MASSIVE datasets, collision, animation, etc.
So, buy into the hype all you want, but understand that you aren't posting in a forum with a bunch of rubes, there are people here that work in the industry and know what they are talking about.
-
-
-
-
-
wow, a few white knights for these guys tonight.
they've been showing this engine off for eight years now. millions of dollars have been poured into. yet they haven't been able to put out an actual game, or even a quick little demo that we can freely download off the net. we don't know what smoke and mirrors they had going on that laptop, so the videos pretty useless.
they need to actually produce something, and until then, it's pointless dribble. -
-
Way to miss the point. In a REAL game scenario you wouldn't have cut and paste instances making up 90% of the content. That alone calls the whole thing into question. No one is saying it's not possible to have that level of detail, people are saying that actual games have much different requirements.
-
-
I'd never dispute that. However, many of what has been said makes everything plausible, in my eyes. Their lack of artists, lack of staff, wanting to work on the major problems rather than spend time making a demo to appease skeptics that have no bearing on those that have invested in their company. They've said they're going to release a trailer showing collision, lighting, and animations. I'm willing to entertain they'll do this, willing to entertain they'll solve most of everyone's concerns.
I mean, people said the demo was fake, then they show otherwise, now they're saying it's impractical, when, in fact, even Carmak said it's something that'll be ready in the future. Carmak, and others, didn't spend years of their lives on this problem, if they had they might've solved it and be demoing a demo that wouldn't leave these gaping wholes which fill with doubt. They had other priorities which prevented that. These guys did put years into the issues so I find it plausible they've found ways to deal with a lot of the problems Carmak, of all people, said wouldn't be a problem in 5yrs.
What did Carmak make his 5yr estimate on? As much as he know's, hell, exists in the world of graphics computing has he put nearly as much effort into this type of engine as Dell and his fellows? The fact is, Carmak say's what Dell is claiming to have done will be done in time. So, it stands to reason that perhaps, perhaps Dell and his fellows have found the solutions to these problems or are at the point where they know how to solve them having put years into building this engine. Perhaps I'm wrong though. It's possible that there have been quite a few others who spent years and year tinkering on this problem, full time. The thing is, I've only seen maybe, 2 other videos for engines that do anything near as complex as what they've done, and that makes it seem that the only people at the forefront of this technology are the ones who everyone is calling BS on and that makes me wonder how much can they actually know about a problem they don't have nearly as much information about.
-
-
-
That engine is running only on a processor. A software only engine. I reason, if my 2500k@4.2 can only do one WU in F@H while my gtx570 can do 4.5 in the same time that's a lot of increased performance to gain from getting that engine to utilize a gpu. Given, I have no idea how exactly F@H utilizes a gpu, but I'm willing to bet that this Atom engine could at least match it or surpass. Also, it's worth noting that Dell has claimed they've had pixel based objects, such as a rocket, interact with objects rendered with Atoms, such as tree leaves. A high claim to be sure, but one you have to take at face value if you're going to fully consider how viable their engine might be.
-
-
Considering that's all anyone can really do given how little information anyone has about this engine the arguments can really just go around in circles. Someone points out how little information they give, say that's suspect, that's sound reasoning, likewise, someone saying that they're understaffed and too busy working on their engine and with investors to bother fully informing an inconsequential public would also
be making a valid point. At this point in time no one has enough information to make any significant conclusion.
I mean, really. All anyone has been doing is grasping at straws. It's moot to point this out here. I mean, you guys are arguing against an engine from a technical standpoint when you have absolutely no technical details about the engine. In fact, all your technical arguments can really just be viewed simply as questions.-
-
I suppose I have been acting that way. I have to admit though it's kind of hard not to. I mean, everyone called BS when the trailer was first released. They were saying it was pre-rendered. That they thought it was pre-rendered, when it was not, goes to show that to some degree it can be viewed, correctly, as 'dark magic'.
It's funny though. So many were calling it pre-rendered, they were proved wrong, and now they pretend like they understand what they didn't believe was real. What do I know though, Perhaps there were quite a few out there that did think it was real upon first sight and only voiced the concerns many now voice. All I know is that they were drowned out, and one of the most vocal of opponents, Notch, used terrible comparisons to this engine, so much so that even the layman couldn't take him seriously.
What stands out to me is that while you're claiming that a lot of people understand voxel rendering I do not see, anywhere, any example that comes anywhere close to what this engine has done, in the manner it seems to do it, and has not gotten anywhere near as much attention, or discussion, not simply due to the claims, but due to the actual demonstration of the technology. It just seems preposterous that so many can fully understand how this engine achieves what it achieves, and it's limitations, when no one else, anywhere, has done what they've done.
-
-
-
-
-
-
-
What they are showing looks very static to me. What about particle effects, fluids, animations? The only use I can see for this technology (when its finally done) is for medical applications doing remote minimal invasive surgery.
They call it unlimited detail, but I am missing many details, because all the grass, stones, palm trees etc. look the same. Trees should have different hights, Stones usually are all completely different, especially in n ature not one leaf looks exactly like the other. What I am seeing here is "unlimited detail on repetetive single objects" not an unlimited detail world, which is what I would expect from this engine.
-
-
-
-
-
I'm nominating this thread for highest number of first time posters that wasn't about Firefall.
Seriously, where did the crazy come from.
qsylvr (has pulse)
shmargin (invested heavily in blast processing)
Tonez (to be fair his first post was a bible quote in some other shitty thread but his heart is with this one)
fw_rams (only one who 'gets it')
Yumcheese (still living in 91 before wireless)
Alexrose (claims to be a physicist)
cixelsydresol (reads blogs about Minecraft)
-
-
-
-
You mean: http://shackapp.com/shackstats/ ?
-
-
-
-
-
Yep and this "Dell" guy knows very well he could never answer to the questions of those with an in-depth knowledge of gaming graphics techniques, only to the layman who is easily wowed by "unlimited detail!". Meanwhile his videos continues to show models pointing in the same direction without any varying geometry, no collisions, no demonstration of dynamic lighting and no animations.
-
-
A lot of people here saying "But you COULD make a game with that engine!" here. You guys are really missing the point; yes you COULD do it, but the point is that out of what we've seen the problems vastly outweigh the benefits of using their engine. While you may get a good bump in small details compared to traditional polygonal rendering, you will also lose an awful lot. Both dynamic lighting and animation haven't been proven to work (Even if they say they have, they've never demonstrated them), and we have no idea what the size of the data is, as well as suspicions that they're working off an incredibly small data-set with copy/pasted objects everywhere. Destruction is probably impractical with their system as well, as from everything I've read they do a hefty amount of pre-processing which means that the environment both is and will remain static.
So yes, you could make a game out of what they have, but you'll be working on an unproven technology that's nowhere near as maleable as current systems, which is both expensive and difficult for a single gain that next to no-one will care about at all. Unless you really, really want to look at each individual flat-shaded, non-dynamic grain of dirt.