Unlimited Detail Real-Time Rendering Technology

  • Thread starter Thread starter Raiser
  • Start date Start date
  • Views Views 4,419
  • Replies Replies 36
Slyakin said:
gamerjr said:
http://notch.tumblr.com/ Ill beleive the guy that knows more about Voxels
Welp, he went into detail about my complaint as well as a fuckton of other things.

Sucks that this is a scam.
Yeah, I just found this as well. If what Notch says is true which could more or less be likely.. sucks.

EDIT: Euclideon Creator Swears Infinite Detail is "Not a Hoax"

Warning: Wall of text
Yesterday we posted a video from Euclideon – a Australian company that claims it can revolutionise video game graphics, increasing visual fidelity by 100,000. This morning we spoke to Euclideon's CEO Bruce Dell – the man Markus Persson calls a "Snake Oil Salesman" – to ask a few questions regarding Euclideon's ‘Infinite Detail' technology.

"I think what I would like to make clear is that this is not the finished product," says Bruce Dell, CEO of Euclideon. "We feel like a mother who put cookies in the oven, and now everyone is surrounding the oven chanting ‘are they ready yet? Are they ready yet?'

"Give us time and the cookies will taste just fine!"

Instantly we recognise the voice - it's the voice from that video. The voice that claimed Euclideon could revolutionise video game graphics, the voice that claimed a new technology called ‘Infinite Detail' could increase visual fidelity by a factor of 100,000. The man Markus ‘Notch' Persson, the creator of Minecraft, openly called a "Snake Oil Salesman".

It's 9am in Brisbane, and we've just woken said Snake Oil Salesman up.

"No! No, this isn't a hoax," Bruce Dell laughs, in response to our first, obvious question. "If this was a hoax then we've convinced the Australian government it was a hoax. We've convinced our board of directors and investors it's a hoax!
"We have a government grant – so no, it is not a hoax! We have real time demonstrations."

The response to Euclideon's demonstration video, which we posted yesterday was instantaneous and fairly mixed. Some were cynical, some called it a hoax, others were more receptive – but it was hardly a convincing demonstration. Markus Persson, writing on his own personal blog, was perhaps the most scathing in his criticism.

"They're hyping this as something new and revolutionary because they want funding," wrote Persson. "It's a scam."

But if it's a scam, then the Australian Government is the mark, having invested 2 million dollars into Euclideon and its technology.


LOOKING FOR SNOW WHITE

We asked Bruce to explain the technology and how it worked.

"Well, basically anyone who is technical is going to say you can't run that many polygons," he began, "but in the past we were trying to explain it in simple terms so people could understand.

"A good analogy would be this: imagine you go to a library to find a book - say… Snow White. Imagine you go to a library and those books aren't on the shelf; they're all lying on the ground. At the moment systems that run point cloud data are doing that, they're putting every point on the screen and there is no order to it. Now imagine you go to a library and all the books are on the shelf and in order – you go to the ‘S' Section, then look for ‘SNO' and it isn't long before you've found the book you need.

"One system is looking at thousands of books," he continues, "and the other system is looking at ten labels. That's the basis of a search algorithm like Google or Yahoo – they sort through all the knowledge in the world really quickly because it's categorised.

"We made a search algorithm, but it's a search algorithm that that finds points, so it can quickly grab just one atom for every point on the screen."

According to Bruce Dell, it's all about efficiency.

"So think about the difference," he says. "If you had all of the points you are seeing on the screen, like in our demo, it's going to take forever. You'll be waiting for a long time. But if you're grabbing only one for every pixel on the screen, then you don't have a trillion dots, you have… well, pick a resolution and do the maths!

"That's the difference. In layman's terms that's how we're doing what we're doing. The workload is so small that at the moment we're running software just fine with real time demonstrations and we're still optimising, because we keep finding more efficient ways to do this."

That appears to be all well and good, but most criticism from the games industry has come from the detail Euclideon has been a little more coy on: animation, physics …

"[V]oxels are horrible for doing animation," wrote Markus Persson in his aforementioned blog, "because there is no current fast algorithms for deforming a voxel cloud based on a skeletal mesh, and if you do keyframe animation, you end up with a LOT of data. It's possible to rotate, scale and translate individual chunks of voxel data to do simple animation (imagine one chunk for the upper arm, one for the lower, one for the torso, and so on), but it's not going to look as nice as polygon based animated characters do."

According to Bruce Dell, the reason no animations have been shown is simple – Infinite Detail is still a work in progress.

"We have animation," claims Bruce, confidently. "We're certainly going to do a lot more work in that area. I have faith that you'll find our animation quite satisfactory, but we have no intention of releasing anything in that department until it looks absolutely 100% because if we release it now, I assure you that no-one will take it as ‘that's where we're up to and we're still working on it', they'll just scream ‘it's not perfect yet! They can't make it perfect! This can't compare to polygons!'"

THE EMPIRE STRIKES BACK

We spoke to an Australian physics engine developer with experience of Bruce Dell and Euclideon. His company dealt with Bruce Dell years ago, when Euclideon was seeking funding for the Infinite Detail project. Said company declined to fund the project, citing issues with memory management, particularly when it came to animations.

According to him any live demonstrations given by Euclideon featured poor art and assets, so it was difficult to gauge precisely how hardware intensive Infinite Detail actually was.

The developer in question asked not to be named, but his primary concern wasn't with the ‘Infinite Detail' tech itself, which he claimed could work with adjustments – the issue was the toolset and the investments required to move an entire industry across to a new standard. Currently every game developer in the world is using tools dedicated to polygons – convincing an entire industry to toss years of investment and research would be a difficult task indeed, especially with an unproven technology.

Bruce Dell disagrees with that assertion.

"I see comments from people saying the games industry will never use this," he begins. "Well, this industry isn't quite so old and stubborn. The games industry is actually quite open and we're in contact with quite a lot of players in that industry."

According to Bruce, the sheer efficiency of his technology will win developers over.

"The present polygon system has got quite a few problems, but not in terms of graphics. Polygons are not really scalable between platforms – if I were to make a character on a PlayStation 3, I can't put him on the Nintendo Wii because he uses too many polygons, so I have to completely rebuild him. Imagine we weren't doing a polygon game, say we were doing a 2D game, if I drew a character on the PlayStation, he's just a bitmap image – this can easily be rescaled. You could do it in Microsoft Paint! ‘Infinite Detail' data is like a 2D bitmap image in that rescaling its size is easy, whereas polygons can't scale like that.

"The big thing is – if you make a game using the present polygon system, you have to rebuild it to rescale it. You don't have to do that with Unlimited Detail.

"The industry's response was, basically, what you have is really good, you do not understand that the industry is used to using polygons and our tools are very good. I took a look at those tools and thought yes, they are very good. We want to get things to the stage where the artists don't have to change anything, just that now they're using unlimited detail."

Not all developers have openly dismissed Bruce Dell and his ‘Infinite Detail' technology, but even the most optimistic have opted for a ‘wait and see' approach. John Carmack, for example, mentioned Euclideon briefly on his Twitter account claiming that "production issues would be challenging" but wondered if the tech might viable "a couple of years from now".

Even Bruce Dell himself admits that he needs time. Come back later, he says, perhaps sooner than we think, and we might get the final product.

"Basically we're in the middle of a trilogy and this is like our Empire Strikes Back," he explains. "We disappeared for so long that I think everyone thought ‘oh, they're dead'. So we thought we'll release a one year report, tell everyone we're alive and then disappear again.

"The intention is to come out again, once we've finished, and then we'll be releasing real time demonstrations."[/p]
icon11.gif
Source
 
BortzANATOR said:
No it wont.
QUOTE said:
Another weakness is that voxels are horrible for doing animation, because there is no current fast algorithms for deforming a voxel cloud based on a skeletal mesh, and if you do keyframe animation, you end up with a LOT of data. It’s possible to rotate, scale and translate individual chunks of voxel data to do simple animation (imagine one chunk for the upper arm, one for the lower, one for the torso, and so on), but it’s not going to look as nice as polygon based animated characters do.

Carry on quoting people rather than doing research and you'll go far in life.

It's only natural that someone who didn't manage to animate voxels properly is going to bitch and complain that it's impossible.

[youtube]Tl6PE_n6zTk[/youtube]

Blocks of voxels connected with bones - problem solved.

By the way, this was done by *1* programmer, for a bloody thesis work and is presented on a "shot-on-shitty-o" computer. Think of what a company could do to improve it.
 
Herf derf i did my research and i know how polygonal modeling works and how voxels work as well. I have spent far too many hours in Maya.
There are FAR too many problems with voxels RIGHT NOW. Im sure that in a few years after this greasy con man gets money for this, that it will eventually gain enough funding to get this popularized. Its easier right now to use polygons cause thats what we are used to. voxels present a glorious freedom from the restrictions of polycounts on rectangular or triangular models, but have fun painting every individual weight on every square inch of the model. Id like to see the weights, the bones, and the individual hierarchy.
And his model looks decent. Ive seen better polygons.
 
One of my main concerns for this is texture mapping. Yes with voxels at this level of detail you don't need bump maps, but what about specular maps and such?

And as stated by Notch (though not directly), this is likely to require way too much RAM to be feasible for a while. Notch's estimate assumes the entire island will be rendered in each frame which is not true (in game design you split your world up into chunks and the camera determines what's rendered and what's not), only a portion would be rendered at a time... but even a small portion of that is still going to be Effing Huge™.
 
Foxi4 said:
By the way, this was done by *1* programmer, for a bloody thesis work and is presented on a "shot-on-shitty-o" computer. Think of what a company could do to improve it.
One programmer using CUDA. It's heavily optimized for a certain amount of CUDA cores; it won't work on anything else than nVidia graphics cards.
A company can only improve the use of CUDA in the programming, but that will not go on infinitely; you will hit a limit after optimizing for a long while.
 
I have a few problems with this in general

1) Comes from a company that has a bit of a shady track record
2) Makes outrageous claims
3) Doesn't provide research papers detailing their methodologies

If you're that confident in your thing, go to SIGGRAPH and do a talk on your topic. If not, then I'm not going to be interested in your voxel technology otherwise.
 
Bumps maps are quick and dirty. They tae mimimal space and do a decent enough job.

I mean after messing with some Halo game models i was suprised at how little geometry detail there was. Almost all the detail comes from bump maping.
 
Berthenk said:
Foxi4 said:
By the way, this was done by *1* programmer, for a bloody thesis work and is presented on a "shot-on-shitty-o" computer. Think of what a company could do to improve it.
One programmer using CUDA. It's heavily optimized for a certain amount of CUDA cores; it won't work on anything else than nVidia graphics cards.
A company can only improve the use of CUDA in the programming, but that will not go on infinitely; you will hit a limit after optimizing for a long while.

Oh, that's really horrible that AMD didn't prepare a nice enough SDK to be even narrowly interesting to homebrew developers. I'm sure the creator of probably the most revolutionary technique as far as animating voxels is concerned (Well, he's not exactly animating the voxels, just attaching them to an animated skeleton to preserve resources, but the effect is about the same...) is truly sorry for the inconvenience.

...and seriously, don't be silly. AMD is going to release a CUDA equivalent soon enough, unless they want to loose the arms race. Also, it's going to work on GFX cards other than nVidia aswell, at the end of the day, CUDA cores are just... cores, simple as that. Takes some optimizing, not a mirracle.
 
Foxi4 said:
BortzANATOR said:
No it wont.
QUOTE said:
Another weakness is that voxels are horrible for doing animation, because there is no current fast algorithms for deforming a voxel cloud based on a skeletal mesh, and if you do keyframe animation, you end up with a LOT of data. It’s possible to rotate, scale and translate individual chunks of voxel data to do simple animation (imagine one chunk for the upper arm, one for the lower, one for the torso, and so on), but it’s not going to look as nice as polygon based animated characters do.

Carry on quoting people rather than doing research and you'll go far in life.

It's only natural that someone who didn't manage to animate voxels properly is going to bitch and complain that it's impossible.

Blocks of voxels connected with bones - problem solved.

By the way, this was done by *1* programmer, for a bloody thesis work and is presented on a "shot-on-shitty-o" computer. Think of what a company could do to improve it.
Notch recently acknowledged he was wrong. He was unaware of this development as it is relatively new (he's not really working on voxel animation).

I'm with Notch on this. He sounds like he knows what he's talking about espcially with the size constraints if using non-repeated geometry. Even loading a chunk of that map into RAM would take HUGE data amounts. As well as that we still need to store the whole island so we'll need to have XXXTB of data storage on hand (I don't know how compressed these point clouds are myself when saved to disk though so may be wrong)
 
notch could be right, but I don't really like hearing what he says.
Coming from a guy who can't update his own game without breaking it most of the time, a guy who literally had more vacation days than work days in a year (You can go back and check yourself and this is only from publicly mentioned vacation days.), and clearly doesn't have all the facts on hand about the subject matter, I just really don't want to hear it from him.

From anyone else, sure.
Ideally from other people who are looking into voxels and such.

I also find it funny how he'll bash on these guys for their voxel work and then immediately praise/promote other people who are doing voxel work as well.
 
Jamstruth said:
As well as that we still need to store the whole island so we'll need to have XXXTB of data storage on hand (I don't know how compressed these point clouds are myself when saved to disk though so may be wrong)
Considering Notch's post:
Notch said:
They made a voxel renderer, probably based on sparse voxel octrees. That’s cool and all, but.. To quote the video, the island in the video is one km^2. Let’s assume a modest island height of just eight meters, and we end up with 0.008 km^3. At 64 atoms per cubic millimeter (four per millimeter), that is a total of 512 000 000 000 000 000 atoms. If each voxel is made up of one byte of data, that is a total of 512 petabytes of information, or about 170 000 three-terrabyte harddrives full of information. In reality, you will need way more than just one byte of data per voxel to do colors and lighting, and the island is probably way taller than just eight meters, so that estimate is very optimistic.
And then this, in his second post:
Notch
* I was trying to show that there was no way there was that much UNIQUE data in the world, and that everything had to be made up of repeated chunks.
* One byte per voxel is way lower than the raw data you’d need. In reality, you’d probably want to track at least 24 bits of color and eight bits of normal vector data per voxel. That’s four times as much data. It’s quite possible you’d want to track even more data.
* If the data compresses down to 1%, it would still be 1 700 three-terrabyte hard drives of data at one byte of raw data per voxel.
No way in hell that you're going to store that much data. Compression can only go a little way before you start running into problems/limits. Also, you don't want km^2 of repetition, it's ugly. Very ugly.
 
Zarcon said:
I also find it funny how he'll bash on these guys for their voxel work and then immediately praise/promote other people who are doing voxel work as well.
His problem is that they're touting it as something new, something amazing. Something with no drawbacks when, in fact, its NOT new, and it has some MAJOR drawbacks.
 
I dont understand how they can call it "Unlimited Detail" if it takes so much storage space?
Yeah it renders much quicker, but you need 3 tb harddrives to play one level.
 

Site & Scene News

Popular threads in this forum