[QUOTE=Muscar v2;25409185]The problem with this is that it's just static, almost impossible to animate and even more impossible to get any physics with.
And simple things as collision detection is very hard.[/QUOTE]
They actually made a small animation demo, it's on youtube somewhere.
[editline]15th October 2010[/editline]
Damn I should of read the thread more, FPtje already posted it.
[QUOTE=FPtje;25416465][media]http://www.youtube.com/watch?v=cF8A4bsfKH8[/media]
Here's an example of a flying bird. I don't know how hard they tried to get this done.
Media tags don't leave me alone.[/QUOTE]
What they do is make one model for every frame, then load them in a sequence.
Think of how much what would take on your computer with 10-20 way longer animations going on all the time. It's not very efficient.
the art and colors remind me of that schizophrenic guy who painted cats weirdly when he got insane
[img]http://seancasio.files.wordpress.com/2009/01/louis-wain2.jpg[/img]
it's just very weird
[QUOTE=Muscar v2;25417237]What they do is make one model for every frame, then load them in a sequence.
Think of how much what would take on your computer with 10-20 way longer animations going on all the time. It's not very efficient.[/QUOTE]
And how do you know this?
You don't.
That map looks hideous. Mainly because of the constant tree spamming.
[QUOTE=SoaringScout;25417572]That map looks hideous. Mainly because of the constant tree spamming.[/QUOTE]
Spamming is what makes this impressive. You shouldn't look at the aesthetics of the artwork. You should rather look at the capabilities.
[QUOTE=Muscar v2;25417237]What they do is make one model for every frame, then load them in a sequence.
Think of how much what would take on your computer with 10-20 way longer animations going on all the time. It's not very efficient.[/QUOTE]
You can apply the idea of "showing only what you need to show" to animation too.
Imagine having a cube with 8 points. Now imagine deforming the cube in an animation, the only thing you need to do is move one or more dots a certain distance in a certain amount of time and you've got an animation.
That's simple model deformation for you. This can go further by adding/removing dots to animate.
Want to speed things up? Only animate the things that are visible on the screen.
Also, OP video shows an animated bug walking around (1:44) it doesn't look like it's made frame by frame.
Do we really need games to get this detailed ?
[QUOTE=Hanibal;25417741]Do we really need games to get this detailed ?[/QUOTE]
Of course. If people didn't want more, we wouldn't be here in the first place.
[QUOTE=Hesychasmos;25409445]This is true.
If they made a game using this technology that was all about destroying a mountain, I would probably play it nonstop for weeks.[/QUOTE]
[u]Blow Shit Up: The Video Game[/u]
[url]http://www.3d-test.com/interviews/unlimited_detail_technology_1.htm[/url]
Interview.
[quote]Q7 What are your plans about Unlimited Technology?
A7 We didn’t expect the resentful media wave, we thought that we could just hide and appear from no where with something brilliant, but unfortunately we got caught working quietly in our corner, and so now we have a lot of people saying “that doesn’t look good enough yet !”, but on our end we feel like a mother who put cookies in the oven and little children are pulling them out and saying “they don’t taste right !” a bit of patience and the cookies will be fine.
The planned form of the SDK is as follows:
IMPORT OF DATA:
Data can be imported from a variety of sources. We expect to directly support some popular scanning and modelling applications and file formats for both point cloud and polygon style data. The SDK will provide a simple way to import data from any source.
ENGINE:
The Unlimited Detail Engine displays 3D models or groups of models which may be positioned, rotated, scaled and viewed from any direction and with any projection. The technique allows geometry of unlimited complexity to be displayed at interactive frame rates without the need for 3D acceleration hardware.
ANIMATION:
Unlimited Detail supports Animation. (Please see attached video clip of an articulated bird. And remember this is a simple first attempt for R&D purposes so it doesn’t look “polished”)
[url]http://www.mediafire.com/file/ndumvromniq/bird_articulated.wmv[/url]
[b]In addition to animation, we will be supporting dynamic and destructible environments.[/b]
MEMORY STORAGE
The Unlimited Detail rendering technology does not require significant system memory to run. The geometry may be compressed, instanced and streamed to allow the best mix of runtime and disk based storage as well as support unique geometry.
SHADOWS:
[b]Unlimited Detail will have both real-time shadows[/b] and pre-calculated lighting as well as a variety of advanced lighting options.
DEMONSTRATIONS AND DOCUMENTATION:
Demonstrations will be available for a great variety of industries, mining, architecture, science, medical, games, and movies/entertainment. In all cases these demonstrations will use industry specific data and will be downloadable from the website.
PRICE:
Pricing will be chosen in such a way that this technology can be accessed by all, including those in developing nations were graphics cards are not commonplace and software based 3D graphics are vital to industry purposes.
MAILING LIST AND WEBSITE:
Nearer to completion we will be releasing short newsletters to inform interested parties of our progress, we will also have a new website in a few months as there have been comments about the inadequacy of the current website, whilst we do agree with the comments please understand we are primarily creators of technology and the recent media wave was unexpected. Please expect a more commercial incarnation of our public image soon enough.
Downloadable Demos will be available before the SDK release date
[/quote]
This is another one of those things where someone goes "Look at us we hvae a ground breaking future technology it can do all sorts of shit!" then a year later you ask someone about it and they just say "What?"
And, The scene is the video lags like hell, on probably one of the newest machines considering they are developers for graphics tech.... How will this run on my 2GB RAM and 9600GSO without being 1 FPS? Not everyone is going to have the latest and newest tech, and this looks like it requires alot of storage and processing capacity, even if they guy in the vid says not, the vid proves otherwise.
Watched this a while ago. Two fundamental questions remain unanswered:
1. Storage. Storing billions of points, even reduced to smallest thinkable form, would make anything beyond the smallest of models consume your entire RAM. Someone did the math above. Unless I misheard the demos run at a resolution of 1024x768 or lower, and the detail of the point cloud is conformed with this. Imagine then the amount of memory you'd need to render a point cloud in HD resolution?
2. Efficiency. If you can somehow cope with models containing billions of points, imagine what kind of time it would take to read it off secondary storage? They also mention they've developed a search algorithm used to find the points in space that need to be displayed - How does this algorithm scale, may I ask? The most effective search algorithms known has a Big-Oh notation of O(logn), where n is the number of elements, and this applies to a set of elements already sorted in a tree-structure. For a set of a billions of entities even this would be too slow. Have they somehow invented something better than what geniuses all over the world has been thinking about since before the invention of computers, and what made Google rich?
Don't we already have Normal Maps to simulate more detail in an object without upping the memory use so much?
And with polygons, all you need is to increase the amount of memory and processing speed to increase the amount of triangles and make everything look rounder.
The examples he took from HL2 were pretty lame, that horse statue near the end of the game is pretty darn round and detailed.
I say we go with polies.
[QUOTE=latin_geek;25418726]Don't we already have Normal Maps to simulate more detail in an object without upping the memory use so much?
And with polygons, all you need is to increase the amount of memory and processing speed to increase the amount of triangles and make everything look rounder.
The examples he took from HL2 were pretty lame, that horse statue near the end of the game is pretty darn round and detailed.
I say we go with polies.[/QUOTE]
Did you even watch the video...
[editline]15th October 2010[/editline]
[QUOTE=Im Crimson;25418095]Watched this a while ago. Two fundamental questions remain unanswered:
1. Storage. Storing billions of points, even reduced to smallest thinkable form, would make anything beyond the smallest of models consume your entire RAM. Someone did the math above. Unless I misheard the demos run at a resolution of 1024x768 or lower, and the detail of the point cloud is conformed with this. Imagine then the amount of memory you'd need to render a point cloud in HD resolution?
2. Efficiency. If you can somehow cope with models containing billions of points, imagine what kind of time it would take to read it off secondary storage? They also mention they've developed a search algorithm used to find the points in space that need to be displayed - How does this algorithm scale, may I ask? The most effective search algorithms known has a Big-Oh notation of O(logn), where n is the number of elements, and this applies to a set of elements already sorted in a tree-structure. For a set of a billions of entities even this would be too slow. Have they somehow invented something better than what geniuses all over the world has been thinking about since before the invention of computers, and what made Google rich?[/QUOTE]
Read the interview I posted a page back.
[QUOTE=Jimmg;25418730]Did you even watch the video...[/QUOTE]
Yes i did.
[QUOTE=Jimmg;25418730]
Read the interview I posted a page back.[/QUOTE]
They don't really offer any answers. On the storage issue they say "the data is compressed, and is read in its uncompressed form" - Okay, but how the fuck does that work in practice? They say that point cloud data uses up only 8% of the memory required by the equivalent polygon-based geometry, but if their models are gonna be X thousand times more detailed as claimed (which probably be well needed to overcome the vintage-voxel look) those would be a massive 8%. I couldn't find any technical details or explanations to how compression/decompression of such massive data amounts will happen in real-time, or what reading uncompressed data directly is about (which sounds really strange; if data is readable loss-less in its compressed form what was its original form like?). Nowhere do they give anything resembling details on their point-super-search algorithm that is apparently faster than anything ever seen, nor do they mention how well this system scales (for instance, what happens if we bump up the resolution?).
Still, they have my benefit of doubt. Though, that narrator in the videos sound remarkably smug and confident for being caught by media attention in what he claims to be infant stages of development.
[QUOTE=Jimmg;25417939][url]http://www.3d-test.com/interviews/unlimited_detail_technology_1.htm[/url]
Interview.[/QUOTE]
I'd love for them to be telling the truth as much as the next guy, if they have defeated the many issues of voxels then congrats to them. But these people are also really bad at admitting their engine has anything wrong with it at all, the guy spends a good chunk of the ending of the second part of the demonstration trying to let us now the the shadow map only looks bad because the map was rushed, and how it's completely not the fault of anything in the engine, that would be all well and good if it only took him the 5 seconds that it needed, but he spends a lot more time than that trying to defend it.
These guys aren't really going into technique at all either. They can't just keep saying it works, and not explain how, if they're compressing thousands of voxels to be 8 percent of what a polygon would take up ( what about a 9 byte triangle that can be any size, I find it highly improbable they can beat that kind of filesize ), and why couldn't we use the same insane compression techniques on polygons to make their footprint that much smaller, likely being smaller than the voxels when they are also compressed? And moving even further into the argument, constantly picking out data from a compressed zip and decompressing all the data realtime cannot be good for drawing and load times.
Honestly the whole thing seems really neat, but extremely fishy and evasive. Their developers like to pretend certain issues don't exist and are known for deleting public questions they can't answer and ignoring people, according to more than one comment on their youtube pages.
[QUOTE=lifehole;25418087]And, The scene is the video lags like hell, on probably one of the newest machines considering they are developers for graphics tech....[/QUOTE]
They were running software mode, btw.
If you've ever used third-party shaders to run TF2 on software alone, it lags like fuck. Literally, 2fps is a high.
[QUOTE=lifehole;25418087]This is another one of those things where someone goes "Look at us we hvae a ground breaking future technology it can do all sorts of shit!" then a year later you ask someone about it and they just say "What?"
And, The scene is the video lags like hell, on probably one of the newest machines considering they are developers for graphics tech.... How will this run on my 2GB RAM and 9600GSO without being 1 FPS? Not everyone is going to have the latest and newest tech, and this looks like it requires alot of storage and processing capacity, even if they guy in the vid says not, the vid proves otherwise.[/QUOTE]
The guy said they were running it in software mode, not in hardware mode. in one of the other videos.
[QUOTE=Im Crimson;25418095]Watched this a while ago. Two fundamental questions remain unanswered:
1. Storage. Storing billions of points, even reduced to smallest thinkable form, would make anything beyond the smallest of models consume your entire RAM. Someone did the math above. Unless I misheard the demos run at a resolution of 1024x768 or lower, and the detail of the point cloud is conformed with this. Imagine then the amount of memory you'd need to render a point cloud in HD resolution?
[/QUOTE]
1024x768 is almost an HD resolution.
Wow that guy's voice is annoying, and he explains everything really shittily. The concept is cool, but I'll have to see another, better demo before I can judge it.
Bruce still uses IE6.
:frog:
[QUOTE=Generic.Monk;25429920]1024x768 is almost an HD resolution.
Wow that guy's voice is annoying, and he explains everything really shittily. The concept is cool, but I'll have to see another, better demo before I can judge it.[/QUOTE]
... What? Full HD, being 1080 progressive, means a resolution of 1920×1080. This is the kind of resolution gamers of tomorrow (and many of today) will expect to play in. As a matter of fact 1280x1024 is viewed by many as the minimum satisfactory. 1024x768 was what most ran Half-Life at back in 1998-99.
My monitor is only 1366 x 768 :saddowns:
I remember this, his website was located at a Van in Australia.
[QUOTE=Im Crimson;25430349]... What? Full HD, being 1080 progressive, means a resolution of 1920×1080. This is the kind of resolution gamers of tomorrow (and many of today) will expect to play in. As a matter of fact 1280x1024 is viewed by many as the minimum satisfactory. 1024x768 was what most ran Half-Life at back in 1998-99.[/QUOTE]
1280x768 is also an HD resolution.
[QUOTE=Onyx3173;25430969]1280x768 is also an HD resolution.[/QUOTE]
Way to see the trees but not the forest, people :v:
Right my mistake, "HD" can imply a wide range of resolutions. "Full HD" is 1920×1080, nothing less. All this mainly applies to TV screens however, most games of today are expected to run on significantly higher resolutions than what's displayed in the demo (mind the 28 FPS, too).
[QUOTE=Onyx3173;25430969]1280x768 is also an HD resolution.[/QUOTE]
:smug:
[QUOTE=Im Crimson;25439399]Way to see the trees but not the forest, people :v:
Right my mistake, "HD" can imply a wide range of resolutions. "Full HD" is 1920×1080, nothing less. All this mainly applies to TV screens however, most games of today are expected to run on significantly higher resolutions than what's displayed in the demo (mind the 28 FPS, too).[/QUOTE]
Or you just missed the point where you were trying to make yourself look better than Generic.Monk. Generic.Monk never said that 1280x768 is the only HD resolution, only that it was [B]a[/B] HD resolution. And just because many of us prefer to play at 1920x1080 or higher does not mean everyone does.
[url]http://store.steampowered.com/hwsurvey[/url]
16% of users are still at 1280x1024 with the only resolution having a higher percentage of users being 1680x1050 with 17.7%.[B][/B]
A couple of things that make me not like this.
- The narrator in these videos sounds very unprofessional and condescending. Also his voice sounds annoying and he really didn't go into much detail at all.
- I REALLY hated how they showed the way LOD works on polygon models. Obviously if you noclip into a open area you would never get to in-game and moved into the playable area OF COURSE the LOD is going to look bad. Proper LOD is invisible and works really well.
- This whole system sounds like a waste of memory. Why use billions of points to make a flat wall when a two triangle polygon can do the same job?
- I just don't see how this system has any advantage over normal mapped polygons. I understand this whole "Search Algorithm" is supposed to work but im not buying it until i run a demo of it myself.
And you are not an investor so you don't have to...
Sorry, you need to Log In to post a reply to this thread.
