View Full Version : Unlimited Detail Technology - What do you think about that?

04-17-2010, 01:55 AM


is that big wow???

04-17-2010, 02:01 AM
"here we have flat shapes" hmmm lost me @ at that point

04-17-2010, 02:02 AM
"here we have flat shapes" hmmm lost me @ that point

04-17-2010, 02:29 AM
grrrrrrr Forum went crazy !

04-17-2010, 02:29 AM
from the demo's it looks like a load of rubbish to me...all talk and no action/quality

and the music in the bg is SO SO SO annoying...almost as much as his dialogue.
he's also a "vague -on"

04-17-2010, 02:35 AM

04-17-2010, 02:40 AM
more like "unlimited B.S" :D

note...no animation, no bones, no deformers, no particles...no "wow"
i do 'get' it's w.i.p. but to impress you have to be impressive...

04-17-2010, 02:47 AM
have u seen CORE? hahaha :D

04-17-2010, 02:51 AM
have u seen CORE? hahaha :D

*chuckle* that was deep REAL deep !

04-17-2010, 03:24 AM
have u seen CORE? hahaha :D


04-17-2010, 03:50 AM
I would say, if what he talks about is acctually working, i shall kneel down and take my hat off. Finaly someone who realises that polygons suck. Ive had the thought my self but im no coder. Why draw alot f stuf that you dont see. All you acctually need at any given time, is the amout of data acctually shown on screen. Therest is unessesary. If you have a shape of 3 million or so polys and zoom out, why would you then need to comput 3million polys that only makes up 300*300 pixels. (Thats only 90000 pixels)

I have no idea if its acctually a software in development, but if it is ill sheer them on, and if I could and was rich i would sponsor them.

Oh, and bare in mind that if they are developing it, its a new technology. Did any one of you see Krafwerks c64 flatshaded 3d. It takes time to go from nothing to something, and if "nothing" is what we see there i cant wait to see what "somthing" is.

04-17-2010, 06:03 AM
its raytracing into a sparse voxel tree. the technique is nothing new nor revolutionary. he is deliberately being extremely vague to avoid that fact, as well as the limitations and issues it poses.

i believe john carmack also talked about this technology a few years ago in relation to tech5 or 6.

pc perspective (http://www.pcper.com/article.php?aid=532)

There is a specific format I have done some research on that I am starting to ramp back up on for some proof of concept work for next generation technologies. It involves ray tracing into a sparse voxel octree which is essentially a geometric evolution of the mega-texture technologies that were doing today for uniquely texturing entire worlds. Its clear that what we want to do in the following generation is have unique geometry down to the equivalent of the texel across everything. There are different approaches that you could wind up and try to get that done that would involve tessellation and different levels of triangle meshes and you can could conceivably make something like that work but rasterization architecture does really start falling apart when your typical triangle size is less than one pixel. At that point you really have lost much of the benefits of rasterization. Not necessarily all of them, because linearly walking through a list of primitives can still be much faster than randomly accessing them for tracing, but the wins are diminishing there.


04-17-2010, 07:13 AM
Saw that a little while ago, but the utter lack of rendering anything "normal", like flat hard surfaces say a space ship interior or a current day interior or exterior is very suspicious. Surrealistic "bubbly" looking stuff is a all nice and stuff, for one game where that style fits, but if you're going to convince anyone this tech will rule the world you need to be able to show it can render some plain good 'ol stuff (but with a lot more detail) and not avoid it like the plague.

04-17-2010, 07:58 AM
It is indeed alot of talk. The reason its in software is because he needs loads amounts of memory and loads amounts of memory throughput. Todays consumer graphics cards can't hold the necessary amount of data in memory to do the processing (mainly culling) so the cpu has to do all the work. That might not be much of a problem with tomorrows (multicore cpu+gpu hybrids) or even todays cpus though. But if he wants unlimited detail he better have unlimited storage. Try putting that on a dvd or transferring that across the internet (with no lag).

It is evident that point data will become more common though. That trend is already very clear.

04-18-2010, 09:06 AM
The concept, as already stated is not totally new, in fact there are a few really good books on point splatting none of which are new books. The ability to cull data from huge data sets down to a desired subset is just simple tree searching mixed with good ole linear alg. The big issue is that you have a huge dataset which does have to be held in memory of some-sort, some-how from which the alg works to cull it down to screen res size. None of this is that much of an issue to overcome other then the damn cost of communication, the great down fall in HPC and all computing.

There are other issues which they miraculously hid, things like animation, dynamics, dynamic lighting. These make the setup much more interesting, not impossible, just more interesting.

Is this worth while to investigate? As a matter of fact yes! I'm interested and even asked about something that's cousin to this technique which Pixar has already started using. Will we see something like this coming soon in say 3D-Coat soon? No its a whole new system from the ground up but seeing the way they move on new tech I would not doubt Infinity-Coat could be more then a couple years out of the question (coding takes time, like it or not Core folks, coding takes time!)

04-18-2010, 09:58 AM
looks great, but if it's true i assume it's gonna take a many years before it's usable.

bugs need to be fixed.
things missing,... Deformations, Caustics, AA, GI, Volumetrics, Bones, etc, etc.

hope it's something though,...

04-18-2010, 10:09 AM
edit; animation

more info;

04-18-2010, 11:06 AM
Deformations exist in the a time series of the point cloud which is ouch data wise since tree search of the cloud at time 1 may give point q but at time 2 may(and should) yield point r meaning deformations, and thus bones and other dynamics, must happen on the raw point cloud, OUCH.

Caustics, refraction, diffraction and other such issues is one heck of an interesting task since you have to provide a spatial mesh even if just functional to do the calcs so is this done on the raw point cloud or the resultant tree search points? From their definition not something easily accomplished. Heck even reflection will be interesting.

AA by definition of the method does not happen in the purest implementation. That being an issue for clean results try and figure out motion blur, yeah good luck with that.

GI, thats where Pixars point rendering comes in and thats what I'm more interested in.

Volumetrics, actually this is pure volumetrics, just add additive opacity to the setup and voila your done.

What I like is the 'bug free algorithm' they claim on the site but oh wait what are those artifacts in the videos they ask patience on. Opps not so bug free after all.