Point cloud data

[BlindSide]

Australians. =/

I saw this about a year ago and I haven't really heard anything new about it. They claim to use a totally unique technology so it's impossible to say anything at this stage until they release more info publicly.

[Virion]

Australians. =/

why =/ lol

[DeM0nFiRe]

Australians. =/

I saw this about a year ago and I haven't really heard anything new about it. They claim to use a totally unique technology so it's impossible to say anything at this stage until they release more info publicly.

As I understand it, their totally unique technology is just their search algorithm. They aren't using any special hardware, they're just doing it on the CPU. I would imagine they are testing it on an i7 of some sort.

[lostclimategames]

I don't think they are worrying about RAM. They are projecting this for use in 16 months. In 16 months if you don't have at least 4 Gb, then you've got a problem.

and you think 4gb is enough to hold billions of points with an x,y,z,spec,color etc, and the movement of those point and the collision code of them....?

Im gonna go out on a limb and say thats not really gonna happen.

[roelor]

Storing all these billions of points in memory would be nonsense. never heard about procuderal rendering and loading from harddrive? (they dont have billions of points btw)

[Insomniacp]

The main thing i didn't like about it is that they never got close to the ground to go between there little pebbles to prove that they are all little pebbles and not just some shading technique. ( may have missed that part thought...)

[Lonesome Ducky]

Am I the only one who finds this suspicious? Alright, suddenly we're told there's an AMAAAZING search algorithm that can sort out only the pixels you can see. But where are the papers? Not even an explanation? No tech demo? Just a video. Oh but we're supposed to wait 16 months before any tangible evidence is to come about. Techniques closely related to what is described by this person have been studied for many years, and yet even experts given that time haven't come up with anything close to this.

A quote from another forums discussing the same:

I don't believe this website at all. Their explanation does not make sense, it's not even possible. They begin to describe that they use ray-tracing (...all it wants is 1024*768 (if that is our resolution) points, one for each pixel of the screen...) but that they have some magical sorting system that no one has ever thought of before....despite that fact RT has been heavily researched for over 50 years.

I just don't buy it. Crappy website, fake explanation, small pictures, and a 3D system of infinite detail that runs on just one core of a CPU (no GPU).

I'd rather put my money on hardware tessellation anyday.

[roelor]

Yeah, Ive thought about this a lot. placing the pixels isnt so hard, knowing where the pixels must be placed... wel thats hell.

[DeM0nFiRe]

Well, I believe that they have what they say they have, but what they have isn't all that impressive. First of all, this isn't the first point cloud renderer. The problem is that the other point cloud renderers I've seen do not search for what pixels should be on the screen. Instead, it decides which points to use solely based on the percentage of the screen that the object takes up. This means that they render a bunch of extra points that don't contribute to the image because they are on the back. If you've got the right search algorithm, you could pick only forward facing ones.

I am a little suspicious, however. There website says feel free to contact them if you have any questions. I sent them an email asking what their memory usage was and what hardware they were using, and never got a response. Maybe they just get a lot of emails, but still.

[DtD]

Just though I'd point out that they uploaded a video of an animation test http://www.youtube.com/watch?v=cF8A4bsfKH8

Although, I'm still skeptical until I see a tech demo I can run on my comp.