It's the graphics equivalent of perpetual motion.

The screencaps certainly don't look impressive. A weird animal drawn in a bunch of pyramids? The textured polygon data looked far better.

The weird animals in pyramids thing said there were 8 billion points. Now assuming each point is a 3d vector of floats, that's 48 gigabytes of memory.

To summarize (based only on watching that video):

Rather than using a limited number of flat polygons, you use point cloud data to achieve greater detail. Since the geometric building blocks are "round" it makes it easy to have truly round geometry as well.

The actual drawing is done in software. They claim to have a search algorithm that searches through all of the point cloud data to find only the pixels that will actually be drawn. So rather than having to load tons of polygons into the GPU, it just gets an image the size of the screen. They claim to be able to do this in real-time and allegedly demonstrated with scenes containing billions of points (IIRC, don't ask me about the actual scale).

Essentially, they claim to be able to have much more detail (they keep saying unlimited, but I'm sure there is still a real-time processing limit, let alone the obvious storage limits) than a polygon-based GPU can handle. They also don't need to have multiple versions of the same models because the point cloud data can just be scaled in real time, keeping it's shape and detail they are only limited by system memory and the algorithm is supposed to be able to go through tons of data efficiently at run-time.

The algorithm is way over my head, but it sounds good in theory. And they seem pretty confident about it so I can't imagine it's a gimmick. If what they're saying is true, in a few years games could have much more detail and perform much better. No more swapping of models at run-time and actually round surfaces in game instead of painted flat ones.

This technology will also probably dramatically change the way graphics cards are made. It's claimed that the processing is done in software, so the graphics cards of today could be obsoleted and optionally replaced by an extra CPU or something.

Then you're going to have too many of them for the hardware to handle.

Looks cool. Hopefully it's not some big hoax

Sooo, they basically invented an algorithm which kind of works like raytracing, but way faster? With only seeing this video, I am more than sceptical.

So they have the biggest collective computer in the world?

I don't believe it. It looks like they are trying to just steal a bunch of investor's money. I never trust anybody who invokes a conspiracy theory.

An "instance" of a point (or sphere) at 8 billion unique locations. Or you mean the "animal" as a whole? Then the pyramid as a whole? Lemme watch it again.

[EDIT]

Reading the comments for the video seems to imply ray-casting. But this "MASS CONNECTED PROCESSING" sounds dodgy, as well as "compressing the data to be very small". Maybe it's lossy or something, but I'm not holding my breath.

Basically what they are doing isn't new, its just they're sort of bullshitting people making them think it is. Here some software that uses that technology for orbitals. They say they use a search engine to generate the points or something and that might be the latest thing to it, besides that theres nothing new to that idea, which is why Im not surpised that most of you think its bullshit.

Orbital viewer does this

"Three Types of Rendering Point probability plots, polygon surface plots, and raytracing (both probability and surface plots)."

The technology google uses to index billions of page requires understanding matrices that have billions of independent bases. The search algorithm can extract a certain base because theres a quicker way to do it somehow, I forgot the actual name of it but I learned about it in linear algebra.

It's all right, I'm sure piccolo is already making a rival engine based on the holographic principle