Note: This is an article on 3d graphics technology. I will try to keep it as approachable as possible but it’s not the easiest material in the world.
There’s a new graphics company that wants to change the video game industry. Euclideon is an Australian company working on a point cloud based approach to 3d graphics as opposed to the long standing polygon based approach. They are not the first people to make or even implement this approach as this somewhat angry blog post points out but they seem to have the most efficient and nicest looking implementation to date.
The video cannot be shown at the moment. Please try again later.
Their video explains the concept of what they are doing very well and throws out lots and lots of really huge numbers. I find their efforts promising but they have many hurdles to get over first.
What did they do right? Euclideon has described their renderer as working like a search engine. It intelligently solves the seemingly insurmountable task of finding what vertexes are visible to the screen and processing their information without having to compute info on every vertex everywhere in the whole scene. This is a lot like how google or bing can take my search query and somehow find 10 to 15 pages most relative to it amongst the trillions of pages they have to pick from in a fraction of a second. They are, according to the video, improving the speed of this rendering and the quality of its output constantly and all that seems well on track.
What’s the hold up? If it’s so great and they solved the unsolvable then why isn’t call of duty using it already? In reality the renders they are demonstrating are just the beginning of a long cycle of development and a need, if this is to be widespread, for changes in hardware. One of the biggest things to note in their demonstration is a complete lack of movement. The camera moves but the elements in the world do not. The trees don’t sway, there isn’t anyone walking by, there’s just a lot of stuff frozen in place. Movement will be a big hurdle in these point based environment because it takes a huge amount of math to move any one vertex.
Matrix multiplication is used to move most things in a 3d environment now. I could go into details on how that works but the important part to understand is every vertex in a scene that gets moved means more math needs to be performed by the processor. Graphics cards and video game systems have been specially designed to perform as much of that math as they can as quickly as possible but going from polygons to point clouds means increasing the number of vertices by hundreds of times. That doesn’t make it impossible but it’ll take just a big an innovation in genius to make that work as did the rendering process.
Graphics cards could be adapted to this purpose and made to devote most of their transistors to multiplication instead of rasterizing polygons but the fact is they will not get that kind of buy in until long after they have made it work on existing machines and even if they do video game consoles will not adapt to it for another full generation which is another 5, 7 or even 10 years.
In the end I think they have a cool idea and a lot of potential but a whole lot of work left to do. There are people criticizing them for promoting themselves and asking for investment but that’s exactly what they need to do. Sometimes open source and spare time projects are just not enough to push a new technology through to reality. This is one of those times when a real office full of paid employees is needed to get it done. I applaud their effort to make real change and the great job they’ve done on self promoting through these well made explanitory videos.
Will an investment in them pay off? I have no idea. It could, or it could be a dead end research path that fades away a few more years down the line. No one can predict that but they are in a lucrative business. The Unreal Engine which is probably the most used middleware in the world makes 30% off of every game that uses it and is used in at least half of all top grossing games in the last year. If Euclideon can get there they will be very profitable and if they are incorporated into unreal then the payoff could be just as good. Only time will tell, but keep an eye out for them.