Download

I want to share my opinion about Euclideon Engine

Hi. Recently I made a video on youtube (v=gM6QkPsA2ds) where I compared Euclideon Engine with Unreal Engine 4 on modern hardware. The reason why I created this video is because years ago Euclideon claimed that their technology can replace polygons to make computer graphics and videogames 100000 times better, making unfair comparisons with low-polygons trees of Crysis 2. Even if the video of 2011 was promising, this never happened (and I had no doubt about it). For me the reason why it didn’t happen is very clear, but surprisingly in the video I received comments where people seem to not understand things that for me (and I hope for you) are very clear, so I wanted to understand what I’m missing, if I’m missing something.

First of all, I worked in the field of computer graphics for so many years, that when somebody claim something like “Unlimited Power” and that polygons are the wrong way of doing computer graphics, my mind produce immediately a bad feeling. Another things that happened is a research. I wanted to know if Euclideon Engine was real, and what kind of technique can be so powerful to perform “unlimited graphics” with the mere use of the CPU. I searched on Google Patents and I found WO 2014/043735 Al “A Computer Graphics Method For Rendering Three Dimensional Scenes”. If you are interested, the following link contains the original document with all the figures at the right place:

http://www.codersnotes.com/notes/euclideon-explained/patent.pdf

After reading the document, I figured out that the technique is just a voxel engine with a search algorithm that make use of octrees, where polygons are converted into a dense cloud of points made of terabytes of data (or eventually taken from the real-world unsing a laser scanner technology). Of course this is a trick of precalculation that waste a huge amount of storage to avoid a huge amount of work on the CPU, with some optimizations like isometric conditions to avoid perspective calculation.

And of course this is not “Unlimited Power”, it cannot be used to replace modern computer graphics because it has several drawbacks. First of all, the term “Computer Graphics” covers a wide range of arguments (the same for “video games”) and when you say that you can do better than the world you have to specify what you are referring to. As Euclideon technology is just a voxel engine, it cannot be used to replace engines or techniques that don’t require it. It’s common sense.

Polygons may be used to draw everything on a videogame, also the User Interface (UI). Imagine how redundant it would be to convert all the buttons and combo boxes into 3D atoms (or voxels or points cloud) to store them into octrees and then use the Euclideon technology to render them instead of using modern GPUs to draw simple polygons. And I could say the same for a platform game like SMB, where you have 2D sprites and tiles, and the same for 3D videogames where open worlds, landscapes or lodding are not required.

My point is that polygons and modern GPUs (which is portrayed by Euclideon like the wrong way of doing it) is the most appropriate, solid and consistent way of doing generic computer graphics, then 3D engine with techniques such as trees and search algorithms can be used to improve the performance, without replacing anything. Polygons and textures may be used to render 2D and 3D models with few resources to cover up all the possibilities. Graphics libraries such as OpenGL, Direct3D or Vulkan are optimized for processing the primitives through a graphics pipeline made of vertex, tesselation, geometry and fragment stages, with programmable shaders. If not enough, you may use compute shaders, OpenCL or CUDA to implement algorithms where vertices, polygons and textures are not required, like you want to do in common computer programming. This is how current videogames, from the lowest to the highest level, are implemented. In my understanding, there is no such thing in this field of human knowldge that can be replaced by Euclideon technology.

For me, it was surprising that such obvious things generated a cloud of disappointements. Why people still believe that a voxel engine with CPU and terabytes of data can replace current technology in the game industry? It seems that very few people can undestand the problems behind the creation of a videogame and the hardware where the game must run, and that’s an incredible amount of work. From this point of view, Unreal Engine 4 did an incredible job, nothing that you can replace with lazy convertion of polygons into 3D points, wasting the CPU and terabytes of data.

For those looking for the “Unlimited Power” i’d suggest look into the project mandelbulb :slight_smile: That’s the first thing came into my mind once reading your article. Thanks for the input, i found it informative.

I remember when they were making those claims about unlimited power. Pretty crazy. I heard that Google will be renting out quantum computing over the cloud eventually. I would think that would be the next step to unlimited power. Even if you were able to use only the CPU for processing graphics, the CPU has limits. Wouldn’t you hit a ceiling eventually? It isn’t really unlimited power then right…?

It depends where do you put your limitations. For instance, once you determine an allowed timeframe for the cpu have to complete it’s job, it becomes limited in that moment, therefore no longer limitless. But if you can allow infinite time for an abacus to finish any amount of operations it can be considered to have an unlimited processing power.

Practically speaking a cpu is only capable to process finite amount of operations in parallel, therefore it cannot be considered unlimited. The same applies for the article as well, as “unlimited power” would be invalid, since the resolution is discrete, therefore it is limited already.

They actually don’t use the word unlimited, but states the technique is faster compared to previous implementations.

‘Unlimited Detail’ has been around for over a decade now, and is very much snake oil. For many years it clearly never even advanced, they just made the same awful demos repeatedly, and never addressing the common observations about flaws and shortcomings.

There are definitely use cases for 3D point cloud / voxel octree based renderers, but games aren’t really one of them.