Circles in computer games

Games are rendered using polygons. Essentially, it fits together polygons in order to create a shape. This works pretty well. Graphics cards are designed to render polygons, do so effectively, and most of the time you don't really notice the fact that everything is made from polygons.

Of course, one of the limitations of making things out of polygons is that you can't create perfect curves. Game devs could make that circle indistinguishable from a real circle but it would take a lot of GPU resources - resources better spent on things like character models. Besides, you don't really notice such a thing whilst playing the game normally.
 
It's all about polygon count, the more polygons the more processing power. Spheres and extruded circles need a high polygon count to make them smooth so they make them faceted to reduce load. It's about a balance between quality and optimisation.
 
Games are rendered using polygons. Essentially, it fits together polygons in order to create a shape. This works pretty well. Graphics cards are designed to render polygons, do so effectively, and most of the time you don't really notice the fact that everything is made from polygons.

Of course, one of the limitations of making things out of polygons is that you can't create perfect curves. Game devs could make that circle indistinguishable from a real circle but it would take a lot of GPU resources - resources better spent on things like character models. Besides, you don't really notice such a thing whilst playing the game normally.

Ah nice! Well explained. Are the use of polygons the mainstream way of making models? Are there any other methods out there or up and coming?
 
Very much mainstream to handle graphics rendering with triangles (smallest 3d primitive) of points (vertices) as it's the simplest geometry surface upon which to draw.

Other methods have been tried in the past and met with limited popularity (voxels), especially since the ubiquitous rise of graphics cards and standard APIs.

That said some companies are still looking at alternatives to these traditional rendering techniques, like this:

 
Very much mainstream to handle graphics rendering with triangles (smallest 3d primitive) of points (vertices) as it's the simplest geometry surface upon which to draw.

Other methods have been tried in the past and met with limited popularity (voxels), especially since the ubiquitous rise of graphics cards and standard APIs.

That said some companies are still looking at alternatives to these traditional rendering techniques, like this:


Very interesting video but that's the most bizarrely contrived voice-acting I have ever heard in my life.
 
Apart from anything else, everything displayed on a computer monitor is made up of pixels. So a "circle", no matter how well defined, will be just a collection of dots.
 
Apart from anything else, everything displayed on a computer monitor is made up of pixels. So a "circle", no matter how well defined, will be just a collection of dots.

But that doesn't really effect the way the shapes in game are rendered, does it? That's just the way they're displayed... which is true enough. But unless you're holding your eye extremely close to the monitor you're not going to notice this :p
 
Isn't tessellation supposed to address the op issue?

Tessellation is essentially the act of dividing a polygon (closed shape) into smaller parts. For instance, two triangles can be created when you cut a square diagonally. By tessellating the polygon into those triangles, developers can then deploy additional technologies, such as displacement mapping, to create more realistic images.
http://compreviews.about.com/od/video/a/What-is-Tessellation.htm
 
Tesselations still rely on polygons (subdivided, and then subdivided again, and again).

Ray-tracing would generate perfect geometrical shapes (within the limitations of your screen resolution), but it's a different technology, and still realistically out of reach.

The classic render pipeline is pretty limited in that sense. You end up approximating all your geometry with triangles, and 'painting' them with 2D textures, whether it's terrain, buildings, characters, particle effects. This is 70's technology, or even 60's tech, perfected over the years. It's well suited to computer hardware.

I used to love writing software renderers. :)
 
Very interesting video but that's the most bizarrely contrived voice-acting I have ever heard in my life.

unless he can keep it up for ~30 minutes, then thats his real voice.

the problem with that video is that everything shown is reasonably well known technology. what would impress me is the ability to animate the objects at that level of detail or get variable lighting, something that hasnt been done with the variation of the technique they are using. this is why polygons are still being used and pretty much no game company has jumped on the "unlimited detail" bandwagon
 
Sometimes you can't notice its not a circle due to distance. At distance games tend to use whats called LOD (Level of detail) where a model changes detail depending on how near you are to it. Its pointless having extra bumps and details modeled when you can't see them at distance.

I've noticed this happens with circles too (sometimes you can notice LOD), so sometimes when you're close, the circle is near perfect, as its okay for power to be spent on the polygons as its in close view with not a lot of other objects in view.

Splines also can be perfect circles I believe. Splines are a little 3D object that is repeated/tiled so ... err I give up trying to explain because I can't. :o
 
Very much mainstream to handle graphics rendering with triangles (smallest 3d primitive) of points (vertices) as it's the simplest geometry surface upon which to draw.

Other methods have been tried in the past and met with limited popularity (voxels), especially since the ubiquitous rise of graphics cards and standard APIs.

That said some companies are still looking at alternatives to these traditional rendering techniques, like this:


lol people still believe that bs , I don't remember the exact number of voxel's they used on that scene but the data size was the equivalent of 200 million tb hardrives or something silly
 
Last edited:
That said some companies are still looking at alternatives to these traditional rendering techniques, like this:

That nonsense has been debunked various times on this forum alone. I think even one time Notch wrote up a blog about how it is ultimately an investment scam.
 
Splines also can be perfect circles I believe. Splines are a little 3D object that is repeated/tiled so ... err I give up trying to explain because I can't. :o
Splines are a way of representing curves from a series of points. As their shape is determined by a mathematical formula they are technically a prefect representation of a curve; however they still have to be converted to a series of points (or, in the case of surfaces, polygons) to be rendered in a traditional GPU pipeline which takes you back to the same problem i.e. how many steps / polygons to use.

This looks interesting...
 
That nonsense has been debunked various times on this forum alone. I think even one time Notch wrote up a blog about how it is ultimately an investment scam.

Did I say it was a polygon beater, or did I say that companies were looking at alternatives to traditional triangle rendering? Fake/exploitative or not, companies aren't all on the same bandwagon. Notch is a bit of a **** at times, and the company has repeatedly said they're not looking for investment as they've already secured enough funding for the research they want to do.

Whether he can pull it off or not is another matter, all that's important is that he's trying and he thinks he can, and he's willing to show how he's trying to do it, which was the point of posting it.
 
Back
Top Bottom