I read somewhere that Ngons were nothing more than triangles masked as an Ngon.
Some testing reveals this is not true, at least to some extent.
Here are two sub-surfed cylinders, one with an Ngon, the other with a Triangle Fan.
The things that look like 6 sided Ngons in the Triangle Fan example are really quads, and the cap on top is really a lot of tris:
As you can see, they are definitely different, but what is the difference technically?
This question is very generic - but you give a very specific example (Subsurf) which happens to treat ngons differently to many triangles (to be clear - differently to a triangulated ngon).
The answer is something like ... it depends: each modifier, tool, export-format and rendering engine can choose to use ngons / quads / triangles - differently. There are no fixed rules here, some tools triangulate internally to perform their calculations, others are able to handle ngon data.
Here are some general rules that mostly apply.
Note that even in cases where ngons are triangulated, the normals from the ngons/quads are often used. So you may get a better looking model when smoothing is applied on the non-triangulated mesh data.
Subsurf does not treat tri, quad and ngon differently:
In each case the result follows exactly the same pattern:
The difference in topology of the two results you showed can be more easily seen if we use simple subsurf (which does not smooth) and then apply the modifiers and look at the wireframes:
The result from subdividing the triangle fan has the extra edges (selected.) These extra edges are the result of subdividing the radial edges which connect the central vertex to the outside edges.
This does not demonstrate any fundamental difference between tris and ngons, because we can produce the same result using only triangles:
On the left is a triangle fan composed of three triangles. On the right our equivalent "ngon" is simply a normal triangle. The result of subdivision is different and follows the same pattern we saw before with the cylinders. The same results would be seen with a triangle fan composed of four triangles and a quad.
In general it's much better to think of tris and quads as special cases of the ngon than to consider an ngon a "bunch of triangles" like a triangle fan or triangle strip.
When you use N-gons you are essentially saying "I don't care how this geometry is arranged -- for now", and you relinquish control of the underlying geometry to Blender's tessellator. If you do an N-gon fill of a profile (edge-loop) you won't be able to control how faces are tessellated under the hood for display or rendering. If you need the N-gon to behave a certain way then you have to edit it manually, e.g., do a face-inset or convert it to triangles/quads manually to control the geometry.
See the example below:
left side N-gon, right side underlying geometry that you don't have to worry about
(until it becomes necessary).
.. or you may never be concerned about the underlying tessellation.
.. or the final output may not be tessellated (in rare cases).
It depends on your use-case.
At the end of the day, it's all triangles - however the idea of NGons is an abstraction that helps the software deal with certain situations that come in handy when modelling (such as being able to slice a plane any which way you like, then extrude it)
