CG Rush Board
3D Fundamentals · Article 2 / 8

Vertices,
Edges,
Polygons

What a 3D model is actually made of — and why the answer is both simpler and stranger than you expect.

Pick up any object near you. A coffee cup, a phone, a chair. Now imagine that instead of physical matter, it was built from tiny flat triangles — thousands of them, fitted together like origami — approximating every curve and surface you're looking at. That's a 3D mesh.

Every 3D model that has ever existed — from the T-rex in Jurassic Park to the walls in your IKEA apartment visualization — is built from the same three ingredients. Once you understand them, you understand the foundation of the entire discipline.

The three building blocks

·
Vertex

A single point in 3D space, defined by three numbers: X (left/right), Y (up/down), Z (forward/back). Plural: vertices. It is the most fundamental unit — everything starts here.

Edge

A straight line connecting exactly two vertices. Edges are always perfectly straight. A curve in 3D is an illusion made from many short edges arranged in an arc.

Face / Polygon

A flat surface enclosed by three or more edges. The simplest polygon is a triangle (3 edges). A quad has 4. Polygons with 5+ sides are n-gons. All of them are converted to triangles by the engine.

Together, a collection of vertices, edges, and faces is called a mesh. The mesh defines shape but carries no color, no material, no texture — those come later in the pipeline.

Step 1 — Vertices
Points placed in 3D space. Each one has X, Y, Z coordinates and nothing else yet.
Step 2 — Edges
Straight lines connecting pairs of vertices. The cube's shape becomes visible, but it's hollow — no surfaces yet.
Step 3 — Faces
Flat surfaces fill the spaces between edges. The cube now has walls — top, front, sides, bottom.
Step 4 — Triangulated
The engine splits every quad face into two triangles (dashed lines). This is what the GPU actually processes.
Cube structure workflow. A 3D cube starts as simple isolated point coordinates (vertices), receives wireframe edges, fills with face polygons, and is finally converted into GPU-native triangles.
The key insight There are no true curves in a 3D mesh. A circle is a polygon with many sides. A sphere is a ball made of flat triangles. A character's smooth cheek is dozens of tiny flat faces placed so close together that from normal viewing distance, your eye perceives a curve. This is the central trick of all 3D graphics.
Interactive 3D Demo — Left-click + drag to orbit | Scroll to zoom
Subdivision:
Vertices: - | Tris: -

Triangles — the only shape the GPU understands

When you model in Blender or Maya, you work mostly in quads — four-sided polygons. Quads are easier to work with, deform better when animated, and subdivide more predictably. Professional artists work almost exclusively in quads.

But every game engine, renderer, and GPU operates on triangles. Before your model is processed, every quad is automatically split into two triangles. Every n-gon (5+ sides) is split into as many triangles as needed. This process is called triangulation.

Triangle (Tri)
3 vertices, 3 edges. Always flat, always safe. The GPU's native format — never causes triangulation problems.
Quad
4 vertices. The artist's working format. Splits into 2 triangles (dashed). Clean and predictable as long as all 4 points lie on the same plane.
N-gon (avoid)
5+ vertices. Triangulates unpredictably — different software splits it differently, causing lighting artifacts. Avoid in areas that deform or curve.
Polygon types. Game engines convert all polygons into triangles for rendering. Quads are standard for modeling, whereas n-gons are problematic and avoided in deforming areas.

Why triangulation matters in practice

When a quad is flat — all four vertices on the same plane — the triangulation result doesn't matter. But when one vertex is raised slightly off the plane of the other three, different applications triangulate it differently. Maya might split it one way. Unreal Engine another. The result: a model that looks correct in your software looks wrong in the engine.

Safe approach ✓

Add the diagonal edge manually on any quad you know will be non-planar. You control exactly how it triangulates in every application.

Problem approach ✗

Trust the auto-triangulator in the export dialog. It will sometimes work, but fail in subtle ways that are genuinely hard to debug later.

The mesh — more than just shape

A mesh isn't just a list of triangles. Each vertex carries a set of properties that control how the mesh behaves in the engine:

PropertyWhat it storesWhat it controls
Position X, Y, Z coordinates Where the vertex exists in 3D space. This is the fundamental property.
Normal A direction vector (normalized) How light reflects at this point. Controls whether edges appear hard or soft.
UV coordinates U and V values (0 to 1) Where on a texture image this vertex maps to. Required for texturing.
Vertex color RGBA color values Color data baked into the geometry. Used for weathering, ambient occlusion, terrain blending.
Skin weights Bone influence percentages How much each bone in a skeleton moves this vertex. Required for character animation.
1.24 · 0.00 · −0.88 0.0 · 1.0 · 0.0 U: 0.42 · V: 0.77 shoulder: 0.6 · spine: 0.4
Position
X, Y, Z coordinates. Places the vertex in 3D space. The fundamental property every vertex must have.
Normal
A direction vector. Tells the lighting system which way this surface faces. Controls hard vs soft edge appearance.
UV Coordinate
U and V values (0–1). Links this vertex to a specific spot on a texture image. Required for any texturing work.
Skin Weight
Bone influence percentages. Tells the engine how much each skeleton bone pulls this vertex during animation.
Vertex property channels. Each vertex is a data container holding spatial position, lighting vector, UV mapping coordinates, and skeleton bone weights.

Normals and Shading — Flat vs Smooth Shading

A normal is a vector that points perpendicular to a surface. In a 3D engine, normals tell the lighting math which direction a face or vertex is facing. This has a massive effect on how smooth a model looks, even without adding more polygons.

When using Flat Shading, the engine calculates lighting once per face. Every triangle has one flat color, making the polygonal structure of the mesh fully visible. When using Smooth Shading, the engine interpolates the normals across the surface of the triangles, blending the lighting and creating the illusion of a perfectly smooth, curved surface (Source: Blender Manual).

Topology — the word that separates beginners from professionals

Two meshes can look identical from the outside and be completely different in quality. The difference is topology: the arrangement of polygons, the flow of edge loops, the distribution of geometry across the surface.

Bad topology looks fine in a static pose. It fails the moment anything deforms, bends, or animates.

Edge loops — the key concept

An edge loop is a continuous ring of edges that travels all the way around a mesh, returning to where it started. Edge loops define the structure of a model. On a character arm, edge loops should ring around the joint — that way, when the arm bends, the geometry flexes instead of collapsing.

Bad topology
Random diagonal cuts, n-gons (5+ sided faces), and pole vertices where many edges meet. Looks fine when static. Breaks when deformed or animated — geometry collapses and pinches at problem areas.
Good topology
Even, parallel edge loops running around the cube. All quads. When this mesh is animated or bent, every polygon has a clear role to play — the geometry flows instead of fighting itself.
Mesh topology standards. High-quality 3D assets require clean edge flow. Poor layout is prone to lighting distortions and mesh collapse during animation, whereas good topology deforms cleanly.
Real industry story Studios like Blur Studio or ILM have a mandatory topology review stage before a model reaches the rigging team. The modeler submits a topology sheet showing every major edge loop and demonstrating the mesh deforms cleanly in every extreme pose the character needs to hit. A model that fails topology review gets sent back — not to be rebuilt.

Polygon count — how many is enough?

The total number of polygons in a mesh is its poly count. Every triangle that's visible on screen must be processed by the GPU every frame. More triangles means more GPU work per frame — so poly count directly impacts performance.

There is no universal right answer. Only the right answer for your specific platform, camera distance, and project budget.

12 triangles
A raw cube. Fine for a distant background prop in a mobile game. Sharp edges, no curvature. Extremely cheap to render.
~96 triangles
One level of subdivision. Rounder silhouette, edges soften. Suitable for mid-range game props — crates, furniture, environmental objects.
~384+ triangles
Two levels of subdivision. Smooth surfaces, suitable for hero props or close-up renders. Real characters have tens of thousands at this quality level.
Detail levels vs polygon cost. Higher polygon subdivisions yield smoother silhouettes and curved edges but require significantly more computing overhead to process.

Optimization is a highly valued skill in the industry. Production hubs worldwide, particularly in Ukraine, which is home to a massive ecosystem of game art and asset outsourcing studios (such as Kevuru Games, Room 8 Studio, and others), specialize in optimizing high-poly sculpts down to strict budget limits while preserving visual quality. This global outsourcing market dictates modeling budgets based on platform requirements.

ContextPoly countExamples
Mobile background prop200–2kBarrels, rocks, small foliage in mobile games
Mobile playable character3k–8kHeroes in mobile RPG / strategy games
PC/Console environment8k–30kChairs, doors, vehicles, architectural details
AAA NPC character20k–60kBackground characters in open-world games
AAA hero character60k–150kKratos, Ellie, protagonist models
Film / cinematic character500k–5M+Thanos, Na'vi, film Spider-Man
Unreal Engine 5 changed the rules Epic's Nanite technology (UE5, 2022) partially removes the upper polygon limit for static geometry. A rock can now be a multi-million-triangle scanned asset — Nanite streams only what's visible at the correct detail level. This doesn't eliminate poly count concerns for characters or animated objects, but it fundamentally changed how environment art is built.

Wireframe — the X-ray view of any model

A wireframe displays only the edges of a mesh, with no surface fills. It's the X-ray view: you see the skeleton of the model rather than its skin.

Wireframe reveals things that rendered views hide — where geometry is unnecessarily dense, where edge loops are distributed poorly, where triangulation will cause problems. Professional artists switch between wireframe and solid view constantly.

Solid view
The final rendered appearance. Everything looks clean. Internal topology is invisible — problems hide here.
Wireframe view
All edges exposed. You can see every polygon, every edge loop, every vertex. In Blender: press Z → Wireframe. This is where topology quality is judged.
Viewport display comparison. Solid shaded view hides geometric flows and problems. Wireframe view exposes the absolute structural anatomy of the model.

What to do right now

Understanding vertices, edges, and polygons isn't trivia — it's the vocabulary you need to understand every other concept in 3D. When someone talks about "dense topology," "clean edge flow," or "poly budget," they're talking about the building blocks you just read about.

The most practical thing you can do now: open Blender, add a cube, and toggle wireframe view (press Z → Wireframe). Add a Subdivision Surface modifier and watch the face count multiply. Move a vertex off-plane and see what happens to the faces around it. Add an edge loop with Ctrl+R.

3D is a tactile discipline. Reading is necessary — but the concepts only lock in when your hands are in the tool.

Key takeaways