Optimizing Three.js Performance: Simulating Tens Of Thousands Of Independent Moving Objects

This post covers my experience building Asterank 3D, a simulation engine capable of showing tens of thousands of particles with unique trajectories. Each particle has a unique path calculated on-the-fly using laws of astrodynamics.

I documented my progress as I scaled the simulation from 60 objects to 50,000 objects. Each section will cover techniques that produced significant performance gains.

Use a ParticleSystem

A naive implementation of the solar system view represents every single object as its own particle. While this is a convenient approach, it becomes less reliable at around 40 moving particles or so on a typical laptop.

My original code was analogous to:

for (var i=0; i < 100; i++) {  
  var particle = new THREE.Particle(new THREE.Vector3D(x, y, z));

Rather than individually adding each particle to the scene, use a single ParticleSystem:

var particle_system_geometry = new THREE.Geometry();  
for (var i=0; i < 100; i++) {  
  particle_system_geometry.vertices.push(new THREE.Vector3D(x, y, z));
var particle_system_material = new THREE.ParticleBasicMaterial({  
  color: 0xffffff,
  size: 1
var particleSystem = new THREE.ParticleSystem(  

To customize the appearance of your particles, you can use sprites to override the default shape.

This change will easily result in several orders of magnitude improvement. However, things like ray tracing/collision detection (which are useful for things like mouseover behaviors) become more complicated.

Browser "multithreading" with web workers

Javascript is single-threaded in all browsers, which means all logic blocks the main UI thread. As a result, your simulation will feel jumpy or unresponsive to the user if the browser is busy computing many new particle positions.

HTML5 Web Workers are similar to threads that communicate with the main UI thread via message passing. Running complex blocking calculations with a web worker will keep the UI updating smoothly.

Web workers are typically defined as JS files that run within their own context. They can't use any of the global variables you use in your main scripts, such as window or document. You can't do UI/DOM work in web workers, but you can perform other calculations.

Here's an example webworker file:

// in worker.js
self.addEventListener('message', function() {
  // We received a message from the main thread!
  // do some computation that may normally cause the browser to hang
  // in my case, I computed the position of an object in space according
  // to Kepler's Laws

  //  now send back the results
    type: 'results',
    data: {
      // ...

Create and run the web worker from your UI code like so:

// in your main js file
var worker = new Worker('worker.js');  
  some_data: 'foo',
  some_more_data: 'bar'

A practical implication of web worker context limitations is that you can't use console.log. In order to log messages to the console, you must pass them to the main thread. This also serves as a good example of how to use web workers to do work and pass messages back to the main thread:

// in worker.js
  type: 'debug',
  value: 'some debug message here'
// in main js file
worker.onmessage = function(e) {  
  var data = e.data;
  if (data.type === 'debug') {
  else if (data.type === 'result') {
    // process results

For a more in-depth look at web workers, I recommend the HTML5 Rocks introduction.

The gains I saw from using a web worker to handle orbital calculations were substantial. Using a single worker allowed me to smoothly render 10,000 moving particles, as the main UI thread was no longer tied up by position calculations. Additional workers could give more gains in theory, but in my case the benefits diminished very quickly.

Web workers are supported by recent versions of all browsers except for IE (IE 10 adds web worker support). They're also supported by the default browsers on iOS and Android, but have been removed from more recent versions of Android. I created a simple web worker compatibility library for browsers that don't have native support for workers.

Smarter execution flow with Timed Array Processing

After adding web workers, I discovered another bottleneck. Despite having separate threads, we must update the animation on the main thread. It turns out updating the position of tens of thousands of coordinates in Three.js was lagging the simulation.

My code essentially looked like this:

// Blocks rendering until complete
for (var i=0; i < 20000; i++)  

This simple approach delays the execution of renderAnimateFrame() and other rendering updates until the loop is completed. If the loop takes longer than your desired framerate, the simulation will lag.

The fix I implemented is based on this article by Nicholas Zakas, author of O'Reilly's High-Performance JavaScript and other books. I wound up implementing my own version of timedChunk:

function timedChunk(particles, positions, fn, context, callback){  
  var i = 0;
  var tick = function() {
    var start = new Date().getTime();
    for (; i < positions.length && (new Date().getTime()) - start < 50; i++) {
      fn.call(context, particles[i], positions[i]);
    if (i < positions.length) {
      // Yield execution to rendering logic
      setTimeout(tick, 25);
    } else {
      callback(positions, particles);
  setTimeout(tick, 25);

In this solution, setTimeout yields execution flow to things like calls to animate(). Even a setTimeout of 0 would work for this purpose, as calling it with no delay simply passes execution flow to whatever's waiting.

Using timedChunk allowed me to increase the number of web workers because it eliminated the UI thread bottleneck.

Using custom shaders

Finally, I saw a huge improvement in smoothness of the simulation by writing a custom shader for my particle system.

This effectively moved all the complex position calculations for my particles to the GPU, which went a long way toward ensuring the speed and reliability of the simulation. Custom shaders are written in GLSL, which is close enough to C that it's not too difficult to translate your math into.

At this point, the number of particles I could show on my desktop became so high that I chose to limit it so laptops with weaker graphics cards could handle the simulation.

You will see significant gains with custom shaders, especially if your particles have independent trajectories. I wrote a quick introduction to shaders that may be useful to beginners.


I went from lagging at ~60 moving particles to being able to smoothly render 50,000+ independently moving particles.

Although webgl performance lags behind native capabilities, it's only a matter of time until the technology and hardware catch up. Until then, I hope this post helps anyone who's running into similar performance issues.

Follow me: @iwebst

How I Built a Canvas/WebGL Visualization With No Graphics Knowledge

I created Asterank, a database of economic value of all the asteroids in our solar system. It condenses about 600,000 data points to the top 100, but the numbers can be pretty difficult to grasp. I sought to create a browser-based visualization of our solar system that illustrates its abundance of resources and mesmerizing beauty. Prior to this, I knew nothing about 3D graphics or WebGL.

Asterank 3D

Choosing a library

I wanted the user to be able to view my models from an arbitrary point in space, so the first step was to choose a JavaScript library for browser-based 3D visualization.

3D in the browser is a fast-developing space. Libraries such as Copperlicht, GLGE, SceneJS, and THREE.js are quite popular, with no major consensus on which is the best. I wound up choosing THREE.js because of its HTML5 Canvas fallback.

Getting started with THREE.js

This basic tutorial is the place to start, introducing the concepts that will guide most of your programming: scenes, meshes, materials, lighting, and the render loop creates animation effects.

The official examples helped with some basic knowledge, and taught me things like how to extrude 3D geometries from simple 2D shapes, but were not terribly helpful overall.

Jerome Etienne's THREE.js boilerplate was immensely useful for hitting the ground running, and he's even developed the option to generate your own customized boilerplate (pictured below).

Three.js boilerplate

Filling the knowledge gap

THREE.js documentation is very sparse and not very useful if you're just starting with 3D. Its main utility lies in the fact that it links to the source code of key classes. This became a recurring theme - the ability to read and understand source code was essential.

The official source for help is StackOverflow. There is a large amount of Q&A knowledge in the github issues for the three.js project from before mrdoob decided that questions should be asked on StackOverflow.

Lee Stemkoski's list of examples is a little-known resource, that, unlike other examples, build on one another. This is probably the best sequence of examples out there, thorough and well-thought. Someone could start here and leave with the knowledge they need for any WebGL project.

Lee Stremkoski's three.js examples

My project was complicated by the fact that I was working with the dev branch in order to get the EllipseCurve object, which I essentially re-implemented specifically for astronomical orbits.

Challenges and Lessons

  1. Orbits - this problem was specific to my project, but I wound up poring through a decent amount of astronomy literature to understand how to correctly render orbits. My model computes basic Keplerian orbits using information mostly contained in this primer. I used an infinite series approximation for true anomaly after discovering that the Equation of Center approximation was inaccurate for some highly inclined orbits (eg. 3200 Phaethon).

  2. Editing TrackballControls for scroll support - this wasn't too hard, but I added a zoom on mouse scroll. It was like binding any other event in JavaScript, but required the willingness to edit existing library source. It was important to keep in mind cross-browser differences in scroll events.

  3. Optimize for performance - keep in mind that many browsers and weaker machines don't handle WebGL or canvas very gracefully. In my project, the number of asteroids is limited and their orbits are computed lazily. I also take into account that some asteroids require less computation in order to get a smooth curve, depending on their orbits.

  4. Always use radians - this goes without saying, but working with astronomy examples that used degrees led to occasional slip-ups.


Finally, an interactive 3D (well, 4D) orbit viewer that isn't a Java applet! I'm looking forward to packaging it and making it more general-use, perhaps for some sort of educational application.

The takeaway: it's easy to program cool graphics in the browser, and you should too.