A1.1.2 · Computer hardware and operation
The graphics processing unit
By the end you can
Describe what a GPU is for and why its design suits graphics, video, machine learning and other highly parallel work.
Your screen shows millions of pixels, and in a game every one of them has to be recalculated 60 or more times a second. A general-purpose CPU, working through a few instructions at a time, could never keep up. The exists for exactly this job.
The key is . Instead of a few powerful cores, a GPU has thousands of small, simple ones. Each is slower than a CPU core, but because the colour of one pixel rarely depends on the pixel next to it, they can all work at the same time.
CPU
0 ticks · 0%
4 cores · each shades 1 pixel per tick
GPU
0 ticks · 0%
256 cores · each shades 1 pixel every 4 ticks
Every pixel can be worked out on its own, without waiting for any other: exactly the kind of job a GPU is built for.
What rendering involves#
turns a description of a scene (objects made of triangles, their materials, the lights and the camera) into a flat image. For each frame the GPU works out where every corner of every triangle lands on screen, which pixels each triangle covers, and what colour each pixel should be once textures, lighting and shading are applied. It stores the images, textures and models it is working on in its own fast memory, .
Beyond graphics#
Around the mid-2000s, people realised that “the same calculation on lots of data” describes far more than pictures. Today GPUs do all of these:
Games and 3D graphics
Positions of millions of triangle corners, textures, lighting and shadows: worked out fresh for every frame, 60 or more times a second.
Video
Decoding a 4K stream to play it, or encoding video when you record or edit, without tying up the CPU.
Machine learning and AI
Training and running neural networks is mostly enormous matrix multiplications: ideal parallel work.
Science and engineering
Simulating weather, molecules or airflow means repeating the same equations for millions of points.
Cryptocurrency mining
Trying billions of hashes in parallel. GPUs were once the main tool; purpose-built chips (ASICs) have largely replaced them.
Check yourself
The GPU
Quick questions, made fresh every time and marked here. Do a few until they feel easy.
Exam practice
Paper 1-style questions
4 questions · 13 marks
Write your answer as you would in the exam, then show the mark points and tick each one your answer made. These are our own questions in the style of the exam, not past-paper questions.
1.Describe the role of a graphics processing unit (GPU).
[2]Describe: A detailed account: say what happens, step by step if it is a process.
2.Explain why GPUs are widely used to train machine-learning models.
[3]Explain: Give reasons: say what happens AND why, or how one thing causes another.
3.Outline one use of a GPU other than drawing graphics for a game.
[2]Outline: A brief account: the main points, without much detail.
4.A school is replacing the computers in its computer science lab. Students mostly use office software and write programs, but some run machine-learning projects. Discuss whether the new computers should have dedicated GPUs.
[6]Discuss: Look at the question from more than one side, weigh the points against each other, and come to a conclusion.
Past papers
Old exam questions on this topic
From HL Paper 1, May 2014 to November 2025. These papers were set on the previous syllabus, so the wording may differ from today's exam, but the topic is the same. Find the questions and markschemes in your school's past-paper pack.
| Paper | Question | Marks | About |
|---|---|---|---|
| May 2019 | 4 | 2 | Why one general processor struggles with 3D graphics |
In a nutshell
- A GPU is a processor built for graphics: turning scenes into the pixels of each frame (rendering).
- It has thousands of small, simple cores that work in parallel, instead of a few powerful ones.
- It suits any job that does the same calculation on lots of independent data.
- Uses: games and 3D graphics, video encoding and playback, machine learning and AI, scientific simulation, and (in the past) cryptocurrency mining.
- It has its own fast memory, VRAM, and takes this work off the CPU.