Technology
How does a GPU work?
A GPU works by splitting a task into thousands of small calculations and running them all at the same time across thousands of simple cores. This massive parallelism makes it ideal for rendering graphics and for training and running AI models.
See it in motion.
Watch a 2-minute animated lesson that shows exactly how a GPU works.
Step by step
- 1It has thousands of small cores, versus a handful of large cores in a CPU.
- 2It excels at doing the same operation on huge batches of data simultaneously (parallelism).
- 3It was originally built to render pixels, where each pixel can be computed independently.
- 4That same parallel math now powers modern AI — training and running neural networks.
- 5It works alongside the CPU, which hands it the heavy parallel workloads.
Frequently asked questions
- Why are GPUs used for AI?
- Neural networks are mostly matrix math — the same operation repeated across enormous data — which maps perfectly onto a GPU's thousands of parallel cores.
- Is a GPU faster than a CPU?
- For parallel workloads like graphics or AI, far faster. For varied, sequential logic, a CPU's powerful cores are usually better.
- What is VRAM?
- Video memory on the GPU that holds the data it is working on — textures or model weights — close to the cores for fast access.