# GPU

> The chip that does thousands of simple calculations at once.

- Level: Basic
- Topics: Hardware, AI
- URL: https://tidbit.legios.com.ar/gpu/
- In Spanish: [GPU](https://tidbit.legios.com.ar/es/gpu/)

A **GPU** (Graphics Processing Unit) was invented to draw video game frames: millions of pixels that all need the same small calculation at the same time. Instead of a few powerful cores like a CPU, a GPU has thousands of small ones that work in parallel.

It turns out that training and running AI models is the same kind of problem. A neural network is mostly huge multiplications of number grids (matrices), and a GPU can do those much faster than a CPU.

That is why NVIDIA became one of the most valuable companies in the world, why data centers are full of GPUs, and why a modern AI model is measured in how many GPUs it took to train.

## An analogy

A CPU is a brilliant chef who can cook any dish. A GPU is a thousand line cooks who each chop one onion. For a banquet of onions, the line cooks win.

## Keep in mind

- CPU: few fast cores, good at varied tasks. GPU: many simple cores, good at repeated math.
- AI training and inference run mostly on GPUs or similar chips (TPUs).
- GPU memory (VRAM) decides how big a model fits on one card.

## Go deeper

- [GPU on Wikipedia](https://en.wikipedia.org/wiki/Graphics_processing_unit)
- [NVIDIA: what is accelerated computing](https://blogs.nvidia.com/blog/what-is-accelerated-computing/)

## Related terms

- [Cores and threads](https://tidbit.legios.com.ar/cpu-cores-and-threads/): Why a CPU is really several processors, and why it can juggle even more tasks.
- [CPU](https://tidbit.legios.com.ar/cpu/): The chip that runs your programs, one instruction after another, very fast.
- [Training and inference](https://tidbit.legios.com.ar/training-and-inference/): Training builds the model once; inference is using it, millions of times a day.
- [Artificial intelligence](https://tidbit.legios.com.ar/artificial-intelligence/): Getting computers to do things that, until recently, only people could do.
- [Neural network](https://tidbit.legios.com.ar/neural-network/): A pile of tiny math units that learns patterns from examples.
- [NVIDIA](https://tidbit.legios.com.ar/nvidia/): The company that sells the shovels of the AI gold rush.
- [Data center](https://tidbit.legios.com.ar/data-center/): A building full of servers where the internet, the cloud and AI actually run.
- [Chip](https://tidbit.legios.com.ar/semiconductor-chip/): A sliver of silicon carrying billions of tiny switches that compute and remember.