Skip to content
Sign in

What is GPU?

A GPU (Graphics Processing Unit) is a specialized processor with thousands of small cores optimized for parallel computations, widely used to speed up AI and machine learning workloads.

Unlike a CPU that handles a few complex tasks sequentially, a GPU excels at performing many simple calculations at the same time. This architecture makes it especially efficient for the matrix multiplications and tensor operations common in neural network training and inference.

Modern AI frameworks such as PyTorch and TensorFlow automatically offload heavy computations to GPUs when available, using libraries like CUDA or ROCm to manage data transfer between CPU memory and GPU memory.

GPUs can be found in data-center servers, cloud instances, and consumer graphics cards, allowing both researchers and practitioners to iterate on models far faster than with CPUs alone.

Example

A researcher training a ResNet image classifier on millions of photos can finish an epoch in minutes on a single GPU instead of hours on a CPU, dramatically shortening the overall experiment cycle.

Why it matters

GPUs have become the default compute engine for deep learning, enabling the rapid scaling of model size and dataset size that drives today's AI progress.

Frequently asked questions

Graphics Processing Unit, originally built for rendering images but now essential for AI computations.