A GPU (Graphics Processing Unit) is a specialized processor optimized for rendering images and video. It excels at parallel processing, making it essential for gaming, 3D work, and AI inference.
GPUs contain thousands of small cores designed to process data in parallel, unlike CPUs which excel at sequential tasks. This architecture makes GPUs 10–100× faster than CPUs for graphics rendering, matrix operations, and machine learning inference.
Main GPU types: integrated (built into CPU die, uses system RAM, slower but power-efficient), discrete (dedicated card with own VRAM, fastest, highest power draw). On phones: Apple GPU (integrated in A-series), Qualcomm Adreno (integrated in Snapdragon), ARM Mali (various OEMs).
GPU specifications: shader count (parallel processors), memory bandwidth (data throughput to VRAM), clock speed (GHz). A GPU with 2000 shaders running at 2 GHz delivers roughly 4 teraflops of compute.
For gaming: discrete RTX 4090 or AMD Radeon RX 7900 XTX (desktops), RTX 4060 or RTX 3050 (laptops). For productivity (video editing, 3D): discrete GPU significantly speeds encoding and rendering. For general browsing: integrated GPU sufficient.
Artificial intelligence is driving GPU demand — language models (LLMs) and generative models run 100× faster on GPU than CPU.