Stream processors are the basic computational units inside a graphics processing unit (GPU). Each one executes simple mathematical operations, but thousands work in parallel to produce results far faster than a single CPU core could manage.
Unlike a traditional processor that excels at complex sequential tasks, a GPU's stream processors are optimised for the same operation on many data points simultaneously. This parallel approach suits graphics rendering (where the same transformation applies to millions of pixels) and other data-parallel workloads like video encoding or scientific simulation.
How they work in practice: When you play a 3D game, stream processors handle tasks like calculating the final colour of each pixel, applying textures, and rendering shadows across the entire screen in parallel. A gaming GPU might have between 1,000 and 10,000 stream processors working together.
What to look for when buying: GPU manufacturers use different names for these units (Nvidia calls them CUDA cores, AMD calls them Stream Processors). Don't rely on core count alone to compare performance. A GPU with fewer, faster cores can outperform one with more cores running slower. Look at benchmarks in applications you actually use, such as games or video editing software. Also consider memory bandwidth, which determines how quickly stream processors can access data.
The total number of stream processors influences a GPU's raw performance potential, but architecture, clock speed, and memory speed matter just as much in real-world use.