GPU architecture
A GPU architecture describes the execution units, memory paths and instructions a family of processors provides. CUDA is the programming platform through which software uses supported NVIDIA devices.
How it works
A kernel is distributed across streaming multiprocessors. Inside each multiprocessor, schedulers issue instructions to arithmetic or memory pipelines. Shared caches and device memory connect those units to the data. The number of units and the operations they support change between generations.
Work through an example
An H100 and a B200 can run the same high-level matrix operation while using different native instructions and accumulator storage. Select the exact device before applying an architecture-specific optimization.
Where the term can mislead
A product family name does not establish every feature: server and desktop implementations can have different compute capabilities. Consult the exact target.
Connect the idea
Take it into practice
Continue with the existing concept, apply it to an interview problem, or follow a course lesson. Access requirements are shown before you open a page.
New to GPUs? Start with the free execution-model concept.
Sources and scope
Examples here are teaching calculations, not measured GPU performance. Consult the linked documentation for the exact device, instruction and tool version. Source review: 25 October 2026.