CUDA stream
A CUDA stream is an ordered queue of device work. Operations in one stream observe its ordering rules, while separate streams can express work that may overlap.
How it works
An application uses events or other supported dependencies when work in one stream consumes another stream’s output. Different streams enable concurrency but do not guarantee it: resources, copy engines and synchronization can prevent overlap.
Work through an example
A transfer for batch B can be submitted while batch A computes, if the buffers, memory type and dependencies permit. The timeline establishes whether the intended overlap actually happened.
Where the term can mislead
Host return from an asynchronous API call does not make its buffer safe to overwrite or free. Keep it alive until the relevant work completes.
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.