CUDA core
A CUDA core is an arithmetic execution lane in an NVIDIA streaming multiprocessor. The label commonly counts floating-point arithmetic capacity; it does not describe an independent CPU-like processor.
How it works
A warp instruction is carried out by the appropriate execution pipeline over its active lanes. How many cycles the instruction takes depends on its type and the target architecture. Integer, floating-point, special-function and matrix operations need different accounting.
Work through an example
Comparing two GPUs by CUDA-core count alone misses clock rate, instruction mix, memory traffic and Tensor Core support. A matrix kernel may spend most of its arithmetic work on a different pipeline.
Where the term can mislead
One CUDA thread does not permanently own one CUDA core. Threads are scheduled work; cores are execution resources.
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.