Introduction to GPUs: Glossary

CPUs are for general computing, GPUs are for specific problems.

GPUs are suitable for data parallel computations.

With the right problem, a GPU can be thousands of times faster than a CPU.

CUDA is designed for a specific GPU architecture, namely NVIDIA’s Streaming Multiprocessors.

CUDA has many programming operations that are common to other parallel programming paradigms.

The memory architecture is extremely important to obtaining good performance from CUDA programs.