About this document

Evaluating CUDA Unified Memory Impact by sohamonapple is a document available to read on EtoBox.

This paper evaluates the Unified Memory technology in CUDA 6.X, which simplifies GPU programming by creating a single coherent memory space for CPU and GPU, eliminating the need for explicit memory management. The authors tested various benchmarks on NVIDIA GPUs, including the Kepler K40 and Jetson TK1, and found that while Unified Memory simplifies programming, it results in an average performance loss of 10%. The study also identifies redundant memory transfers and page caching faults as primary reasons f

Author

sohamonapple

Language

EN