Comparison between CloudXYZToCloudNormal and CloudProcessingGPU

Hello, I would like to understand how to compare the ‘Calc Normal of point cloud and filter it’ step in Mech-Vision when using ‘CloudXYZToCloudNormal’ and ‘CloudProcessingGPU’.

Test Result

We tested this on an RTX 4060 Ti and found that the time taken using 'CloudProcessingGPU’ was 0.170442 s, while the time taken using 'CloudXYZToCloudNormal’ was 0.037170 s.


Here are the answers to this question:
1. Data transfer from memory to GPU takes time, and for scenarios with small computations, GPU may not have an advantage over CPU.
2. When we use the GPU without the highest priority, it can be influenced by the current system display load.
3. This is related to the graphics card bandwidth.