Why is CUDA faster then OpenCL

Hey guys,

I just wonder why is CUDA faster than OpenCL.
Nvidia develop special arhitecure to sustin CUDA?

I am curious too. It used to have comparable speed, now became slower.

Is it becoming slower, or is CUDA becoming faster? I would think it’s the latter, because NVIDIA clearly makes significant investments into the CUDA ecosystem as a major revenue-generating technology and unique selling point, while there seems to be very little investment into OpenCL. So as CUDA improves continuously while the OpenCL implementation remains largely stagnant, it would follow that the performance gap between the two software platforms would increase over time.