Does it even has a performance impact? Or why is it relevant? Or is it blocking you on using latest cards?

Anyhow updating CUDA is not hard. In my OpenCL Fork I actually upgraded it to Cuda 11.3 before ripping out the cuda code.
I think you only have to modify the BitCrack.props file to upgrade the cuda version.

https://github.com/Uzlopak/BitCrackOpenCL/blob/master/BitCrack.props