yeah, that's pretty much why I stopped buying Nvidia after GTX 1080. Cuda was bad in terms of their practice, but not that impactful since OpenCL etc can still tune and work properly with similar performance, just software developer/researcher love free support/R&D/money to progress their goal. They are willing to be the minions which I can't ask them to not take the free money. But RTX and then tensor core is where I draw the line, since their patent and implementation will have actual harm in computer graphic and AI research space but I guess it was a bit too late. We are already seeing the results and Nvidia is making banks with that advantage. They are essentially just applying the Intel playbook but doing it slightly different, they don't buy the OEM vendors, they "invest" software developers/researcher to use their closed tech. Now everyone is paying the premium if you buy RTX/AI chips from Nvidia and the capital boom from AI will make the gap hard to close for AMD. After all, R&D requires lots of money.