So do you recommend I stay with the 32gb unified memory and that should be enough for good long five years with the usecase? Offcourse as a ML engineer I have access to GPU clusters and AWS Cloud Compute, however, as I mentioned I would like to train ML models sometimes and run Docker, K8S, etc. on it. Also the benchmarks shown here look compelling: https://youtu.be/u9ECps9b664