[brutus1213]

I'm not a HW person but as I understand it, the latest chips AMD produces seem to have something in common with NVIDIA GPUs .. they have these blocks that they can essentially copy and paste. In NVIDIA parlance, these are SMs. I think what the OP is referring to may be the interconnect and this sort of simple architecture. You get the power from scaling the blocks (not from a more complex individual block). My previous gen Ryzen is slower on an individual core basis than the comparable Intel part but I went for it because of the core count. What is interesting is that as TSMC improves their process and they both move to a smaller node, it makes it possible to fit more blocks on an area of silicon and it gets better (power-wise?) Would be interesting to hear from an EE if my thinking is on the spot or if I am in error.