A lot of this is happening because of die shrink and the way monolithic design of cpus work, vs amd approach which is make a bunch of highest tier die but make the cores if they get printed bad get either used in lower tier processors or you take all the best preforming cores and but them into one top tier chip. The issue with power draw is nivida didn't invest in scalability of its node or process so they are just cranking voltage to try to beat Amd. Amd is working on dual gpu cards again but using the infinity fabric which let's them sew cores on cpus together, that with the fact that nivida has been investing in expensive exclusive cores like cuda and Ray tracing while amd is just building weaker but more plentiful rt and there version of Cuda into the basic cores, which if your following is scalable, also easier to manufacture since in made on fab. :shrug-outta-hecks: Also nivida is behind on the node or die size, so samsung has been to some extent creating nividas gpus cores at a minor loss but are glad for the opportunity to do so.