Prior release of Turing, nVidia had better performance per transistor per clock. With Turing, they degraded themselves due to investment which did not exactly improve classical rendering methods. And as result Vega is practically equal with Turing in this comparison. But Navi in its 1st iteration is better in this department. And we can only expect better performance per transistor per clock. Issue at hand is that AMD is AMD and they like to use available clocks and therefore set higher TDP/TBP. On other hand Navi can do 2GHz and clocking it out of the box to just 1500~1600 MHz must look like quite some waste. So at time nVidia delivers 7nm EUV GPUs next year, AMD will deliver their alternative. And here is what I expect: nVidia will double down on RTX capability which will again weaken their performance per transistor per clock. AMD will deliver final Navi configuration which will improve again performance per transistor per clock. AMD will have quite some advantage in terms of manufacturing cost of GPUs vs. delivered performance. It may come again at cost of higher power consumption. But that remain to be seen. And we do not really know real power draw characteristics of 5700 XT and what we can do about it.