You say that with straight face. Knowing that AMD was able to develop Zen2 on 7 nm process node?
The power targets were lower and clock targets were higher than when they came out of bakery. AMD had a choice. Either they clock them to death, or they are not faster than RTX 2070. Power targets were not met. Clock targets were. And this is solely because of the process, because it is better than people expected it to be, on clocks front, and worse than people expected it be, on power front.
A correction would be Nvidia has better Engineers in their GPU division than AMD does in their GPU division. Particularly their process and error correction engineers which relates to your second post. The history of GPU's since the Shanghai team has taken over has been lackluster compared to the Markham team, everyone can agree on that. The process and error correction team has allowed Nvidia to mostly set their volts rather optimum compared to AMD(much smaller of an overvolt). This is because there variance in samples quality is much smaller. The difference in sample quality is + or - 5% vs + or - 10+% for AMD. This causes AMD to push more volts to maximize yields and stability at a particular clock.
Your second post is a complete red herring that does not address the rebuttle to your original post. It is ridiculous to suggest that a move to a smaller node won't bring efficiency/performance gains. That's literally the trend of pretty much every nodal shift. Your largely unrelated post about AMD hitting clock goals and not hitting power goals is not only unproven it doesn't related to your original post(hence a red herring).
Also who said it was better from a clocks point? 1755mhz as the real world clocking speed(assuming AMD is being truthful here) is not that far above the 1730mhz average clocks the Radeon VII hits. If AMD designs is more similar to pascals/maxwells we should be seeing higher clocks than this. Nvidia was able to improve their clock speed during the maxwell generation nearly 20% from the architectural shift(without a power hit) and without a new node. The rx 590s were already hitting 1560 as their gaming clock, so a modest 12.5% improvement over this is not that impressive when you combine new architecture + new node. New node + Maxwell = Pascal which yielded some 60-70% increase in frequency vs kepler.
If you don't think turing is going to clock badly on 7nm, you have another thing coming because Turing has a tonne of clocks left in the tanks. How do we know this? Look at the LN2 speeds of the card.
https://wccftech.com/nvidia-geforce-rtx-2080-ti-galax-hof-oc-2-9-ghz-3dmark-world-records/
Nvidia is able to reach 2.9ghz on ln2. You might question how they are related but what I have noticed which has allowed me to make accurate clock predictions is LN2 rates give a preview of what a card to can overclock to on a new node.
What does a GTX 980 ti clock too on ln2, 2.1 ghz.
https://www.kitguru.net/components/...-hits-2-1ghz8-4ghz-with-ln2-sets-new-records/
What does a GTX 1080 ti clock to on air? 2-2.1ghz.
The same was true of Fury X vs Polaris. Fury X under ln2 could reach between 1400-1500mhz. A very good rx480 could overclock to that speed.
Now lets look at Vega, what does Vega do under ln2 atleast 2ghz and probably a bit further according to buildzoid(no one did serious testing under ln2 since no world records were achievable with the card).
What does Radeon VII overclock to? Again 2.1 ghz.
People were shocked at high pascal clocked and expected the same from GCN on the move to 14/16nm. People saw 1750mhz and 2.1ghz overclocks and expect the RX 480 to hit, 1.6 stock and nearly 2 ghz overclocks. But all we got was 1.28 ghz stock and 1.450ghz overclocks. Most were surprised by this, but if you were paying attention to the ln2 overclocks, you would not have been disappointed and actually expected this.
I don't expect turing to have stock clocks of 2.9 ghz or overclock this high under air, but I could easily seeing stock clocks of 2.3-2.4 ghz and overclocks in the 2.7ghz range(which is a common overclock of ln2).
So what does this mean? If Nvidia kept clocks at current clocks which is around the 1750 to 1800mhz range, they could likely half their power consumption(something AMD also claims with Radeon VII) because this would be 30% below their achievable 2.3ghz-2.4ghz standard clock.
This is consistent with TSMC specification of 50% power reduction at the same performance or 30% higher clocks at the same power.
So what does this mean? We are not going to see a regression in performance per watt that you weirdly want? It is pretty much impossible.