So, not scalable at higher speed or core counts or both then. It was a small chip on 7nm from tsmc, I don't remember hearing yield issues.
Also not scalable then, since it was only designed for mid range performance. If you believe that, then it means there's something about the design that doesn't work at high performance. But that's not how historically GPU have been design. All the lower tiers are basically the big chip with clusters removed, there's not a whole lot of difference in design. Whatever is their biggest chip is their best chip. Where the performance lands is what it is.
Let's do some simple math, including all the cost of the card, 5700 ended up selling for $400. So, you think that basically having the same board and maybe add another 4GB of memory and have a 50% bigger chip that sells for $800-$1000 wouldn't have bigger margins? And it's not like you're wasting the chip, bin the bad ones down to 5700 and 5500. If that's really how it is, no wonder NV is on top, AMD is not even trying. I prefer to think that AMD tried but could not scale up for some technical reason. No shame there, sometime things just didn't work the way you expected and the next version will be better. Also, I hope you're wrong about the wafer availability, cause there's tons more pressure now at 7nm. New xbox, ps5, zen 3, new iphones. If they have the same mentality, how many big Navi will we even be made?
Why bother even, with the price of 3070 and 3080, might as well not even make any big Navi. /s