Yeah NVidia has a long history of artificial product releases that are cut down primarily to suit marketing/product stack needs than technical needs, RTX2060 is an excellent example of this, it's barely cut down at all from the RTX2070 and the key ways it is cut down don't really impact yields/cost but have a big impact on performance, it's an artificially capped GPU to fit into a price point neatly. In reality the BOM cost difference between the RTX2060 and 2070 is likely fractional, one is just a lower margin high volume product and the other is a high margin low volume product.
In contrast, AMD usually releases products that are primarily only binned/cut down in ways that improve yields/reduce costs, IE are logical from a technical perspective. This gives them less ability to counter NVidia with product stack changes, but a lot more room to counter them with price cuts, which is why AMD are rarely beaten on price, they will have much lower overheads on their individual SKUs.
This is far more obvious in the CPU market though, given Intel more or less invented that strategy of fusing off random bits of the silicon just to differentiate a chips marketability rather than to actually improve yields. AMD still do it too (Removing SMT is the easiest to spot example of this, no actual yield/cost difference, it's purely for marketing reasons) but much less often.
It's kinda like with cars, where electric windows or many other features were far cheaper to manufacture than their manual alternatives for many years, but still came at a price premium for marketing/product stack reasons.