It's clear why Nvidia did not go down this route: the per-die cost would have been identical to the RTX 20xx series, so the only way to have been able to sell them for a lower cost would have been to harvest dies that only had defects within the RT core and tensor core area specifically. That's only a small proportion of total die area (IIRC under 20% total die area) so this would have been an exceptionally small supply of dies which could have been used for a notional 'GTX 20xx" card but not instead used for a RTX card lower down the scale (e.g. GTX2080 vs. RTX 2060).
Offering a very short supply of cards that performed 'the same' in older games at a lower price point would have left Nvidia either facing the gibbering masses accusing them of deliberately manufacturing fewer of the 'better' cards - as if Nvidia had control over defect placement during fabbing - or selling the same die without defects as a fused-off GTX card rather than a full-featured RTX card, and losing money on it.
In the end, demand for the RTX series proved not to be an issue, and DLSS and RT have been widely adopted.