In my opinion, I don't think we'd have one on the die without the other. Nvidia is trying to lump sum the value of these features in consumer space so they don't have to make a different kind of die for professional products.
This might very well be the case. Still my statement holds, RT-cores take about 3% of die space.
Or corollary : if Nividia would have decided to go with Tensor-cores but not with RT-cores, the die size saving would be 3%.
Overall, the higher price of RTX cards can not remotely be explained by the die size adder due to RT-cores - that would be less than 3$ per card - thats noise.