Buying nvidia graphic cards for Al, Deep Learning and Data Science work

I am planning to buy Nvidia GeForce GTX 1660 Super for my Data Science & Deep Learning (DS & DT) training (apart from playing video games) with tensorflow-gpu but according to the supported list of CUDA enabled devices, GTX 1660 Super are not listed.

Some tech blog on the web said that GTX 1660 Super is good graphic card to start with in the DS & DL area. But then someone pointed out that GTX 1660 Super doesnt have Tensor core and ones should go for RTX 2060 instead since it has Tensor core.

Is that suggestion true? What does Tensor core could offer in the context of DS & DT? Should I just stick to buy GTX 1660 Super?