While I do like using passmark for comparing video cards, its currently quite biased to Nvidia video cards. When two cards of similar performance are compared, such as the gtx 970 and r9 390, the Nvidia card will win, even though the AMD, in this instance, generally performs better. AMD cards are about 10-20% less than what they should really be.
Don't just use one benchmark as the be all end all for which card is better, you need a range of different benchmarks because different cards excel in different areas and games.
Also, 1060 3g is currently a card with extremely obvious planned obsolescence, just get a used gtx 970/r9 290/r9 390 instead for cheap.
I agree that a variety of benchmarks is needed. However, I think that we would both agree that 3DMark is still a reliable option for comparing performance.
I don't think that AMD cards are necessarily portrayed 10-20% worse than they should in games. The performance difference between the card comes down to the relationship that NVIDIA has to developers when it comes to better optimizing their cards for certain situations (just like how NVIDIA's hairworks destroys AMD cards in The Witcher 3).
Even going beyond that, NVIDIA cards utilize much better tech compared to AMD, which makes real gaming performance much better on an NVIDIA card (not to mention that I much prefer NVIDIA control panel to AMD's catalyst, and NVIDIA also has some awesome benefits like Shadowplay).
All in all, even when it comes to gaming, any of these cards would be an excellent choice, and will provide you with FPS upwards of 80/90 in most modern AAA titles at 1080p medium/high. I give the used 960/970/980 or a new 1060 3/6gb the upper hand because of the extra benefits and tech, but if you like AMD and want to feel edgy then go for a 570/580, as those will also give you some great performance.