You're comparing apples with oranges.
Different chips designs have different pipeline architecture and ati's cores are hardly anything like nvidia's cores.
If you look further, you can see that despite having 3-4 times the amount of cores, ati's chips are in the same bracket when it comes to transistor count so there is no way those ati cores are even close to the same complexity as nvidia's.
Frankly, specification numbers haven't been a good way to estimate real life performance for quite a while... Nowadays, you need to understand the different needs of different applications and their respective bottlenecks to even make an educated guess of how these numbers will translate. If you're in a hurry, forgot all of those and just look up the benchmark for what you'll use it for.