From my experience, ccminer-alexis78-ms2013-cuda7.5 is a tiny bit faster than ccminerKlausT
I set intensity to 16.63, but if you get out of memory errors, you can try to lower that.

Overclocking both core and memory helps, but the exact clocks depend on what 1070 you have, some have higher TDP, some have different memory etc. (I have 3 different ones and they are all different) and it also depends if you want maximum hashrate or efficiency.
You should see more than 1200Kh/s if going for hashrate, and about 1100kh if going for efficiency.

Hsminer is about as fast, but also has a dev fee.