From an ASUS Zenbook (UX301LAA) with Core i7 Haswell CPU I ran some power monitoring tests of the battery when running Linux 4.4 stable on Ubuntu 16.04, Linux 4.5 Git, and also the Linux 4.6 Intel DRM-Next code (since for Linux 4.6 there is FBC and PSR enabled by default).
Everything else remained the same during testing... First up is just when idling from the Ubuntu-based ultrabook/laptop:
Linux 4.5 is lower than Linux 4.4 across the board compared to Linux 4.4. However, when pulling down the DRM-Next code for Linux 4.6, the results get even better! Linux 4.5's power use dropped by about 9% compared to 4.4 and Linux 4.6 looks like it could drop this system's power use by another 10~13%. That is if Intel doesn't end up having to disable Frame-Buffer Compression again like they have in the past due to regressions.
Along with lowering the power use, the CPU remains cooler on the newer kernels. The idle data can be found via this OpenBenchmarking.org result file.
The Linux 4.5/4.6 benefits though are namely when idling. Under load, there isn't much difference in the results.
You can look at the load data via this OpenBenchmarking.org result file. If there is enough reader interest, I can carry out some Linux 4.5 / 4.6 DRM-Next power tests on a few other Intel laptops/ultrabooks; let me know via the forums or on Twitter.