Testing done at Phoronix on the Linux 3.11 development kernel has found lower power consumption for some Intel systems (Ivy Bridge and Haswell) and other notebook testing has shown Linux 3.10/3.11 is better than the past.
I'm still in the process of expanded power consumption testing on modern Linux kernel releases, but it's a slow process with having a limited number of notebooks -- all of which are Intel CPU based, several of them being Apple systems, and most of the others being Lenovo ThinkPads. I've also done some testing of desktops with a USB-based WattsUp meter measuring the AC power draw in conjunction with the Phoronix Test Suite.
I'm in the process of writing up a multi-page article that shares more of my findings and results from a multitude of systems, which will likely come out in the next week or two. Yes, CONFIG_NO_HZ_FULL can play a role in power usage. Though for illustrating the mixed power results, today I uploaded results from an HP EliteBook that served as the Intel Software Development Vehicle (SDV) laptop for Sandy Bridge.
From the Core i5 notebook, the Linux 3.10/3.11 power consumption is actually a bit higher in some workloads than on Linux 3.8/3.9. Those new data points are available on OpenBenchmarking.org from 1308098-SO-INTELLINU35.
If you wish to support this timely and investigative testing for the continued Linux power consumption (and performance-per-Watt) situation, please consider making a PayPal contribution or subscribing to Phoronix Premium.