The tests in this article were some baseline testing done to see if the power consumption has changed at all on recent kernel releases for an Intel Core i7 4770K "Haswell" system with HD Graphics 4600 (other system details mentioned on the OpenBenchmarking.org result file pages). The Linux power regression in the aforelinked story is from some older Intel CPUs in a multi-socket workstation setup, but for reference for Linux desktop users on modern Intel Haswell hardware is what's my focus with today's tests.
For this testing I ran some idle and load benchmarks on the mainline Linux 3.9, Linux 3.12, and Linux 3.13 Git kernels. All testing was facilitated via the Phoronix Test Suite and the AC power consumption was monitored in real-time using a USB-based WattsUp power meter that can interface directly with our open-source benchmarking software.
The idle data (and system software/hardware details and logs) can be found via 1312137-SO-INTELLINU17 on OpenBenchmarking.org.
For the i7-4770K system on the tested kernels there was no real change in power consumption when the system was idling.
There was no real change in power consumption (or performance for that matter in most workloads) when running various benchmarks. The load results are found within the 1312138-SO-INTELLINU33 result file.
Overall, there really isn't anything to gawk at for power consumption for this Haswell desktop system on recent Linux kernel releases, at least for the workload and software configurations tested.
Last month I also ran some extensive Linux kernel power tests from an Ivy Bridge ultrabook. Further Linux power testing will come in January.