Linux 4.7 Notebook Power Testing: Possible Regression When Idling

With the Linux 4.7 kernel expected to be released this weekend, I did some Linux kernel comparison power measurements from a Lenovo ThinkPad X1 Carbon notebook. Here are those results.

The X1 Carbon with Core i7 Broadwell processor is sadly the newest notebook/ultrabook I have in my possession, so I used that for testing. From Ubuntu 16.04 I tested the mainline kernel PPA builds of Linux 4.4, 4.5, 4.6, and 4.7 Git as of yesterday.

Here's what I found while monitoring the power draw on battery during a period of idling for five minutes, automated via the Phoronix Test Suite:

The 4.7 Git behavior was noticeably different then earlier kernels. The 4.4/4.5/4.6 kernel results were largely flat while the 4.7 Git results were fluctuating a bit and never flatlined during this automated process... The P-State performance driver/governor were used and all other system settings the same along with making it easily reproducible by using the Ubuntu Mainline Kernel PPA. So the average power draw with Linux 4.7 was 7.9 Watts while idling with this ThinkPad compared to ~7.2 Watts with earlier kernels.

While measuring the battery power rate during the course of many benchmarks, it ended up being comparable to the earlier kernels. You can see those load benchmark results and power data via this OpenBenchmarking.org result file.

Today and tomorrow I'll do a similar idle power comparison on my other notebooks (Haswell era) to see if the results are similar to this Broadwell notebook on Linux 4.7. Have you experienced anything similar with Linux 4.7? Feel free to share your thoughts by commenting on this article in the forums.