Testing For The Latest Linux Kernel Power Regression

The third and final system used for this quick kernel power testing was an ASUS Zenbook with a Haswell Core i7 4558U processor sporting Intel Iris Graphics 5100. This system is actually the main system I use for production purposes and was connected to a 2560 x 1440 display while running Ubuntu 14.04 LTS. With this being my main system, I only tested it against the Linux 3.15 and 3.16.1 kernels (with choosing Linux 3.16.1 vanilla since that's what I've been running on it).

When running Tesseract between Linux 3.15 and 3.16 the power consumption is up by about two Watts. I also tested this system when forcing frame-buffer compression and then on the same kernel it dropped by about a half-Watt.

The Tesseract performance creeps lower on the newer kernel...

Next up was the LAME MP3 encoding test and here the average power consumption with the Linux 3.16 kernel was slightly higher than on the Linux 3.15 kernel but it amounted to just over a one Watt difference. Intel frame-buffer compression had no big impact here.

The one Watt difference might not be attributed to CPU scheduler differences with the Linux 3.16 kernel with no apparent performance changes.

This investigation is still ongoing so there isn't much more to add right now... Long story short from these first three systems tested, there is a power consumption difference between Linux 3.15 and 3.16 commonly exposed by running the LAME MP3 encoding program via the Phoronix Test Suite. For one system this power increase came with improved performance while on the other two there were no major differences.

On the matter of Intel driver and frame-buffer compression, I haven't found any hugely different results at this time.

My tests are still ongoing so keep checking back on Phoronix in the days ahead as I find the time to test out more systems and investigate further.