Another Intel Linux Power Regression Is Being Investigated

Power regressions are still easy to come by with the Linux kernel and other areas of the open-source stack... Multiple users have been reporting of a recent power increase on newer versions of the Linux kernel, which seem to track down to the Intel i915 DRM driver.

Going back to earlier this week has been reports of an Intel DRM kernel driver regression. Reported by Eric Rannaud, he explains, "Between 3.15.4 and 3.15.8, there was an increase in idle power consumption on Apple Macbook Pro 15 (late 2013) on a freshly booted system (no wifi driver loaded; brightness set to 4/100; X running; no desktop environment, except Awesome), from 6.5W to about 10.5W, as reported by powertop."

Eric bisected the changes in the kernel stable tree and found it was attributed to frame-buffer compression (FBC) now being disabled by default on Intel Haswell hardware and later. FBC was disabled in the Intel driver by default since it was causing black screen issues, slower performance, and other problems. While this was the believed cause, Intel developer Ville Syrjälä commented, "FBC works best when the screen contents don't change. The more activity on the screen the less effective FBC becomes. 4W sounds way too much for FBC however. 0.4W is closer to what one might expect from FBC based on my observations. 4W sounds more like the difference between min vs. max display brightness to me." But after playing around with the i915.enable_fbc=1 module parameter, Intel developers have been surprised with this four Watt difference over FBC support on Haswell.

Today, Sean V Kelley of Intel jumped in on the thread and reported with a mid-2014 MacBook Pro Retina using the Linux 3.16 kernel he's seeing definite Wattage differences as a result of the frame-buffer compression support, which amounts to a 4+ Watt difference.

As of right now, this power regression is still ongoing. There's been yet to be reviewed/merged patches for improving the Intel driver's frame-buffer compression support, but until that happens we're unlikely to see Intel re-enable FBC by default due to the adverse problems it causes some users.

I've started my own testing of the recent kernel releases on a few different Intel ultrabooks/laptops at my office. Interestingly, I'm also seeing a power regression between Linux 3.15 and 3.16 on an Ivy Bridge system (not Haswell and later where the FBC state was changed). Stay tuned for more Phoronix power testing as I jump to check in on the FBC power impact for different systems and to see what other power regression(s) might be lingering as well.