Possible SATA Power Management Improvements For Linux

[phoronix]

Phoronix: Possible SATA Power Management Improvements For Linux

Linux kernel developer Matthew Garrett has revised his work around improving power management for modern Intel systems via SATA...

http://www.phoronix.com/scan.php?page=news_item&px=SATA-Linux-Power-2016

[davidbepo]

i hope they get merged that would help extend my laptop battery (i5-5200u cpu) i think it will make very small ddiference but any improvement is welcome

[ArthurBorsboom]

What makes you believe it will make a very small difference?

[milkylainen]

Maybe this will unf-ck my Skylake laptop loosing it's SSD when coming out of suspend every now and then.

[davidbepo]

What makes you believe it will make a very small difference?

i dont think the sata link can consume more than a ssd and my ssd (850 evo)uses 3 watts

[ArthurBorsboom]

Honestly I have no clue either, so I have searched a bit.

According to this presentation of Intel (2007), SATA power management can save on average around 0.6 Watts and up to 1,5 Watt per SATA channel. My laptop has two SATA channels, so based on this information, the maximum estimated power savings could be 3 Watts on my laptop.

Does anybody have more recent power savings numbers?

[FLHerne]

The point is that in recent generations the CPU can't enter its lowest sleep states without the SATA link being powered down.

It's not just the direct savings from the SSD and SATA controller.

[Luke]

I ran into a world-class ugly issue when retrofitting one and old Sandforce-based SSD to a Pine Trail (circa 2009-2010) Intel Atom netbook: If the machine tried to go to sleep while running software updates (this being possible is a bug elsewhere!), the entire root filesystem would become badly corrupted, easiest fix was to restore from one of the backup images I keep. Workaround was to put the sleep inhibit applet in the mate-panel and always disable power saving while running an update job. Not enough CPU to do much else while that is running, so updating in background out of the question.