Thread: Touch Power code regression?
View Single Post
Old 02-13-2013, 08:18 AM   #39
Shai-Hulud
Silicon Book Worm
Shai-Hulud has much to be proud ofShai-Hulud has much to be proud ofShai-Hulud has much to be proud ofShai-Hulud has much to be proud ofShai-Hulud has much to be proud ofShai-Hulud has much to be proud ofShai-Hulud has much to be proud ofShai-Hulud has much to be proud ofShai-Hulud has much to be proud ofShai-Hulud has much to be proud ofShai-Hulud has much to be proud of
 
Shai-Hulud's Avatar
 
Posts: 129
Karma: 27430
Join Date: Jul 2012
Location: England
Device: Kobo Touch
Quote:
Originally Posted by GeoffR View Post
If it was a firmware problem, then wouldn't more people be encountering it?
I posted this with my eyes half shut at 0252 local time - there's dedication for ya What I should have said is with respect to two Kobo Touch's on ostensibly identical hardware platforms, yet with one running the Touch firmware and one running the Glo firmware, the one running the Glo firmware has a dreadful battery life in sleep mode (as also expresed by Mrs_Often et al in a different thread). Quite how far this extends to others I don't know which is why we post on these fora to begin with. Clearly you don't suffer from quite the dreadful battery life others are having with your Glo which is useful to know despite it muddying the water still further. But I'm pleased that someones device works close to as advertised!

My latest discovery was prompted by a comment made by Mrs_Often in response to my original query and which is incidental to it, yet corroborates what (some) others are seeing with their Glos/Touches and further suggests that there is more than one power issue involved here.

So far, observations reveal:

1. The issue I had that originally prompted this thread and which is (I'm guessing) the CPU getting stuck in a loop when trying to enter power-off mode and chewing through battery (but this aside is otherwise ok when left in sleep mode only).

2. The issue with Glo's and recent Touches that run the Glo firmware where the battery life is simply dreadful when left in sleep mode anyway BUT ALSO fail to reliably enter power-off mode when set to do so after 30 mins. Whether the "CPU looping" (for want of a better descriptor) also occurs periodically to compound this is unknown.

3. The use of linux kernel 2.6.35 which suffers from a documented power regression deficiency of around 20% when compared to 2.6.34 (as mentioned on Phoronix and other linux tech sites).

Now common sense says that if the newer Kobo Touches require a different firmware than the old models, then there has been some hardware change made inbetween. But until Kobo acknowledge this publicly or it is otherwise proven to the contrary, we have to assume that hardware differences are a red herring.

With regard to the kernel: yes 2.6.35 is a bad choice for power comsumption generally but as both the old and new models (appear to) have the same kernel there is still a relative change between old firmware and new firmware to the detriment of battery life.

In conclusion (based on MY experience) all I can say is this:
- On the Old Touch, this battery eating state occurs intermittently (actual trigger unknown) when the device tries to enter power-off from sleep (15m/30m) and fails to do so. When set to Sleep mode only, this doesn't to seem to unduly affect battery life.
- On the "new" Touch, this intermittent failure to enter power-off mode from sleep also occurs and depletes the battery life just as quickly BUT the same behaviour is on by default if the device is set to sleep-only. Annoyingly I can't think of a way to determine if the battery is under excessive draw whilst reading normally, short of following the "30 mins a day" for a month with an explicit power off after each session.

And frankly all this monkeying about is getting in the way of me "just reading" which is what I bought the bloomin' thing for!
Shai-Hulud is offline   Reply With Quote