suspend status

[Nate Lawson]

Tobias Roth wrote:
> On Fri, Dec 10, 2004 at 08:56:06PM -0600, Eric Anderson wrote:
> 
>>Refigure your math - if 100% cpu means 3hrs 18 minutes of runtime left, 
>>then that about 200 minutes of runtime.  So 1% equals 2 minutes of 
>>runtime roughly.  So, you suspend - and wait up 90 minutes later.  If it 
>>would have been running like normal, it would eat up 1% per 2 minutes, 
>>so about 45% of your battery - but it didn't, it only ate up 20%.  So 
>>ath that rate, it was using less than half the power as when in 
>>non-suspend mode.
> 
> how long would the same laptop/battery survive when suspended from
> windows? i always had in mind that a suspended laptop is supposed to
> live for more than a day, which clearly is not the case in your example.
> 
> you often hear comparisons here about how much less battery windows uses
> when compared to FreeBSD (or rather, how much better windows battery
> saving techniques are). detailed comparisons of bsd <-> linux <-> windows
> with good guesses of why the discrepancies are there would help.
> 
> i am just trying to say that battery saving in suspend probably IS bad
> in FreeBSD (as compared to the possible optimum, as windows shows it).
> it's not just bad math in the above example.

I'd like it if people took the time to compare various features on 
windows/linux/freebsd including average temp during your normal use, 
battery usage, etc.

Both my ThinkPads last a very long time in S3 (weeks).  However, I think 
BIOS code handles most of the system power issues which may hide things 
that the OS doesn't do as well.  If a single device driver doesn't 
properly power down a device, it may be the culprit.  Such problems are 
hard to find, obviously.  I do know that some display adapters need more 
  magic than X is currently using.  Also, bms@ pointed out that we 
possibly need to hook X into suspend/resume more.  I started some 
patches for that (on my web page) but didn't get very far since I don't 
know the X server code and it wasn't clear that the module in question 
(bsd_apm) is even compiled on FreeBSD.

The other thing you can try is my suspend power patch (committed in 
-current, patch posted recently here for 5.x).  It may help since it 
explicitly powers down/up acpi and pci devices in suspend.

-- 
Nate