nullstream weblog - UMPC Vista Battery Life Comparison Summary

« Heads up (UMPC) | Red Ring O Death »

UMPC Vista Battery Life Comparison Summary

Mobile Tech

May 9, 2007 03:55 AM PST

My device: Original Samsung Q1, 1G Ram, XP Tablet and Vista Ultimate dual bootable. I am currently experiencing a large difference in battery life between the two for the same use case.

Ok here is what I know...

I gave up on running full battery life discharge tests and I installed NHC (notebook hardware control) so I could watch the power being consumed as measured by the battery itself. This was very enlightening. Note: I'm only using NHC for monitoring. The CPU dynamic switching and voltage adjustment features are useless with the Celeron M. I found that nothing you do with the min / max processor % have any impact on power use. Setting max to 50% screws up the scale in task manager however and doubles you reported proc usage. Also with some tools (NHC included) leaving set like that makes it assume your proc frequency is moving around when it isn't. My recommendation is to just set min and max to 100% to disable all the monkey business. That's what I did for these tests.

For starters, yes it turns out my standard 3 cell battery is seriously messed up. Nasty wear factor, voltage that sags as the battery drains causing the power used to gradually increase. (Due to efficiencies dropping as the voltage falls below norms). It has a pretty non-linear discharge. Bottom line here, the 3 cell battery is a gonner, time to switch to my shiny new 6 cell that recently arrived. Its too bad because I like the form factor and weight with the 3 cell. Anyway this explains why my total run times don't align with the results others have had, but it still can't explain the repeatable battery life delta between XP and Vista on the same hardware, same battery with the same test cases. So time to just pay attention to power draw and do some direct comparisons.

I took lots of data with each battery, at different points with different settings etc. But rather than bore you with that, I'll attempt summarize in my usual rambling, long winded fashion.

Notes and disclaimers:
Power draw fluctuates a lot (more with Vista than XP) so I had to let things settle, and shoot for an average. I didn't bother trying to get actual data logging enabled since this was supposed to be a quick sanity test. The data was taken at minimum brightness, after the adaptive dimmer kicks in using my brand new 6Cell topped off. I am attempting to just measure battery life at idle. So I'm shooting for a best case here, not a worst case. And another disclaimer - I've disabled lots of services in my quest so my results may not match your system. The results are pretty consistent for me across multiple runs however. And yes I fully acknowledge that this could just be my hardware / configuration. Since others have reported similar battery life losses, however, so maybe this time I'm not just being an idiot.

Test Results
Vista: wifi on, bluetooth on, hard wired lan disabled: minimum brightness after auto-dim. Attempted to let the HD settle:
Typical min power use: 10.3 - 10.4 Watts

XP: wifi on, bluetooth on, hard wired lan enabled, minimum brightness after auto-dim. Easier to let the HD settle:
Typical min power use: 8.3 - 8.5 Watts.

That is a pretty serious difference. After a lot of playing around with settings and enabling / disabling stuff I found the biggest contributor to Vista's power use (on my Q1): Bluetooth!

Vista: wifi on, bluetooth disabled in device manager, lan disabled, min brightness after auto-dim...
Typical min power use: 9.2 - 9.3 Watts.

So over 1 watt used just to have the bluetooth radio on? I didn't even have any bluetooth devices connected so it should have been idel. I also have BT sharing and auto device connection turned off which should help.
My first thought was that I had something messed up with the drivers so I un-installed them, and re-installed them - same results.

Something is obviously wrong. Driver? OS install? Implementation? As usual just when I start to think maybe my hardware is messed up I remind my self that XP doesn't exibit this same behavior on this system. When I disable the Bluetooth on XP it only drops only a small amount and is lost in the noise (.1 - .2).

So turning bluetooth off gets me closer to XP power draw, but still nearly a 1 watt more. -Trying to find the source of this missing watt is a problem for another day.

Exchanging Bluetooth for battery life is not going to be an option for me since I have a large assortment of BT goodies I use with my computers. (Too numerous to list here, although it might be fun to line them all up someday for a photo shoot. Then again, no need to give my peers yet another reason to mock me).

I need someone else to help verify these results. Could there be a bluetooth power bug in Vista similar to the old XP USB bug?

Comments (5)
Brandon Paddock, May 9, 2007 04:11 PM:

If you disable Bluetooth in XP - do you see a similar drop?

Is this a clean install of Vista? Or has it had time to finish indexing and its other "moving in" tasks? As that could make a big difference.

John, May 9, 2007 06:00 PM:

I can hardly measure the drop in XP when I disable bluetooth. It gets lost in the noise.

Yes this is a clean vista install to a new partition. I have disabled indexing (sorry, no offense) for now as well as other services to try and help the battery situation.

Since the original Q1 is not officially supported with Vista there is a chance that the drivers are not optimized yet. All of us in the 'Vista Q1' community are using the same drivers though, so I'm hoping to get some verification on my results.

My next mission is to try and figure out what accounts for the 'other' watt of power difference while idle. If it is software based it is slipping under the task manager CPU radar. -Not that that is hard to do.

Paul, May 9, 2007 07:24 PM:

John doesn't need search indexing, he's a write-only kind of guy.

Brandon Paddock, May 9, 2007 11:27 PM:

Have you tried using the "Power Saver" profile?

John, May 10, 2007 09:29 AM:

Well, the current belief in the Q1 community is that the 'samsung' optimized power profile provides better management that Vista's built in 'power saver'. I of course have experimented with both extensively (or is that obsessively). I'm still waiting for someone to repro my bluetooth use results. (CTitanic are you out there?)

But I've been playing with trying to track down the last watt however. I'm using Russinovich's 'context switch delta' technique for finding processes that fly under task managers radar. Actually the new process explorer will give you a clock cycle delta in Vista which is very handy. So far I have not made any break-throughs. I've stopped services and suspended processes till the cows come home (or at least until Vista becomes un-usable) and so far no magic bullet. I will say though that explorer.exe seems to really be chewing up context time and cycles. Many many more context switches than on XP. (I can't measure cycle delta on XP). In fact it is so far off that it is probably that the sampling frequency is different and thus I can't make an apples to apples comparision. Of course what explorer.exe and the kernel use are excluded from the task manager CPU use measurements. (Who measures the measurer?)

It is still possible that there are more hardware driver issues at play like the one I found with the bluetooth driver, but I know all to well (from my former life) how much new code has been added to core Windows. All that code doesn't run for free.

All links will be marked with the nofollow tag, making them useless for search rankings. Any posts containing spam URLs will then be deleted.