[ltp] X60 - (bad) experience, others?

[Nils Faerber]

Wow, thanks for the lengthy reply!
More below...

Theodore Tso schrieb:
> On Tue, Jan 29, 2008 at 11:41:00AM +0100, Nils Faerber wrote:
>> The standard battery that comes with it only lasts for 1.5h to max. 3h,
>> but 3h means almost no CPU activity, backlight at lowest possible
>> setting (almost unreadable) or off and all wireless turned off. If it
>> shall be halfway usable the maximum I can get is 2.5h. Disappointing for
>> a modern expensive mobile hardware - compare this to e.g. Apple's notebooks.
> 
> There's nothing magical here.  It's all about battery size.
[...]
> 13.3-inch Macbook air (1.6Ghz)	      37 Watt/hours (5 hours @ 7.4 W/hr)
> 13.3-inch Macbook  (2.0Ghz)	      55 Watt/hours (6 hours @ 9.1 W/hr)
> 15-inch Macbook pro (2.4Ghz)   	      60 Watt/hours  (6 hours @ 10 W/hr)
> 17-inch Macbook pro (2.2Ghz)	      68 Watt/hours  (5.75 hours @ 11.8 W/hr)

Uh, OK, did not know that there is such a huge difference in the pure
capacity between Apple and IBM/Lenovo. This of course explains the huge
difference in run-time (I never owned a power book, I should also say...).

> The battery life times are taken from the Apple web site, and are the
> standard marketing "up to" lifetimes.  The battery life times I've
> quoted above is for my X61s, with backlight at 40%, and machine in
> "airplane mode" (USB, wireless turned off, system configured in
> laptop_mode, etc.) where I have my power consumption tuned down to 9.3
> watts/hour.  If I take the backlight down to 10%, I can get it to 8.8
> watts/hour, but that's not really a usuable configuration, where as
> reading e-mail while on an airplane at 40% backlight and minimal disk
> usage is at least a believable scenario.

With 2.6.22 I also got down to ~9.5W-10W which is quite OK. But this
involves a lot of manual tweaking and "not doing" caertain things. The
X31 I had earlier did not need that careful tweaking. That's what annoys
me a litttle about those new machines. If you do not pay close attention
then your batery run-time is just a quarter of what could be achieved.

> The 4 cell "standard" battery is basically simply too lightweight for
> what you want.  But in terms of actual power usage, the X60 series
> aren't that much different from what Apple has; Apple is just simply
> giving you a bigger battery --- and true to Steve Job's design
> asthetic, he gives you only one choice. 

He ;) Having a non-replaceable battery would annoy me pretty much too.
And by now I also bought the extended battery and using kernel 2.6.22
without Wifi and low backlight it survives quite long, yes.

> Lenovo gives you three possible batteries, plus the option of
> attaching an 2nd, "extended life battery" to the bottom of the laptop,
> connecting via the dock connector.  If you use the 8 cell battery plus
> the extended life battery, you can get 11 hours of battery life.  :-)

Uh, sounds appealing but expensive ;)

> Now, that's not to say that we can't do better; MacOS is much better
> at automatically enabling low battery configurations where as I got
> down to 9.3 W/hr by aggressively tuning my system.  But it's not like
> the Apple hardware is fundamental superior; it's just software, and we
> can fix software, since it's Open Source.  :-)

This is surely true and I should have been more specific. Several years
ago I was impressed by the battery lifetime of the older power books
still using the PowerPC CPUs. There even Linux had quite long run-times
without extensive software tweaks and MacOS was even better. And they
did not have fans... Today with Intel hardware inside Apple's notebooks
are not much better off than Thinkpads, true.

>> The whole device is always quite warm, especially the region below the
>> right handrest - I suspect that WiFi is sitting there because it gets
>> even warmer when power saving on WiFi is not enabled or if it is used
>> more intensively.
> Yes, that's where the WiFi is.  You need to completely unload th WiFi
> driver if you want the biggest battery savings, BTW.  I've already
> complained to folks I know at Intel about this.

Just for the record: Just tested this with 2.6.22 and the ipw3945
driver: No change at all. The IPW driver works as expected, i.e. when
downing the interface, putting it in power-save mode and enabling
rf_kill it is already in lowest power state.
And with the latest compat package of IWL drivers for 2.6.24 it also
makes no difference there too, i.e. enabling rf_kill has the same power
saving effect as to rmmod the module.

>> The warmth also results in an ever spinning fan which is additionally
>> also quite noisier than the ones in X23, X24 or X31 (which were almost
>> inaudible at their lowest speed).
> Faster processors means more heat.  :-)

But newer models do not necessarily mean noisier fans :(

And that the X60 CPU cannot be clocked below 1GHz is quite rediculous.
And disabling undervolting and underclocking is also not very helpful
(pre Core2 CPUs were able to undervolt/-clock which significantly
reduced heat - why did Intel disable this?).

>> The situation with Wifi drivers is not improving over the last 1/2 year.
>> There is the IPW3945 driver which works quite fine but requires non open
>> source parts and is difficult to install. The newer open source IWL3945
>> does not work well (I have issues associating to APs) and I have the
>> impression that its power saving is not that good too resulting in even
>> less battery life. And since the release of 2.6.24 neither IPW nor IWL
>> compile out of the box - heck, what's going on there? IPW will not get
>> updated and there are third party projects to patch the latest IWL
>> drivers for 2.6.24 - the IWL driver in the stock kernel 2.6.24 is at
>> least three months old.
> Yep.  Blame the intel driver folks.  I'm not sure what their
> development methodology is, but their dev tree isn't well synched with
> the mainline tree, for some reason that I've not been able to
> determine.  This means that merging the latest IWL driver with the
> latest kernel is not trivial.

Seems like it - and I do not have the motivation to dive into that big
chunk of code either.

>> And while we are talking about 2.6.24 instead of getting better battery
>> performance that was rumored to come with cpuidle I get up to 4W of
>> battery drain - additionally. The lowest I can get with dim backlight
>> and no Wifi on 2.6.22 is about 9.5W - with 2.6.24 it is about 13.5W. I
>> found, using powerop, that one of the reasons might be that 2.6.22
>> manages to keep the idle CPU in C3 for 95% of the time while 2.6.24 C3
>> is only used for maybe 20-30%.
> 
> I have not seen a power usage regression on 2.6.24.  The power
> utilization at idle for me is still around 9.3 Watts in my "airplane
> mail reading scenario".  I was in C3 98% of the time, with an average
> residency in C3 of about 50ms. 

This is really strange!
This always worked for me and I use almost the same kernel config for
2.6.24 and 2.6.22 - but now only 20-30% in C3 at idle.

> If you're not getting that, it's probably a matter of software
> configuration issues.  Or, try suspending (to ram) and resuming the
> laptop, which seems to synchronize the timers to increase the average
> residency time.  See my blog entry here:
> 
> http://thunk.org/tytso/blog/2007/10/29/tip-o-the-hat-wag-o-the-finger-linux-power-savings-for-laptop-users/
> 
>  ... for more details.

Update:
I basically found the guilty: uhci_hcd!
The strange thing is that auto-suspend is enabled (and also on all
devices) and that I did not have to remove it for getting 9.5W with
2.6.22. So something from 2.6.22 -> 2.6.24 changed in uhci_hcd and the
change was not good, i.e. causing the CPU not entering C3 in 70-80%.
Again, this was not necessary with 2.6.22...

[...]
> Well, I like having my environment on one laptop, and not have to move
> files back and forth between a "desktop" and "laptop" computer.  And
> unfortunately I fly enough that I want my laptop to be my primary
> computing environment.  It helps that I mostly fly American Airlines,
> which has power plugs available even in the Cattle-class seats, but my
> main answer is just to bring more batteries with me.

He, if you can afford enough... if I pay 1100EUR for the laptop alone
extra batterie_s_ quickly add up to a respectable sum. And since
batteries have a limited lifetime (in contrast to the notebook itself)
this should be considered very carefully.

And while we are at batteries: I just also found that my X60 discharges
the battery even when switched off!? Two days switched off eats almost
50% of the standard battery.
Is that normal? If not I should probably check that more thoroughly and
call Lenovo service :(

>> When I was at OLS last year the most important feature request of Intel
>> people there was thermal management in the kernel - for sure, because
>> their hardware is always over-heating! They presented a prototype
>> hardware in the form factor of a PDA with their newest and latest
>> low-power mobile CPU - wow. Guess what, it had a fan! Sorry, but this
>> cannot be honestly state of the art in low power CPU design. 
> The problem is the tradeoff between speed, power, and heat.  Sure, you
> can get something like the Intel XScalre (aka arm) processor --- but
> then you will have people complaining about how slow the machine will
> be.

There are lots of levels in between. The XScale (and similar parts like
the TI OMAP, Samsung S3C series etc.) are made for really low battery
devices liek mobile phones. But even there they do a great job but are
not so general purpose as a notebook, sure.
What I am wondering is why do companies like Intel always want to brag
with MHz and GHz instead of adding a second line of really low-power
optimized CPUs? The X31 I mentioned already had a 1.4GHz CPU with old
technology and already quite power savy. Shouldn't it be possible to
make a CPU with cutting edge technology (smaller processes, smarter
architecture with more than one clock for different functional units -
just as needed, independantly en- and disabling functional units, etc. I
can live with a slower CPU if I do not get a lap heater and have longer
battery life - 99.9% my CPU is idle anyway.

I wish we could get back static designs again ;) When idle simply switch
it off completely...

>> Oh well...
>> If you feel you have a great tip to save power and/or prevent heat
>> dissipation on the X60 I would be glad if we could share those tips here ;)
> 
> See my blog entry (listed above) for a whole bunch of tips.
> 
> My suggestion to you would be to get a bigger battery --- I would go
> for the 8 cell extended capapacity battery ($179 on the web, $143.20
> if you can find an IBM/Lenovo employee to give you a discount code),
> and the extended life battery ($179 on the web, $143.20 EPP
> friends/family price), and just be happy.  Yeah, it's heavier, but
> remember, an Apple MacBook with the 13.3 inch display is 5 pounds.

The extended battery I got from ebay is quite OK.

> The x61 laptop is 2.73 pounds with the smallest battery (weighing 0.44
> lbs).  The 8 cell battery weighs 1.05 pounds, so with the 8 cell
> battery the x60 laptop weighs in at 3.3 pounds, and if you add the
> extended life clip-on battery, it's another 0.82 pounds, for a total
> eight of 4.1 pounds.  At that point, you can have a 3.3 pound laptop
> with 8 hours of battey life, or a 4.1 pound laptop with an 11 hour
> battery life.  And you can feel superior to the person in the next seat
> over with a 5 pound Macbook laptop with only a 6 hour battery life.  :-)

Last time at some airport I saw a really thin and tiny young lady in the
row opposite to me. She put her quite huge bag on the floor and pulled a
laptop out of it - but what a monster! Must have been some 17" 16:9
display, all buttons glowing blue, glare screen etc. - the most ugly
things I have ever seen! This machine was almost bigger than the whole
woman. And must have had almost the same weight...

So... after I found that 2.6.24 UHCI behaves worse than 2.6.22 I can
also get down to 9.somethingW just that Bluetooth also does work without
it. And with power_level=5 even iwl3945 behaves concerning power drain
(~10-11W associated with an AP 2m away). So that is basically OK again.

Probably I diff through .22<->.24 uhci_hcd...

Anyway... many thanks for the replies so far!
At least they cheared me up a little... and sorry if I mourned too much.

> Regards,
> 						- Ted
Cheers
  nils faerber

-- 
kernel concepts GbR        Tel: +49-271-771091-12
Sieghuetter Hauptweg 48    Fax: +49-271-771091-19
D-57072 Siegen             Mob: +49-176-21024535
--