X200s - amazing power consumption

[ThinkRob]

Update: Turns out it can get a bit better: 5.7W is my new record.

I'm really, *really* impressed with the power efficiency of my ThinkPad X200s.

How awesome is it? It idles at 6.5W with 8/14 bars of brightness and WiFi on.

Now it's true that I did configure my system to be as power efficient as possible (ULV chip, LED backlight, low-draw SSD, etc.) Still, that's quite impressive. That means that the 6-cell battery will last for a US/UK flight! It actually could do a bit better, but some of the grsec/PAX options that I've built my kernel with tend to increase CPU usage (and thus power consumption.) I don't have a 9-cell, but I'd estimate that it could let me break 13 hours.

Screencap:



Hardware:
SU9300
LED-backlit WXGA+
128GB Samsung SSD
IWL5100
Ericsson F3507g

My kernel is based on 2.6.32.7 with the following patches applied:
1) SATA-poweroff patch - powers off unused SATA ports (ported from 2.6.27)
2) grsecurity
3) CK patchset (-ck2)
4) i915 FB range patch for grsecurity (allows the use of CONFIG_GRKERNSEC_KMEM to deny writes to kernel memory)
5) CONFIG_COMPAT_VDSO removal

The result? A light, long-lasting ThinkPad with a exceptionally-hardened, highly-responsive kernel.

Now as soon as I fix the bendy keyboard issue, I'll be all set. (I've located a 42T3737, but until it gets here I've shimmed my existing backplate to make it somewhat bearable.)

[Flintz]

Which distro are you using?

Interesting topic as I am thinking about switching to Linux as well.

How did you deal with the Ericsson WWAN card? Did you disable it? Because after a short test I found out that between all wireless switched on and all switched off (via the hardware switch on the left side of the notebook) I had a difference of 3-4 Watts in power consumption. However I didn't figure out yet which wireless is the reason but I am guessing the Ericsson card.

Do you have DRI/Effects enable in X11?

[ThinkRob]

Which distro are you using?

Well at the time I took that screenshot, I'd have to describe it as Jaunteeze -- a hacked-up combination of Ubuntu Jaunty and Debian Squeeze.

Currently, I'm running Debian Stable with the same kernel, and there's no notable difference in power consumption.

How did you deal with the Ericsson WWAN card? Did you disable it? Because after a short test I found out that between all wireless switched on and all switched off (via the hardware switch on the left side of the notebook) I had a difference of 3-4 Watts in power consumption. However I didn't figure out yet which wireless is the reason but I am guessing the Ericsson card.

There's a difference of about 1W for me. YMMV.

Do you have DRI/Effects enable in X11?

DRI yes, compiz/XFCE compositing no.

[Flintz]

I found out that it wasn't actually the Ericsson card which increased the power consumption this much.
The stock wifi drivers in Ubuntu 9.10 don't support power management, so as soon as I compiled the driver myself everything was fine.
I achieved a power consumption of around 7,5W but I am booting linux from a USB pen driver, so this is increasing the power consumption a bit.

Maybe I will install Ubuntu on my SSD soon and can post some more information then.

[ThinkRob]

I found out that it wasn't actually the Ericsson card which increased the power consumption this much.
The stock wifi drivers in Ubuntu 9.10 don't support power management, so as soon as I compiled the driver myself everything was fine.

Yep. It looks like power management isn't enabled for the 5100 card in 2.6.28. I'm running 2.6.32.7, and I can certainly confirm that is both enabled and effective.

I achieved a power consumption of around 7,5W but I am booting linux from a USB pen driver, so this is increasing the power consumption a bit.

Probably quite a bit, actually. USB 2.0 is fairly power-hungry (although 1.1 is much, much, much worse since it requires polling), and since it's CPU-driven it will likely keep you out of the deeper C states even under a fairly light load.

It's also worth noting that the drive that I'm using is quite power-efficient. Not all SSDs are though, so YMMV.

Also, do you have an LED-backlit screen? If you've got a CCFL backlight, I'd expect it to suck up quite a bit more power.

[ThinkRob]

Here's the SATA power power-off patch that I use.

From my experience it has a small, but worthwhile impact on power consumption. It will apply to 2.6.32.7, but will require fuzz.

http://pastebin.org/85937

Enjoy!

[Flintz]

Ok, I installed Ubuntu 9.10 yesterday on my harddrive.

Surprisingly the power consumption didn't go down as much as I thought. With the stock 2.6.31 kernel and recompiled iwlagn driver I am now at around 7.4W power consumption for total idle, wlan activated and screen brightness at minimum.

i still have a CCFL screen, so I guess I won't reach your power consumption with half the birghtness.

Anyhow, I know from my windows installation that my particular X200s can reach around 6-6.3Watts in idle. My aim now is that I reach something around or just below 7W.
I think the only chances I have is to try the kernel patches you used.

BTW, did you manage to geht a visual feedback when pressing the MUTE button? It works here but the gnome icon doesn't change and I also don't get that fancy black notification I get when pressing volume up or down.

[ThinkRob]

Ok, I installed Ubuntu 9.10 yesterday on my harddrive.

Surprisingly the power consumption didn't go down as much as I thought. With the stock 2.6.31 kernel and recompiled iwlagn driver I am now at around 7.4W power consumption for total idle, wlan activated and screen brightness at minimum.

IIRC, 2.6.31 lacks some of the power consumption improvements that were added in 2.6.32.x. Further, Ubuntu's "stock" kernels tend to be *very* different from the vanilla versions of the same. Just FYI.

BTW, I assume you've tried all the usual tweaks? (Things such as Intel HDA powersaving, using hdparm to enable drive powersaving, vm sysctl tweaks, USB autosuspend, etc.)

i still have a CCFL screen, so I guess I won't reach your power consumption with half the birghtness.

Perhaps, but I'd be surprised if that accounted for a whole watt's difference. Do you have an LV chip or a ULV one?

Anyhow, I know from my windows installation that my particular X200s can reach around 6-6.3Watts in idle. My aim now is that I reach something around or just below 7W.
I think the only chances I have is to try the kernel patches you used.

See above. Also, check the tips over at ThinkWiki.

BTW, did you manage to geht a visual feedback when pressing the MUTE button? It works here but the gnome icon doesn't change and I also don't get that fancy black notification I get when pressing volume up or down.

I don't use GNOME, so I can't comment on what it does/does not support. I do use TPB (built from SVN), and it detects mute just fine. You might want to try passing the boot parameter acpi_os="Linux" to work around one of Lenovo's BIOS quirks for the X200 series.

[harrigan]

Interesting thread, thanks for posting your info. For comparison I ran a similar test and got 7.1W.

Vanilla 2.6.32.7, 8 bars of brightness, WIFI associated, xterm, conky.

Hardware:
SL9600
LED WXGA+
32GB SSD
IWL5300
no WWAN

Do you have any idea how much of the difference can be attributed to the SL vs. SU CPU?

[Flintz]

I downloaded 2.6.32 kernel and now I am at 6.1W idle! Yeeeehaw (With minimum brightness though), still I am quite satisfied.

Thanks for the input ThinkRob, now I can enjoy my x200s really under linux

[ThinkRob]

Do you have any idea how much of the difference can be attributed to the SL vs. SU CPU?

I wouldn't be surprised if the LV CPU accounted for the 0.6W difference.

Also, I don't know what SSD you're using, but if it has a higher idle draw, that also might make a difference.

[Esben]

Incredibly low power consumption you've got there, thanks for sharing. I thought Linux was behind when it comes to minimizing power draw, but I can see that I'm wrong. For comparison my X300 with Windows 7, minimum brightness and WLAN/Bluetooth off can idle at ~7W.

[ThinkRob]

Incredibly low power consumption you've got there, thanks for sharing. I thought Linux was behind when it comes to minimizing power draw, but I can see that I'm wrong..

In terms of actual capability, for most ThinkPads it's on-par with Windows. There are some notable exceptions (specifically a few of the ATI cards don't do power management yet), but for most of them the differences should be negligible (this is especially true for the newer Intel-based machines.)

That said, an out-of-the-box install of Ubuntu or some other distro will likely consume more power than a brand new factory-provided install of Windows -- but that's not due to a limit of Linux as much as it is the fact that the latter has been pre-tuned by the manufacturer whereas the latter is a more generic configuration. If you're willing to go on ThinkWiki and make a few tweaks (as well as do a bit of Googling), it's quite realistic to reach the same levels with a modern distro as you can with Windows. Just for reference, the tweaks that I made to my machine took about 20 minutes, including the time taken to look up the stuff online, etc. It took a little longer to compile my own kernel, but given that I do that anyway (so as to use grsec/PaX) it wasn't that much of a hassle. Someone who didn't need the custom kernel patches that I do could easily just install a pre-compiled kernel to reap the same power-saving benefits.

[harrigan]

Also, I don't know what SSD you're using, but if it has a higher idle draw, that also might make a difference.

It's an OCZ Vertex which claims a 0.5W idle.

Out of curiosity, I applied the grsec and ck2 patchsets and it bumped up my idle consumption to 7.2W. Killing off conky dropped my wakeups-from-idle from ~80/s to ~30/s and my idle consumption to 6.9W.

[ThinkRob]

Out of curiosity, I applied the grsec and ck2 patchsets and it bumped up my idle consumption to 7.2W. Killing off conky dropped my wakeups-from-idle from ~80/s to ~30/s and my idle consumption to 6.9W.

Conky can suck down battery if you've got it configured to update too frequently. I've got my conky config set to update every 4 seconds, and it produces almost no overhead, wakeup-wise.

Grsec's effect on wakeups is entirely dependent on the features you select. IMHO, PaX's free-memory scrubbing is one of the larger drains, since it adds overhead to every memory allocation (thereby causing more bus traffic, more CPU usage, etc.) I don't think any of the other grsec features have a notable impact on battery life, but I haven't benchmarked it to be sure. It's kind of a non-issue for me, since I don't see running without grsec as an option.

I didn't get a chance to play with the ck2 patchset too long before it started causing problems (it doesn't seem to play well with my grsec config, plus it was causing sporadic panics on module loading/unloading), and since security is far, far more important for me than the latest shiny interactivity patchset, out it went.

[ThinkRob]

Ok, new low for this machine. After playing around with the vm sysctls, I was able to get a bit lower consumption when kinda-idle and with my backlight on with low brightness (2 or 3 steps IIRC). Wireless was on, WWAN off (i.e. card totally powered down).

5.7W:


I think that's about the best I can do without switching to a new drive and/or ditching PaX.

BTW, harrigan, did the SATA poweroff patch make a difference for you? I'm pretty sure it produces a drop for me, but I'd like some outside verification (else I'll drop it from my patchset.)

[harrigan]

BTW, harrigan, did the SATA poweroff patch make a difference for you?

Maybe a tiny amount. Before the patch it would settle at 6.9W and then stay there. After the patch it bounces between 6.8W and 6.9W.

What sysctls did you use? That's a pretty good drop because at 2 steps brightness my system only drops to 6.6W.

[ThinkRob]

Maybe a tiny amount. Before the patch it would settle at 6.9W and then stay there. After the patch it bounces between 6.8W and 6.9W.

That's pretty consistent with what I've experienced. It seems to drop it by about 0.1W. Guess it's worth it, given that it has no overhead and requires no user intervention.

What sysctls did you use? That's a pretty good drop because at 2 steps brightness my system only drops to 6.6W.

Mainly writeback frequency, but also dropped swappiness to 0 and some other things that I don't recall at the moment.

Keep in mind though, you have a SL9600, so that's gonna draw more at idle than the SU9300. I also think your SSD might be a bit more power-hungry, so that could be part of it too.

[harrigan]

I also think your SSD might be a bit more power-hungry, so that could be part of it too.

Indeed, the SSD was mostly to blame. I swapped in another drive with an almost identical OS image and now I'm idling at 5.9W with 2 steps of brightness.

[ThinkRob]

Indeed, the SSD was mostly to blame. I swapped in another drive with an almost identical OS image and now I'm idling at 5.9W with 2 steps of brightness.

Nice. After some further userland customization, I'm idling at 5.7-6.1W.

[lordt8]

I just got my x200s, 7466 3SG with the optional WWAN module,
if your not to keen on going after the specs i'll spell them out
CPU:1.86 GHz
RAM: 4GB
HDD: 250 GB 5400rpm
WIFI: intel 5350
WWAN: ericsson 5430g
LCD: 1280x720 LED
optional extras: webcam, turbo memory, modem, fingerprint reader
battery: 4cell, 9cell on order will ship in 8 days
os: windwos 7

well that covers the basics.
First the 4cell is from another x200s, it was cycled 15 times in about 11 months, its at 97,5% original capacity.
Well why am writing this post now?
POWER USAGE,
i got this cos my sony tx3 died after 5 yeats of service, that thing had a core solo 1.06 ghz cpu 11,1" lcd, it was using about 7 watts most of the time.
The x200s draws above 13 wattas at anygiven time. At first i tought bloatware, so i did a clean win7 install, hasn't got better, the cpu usage is 1-2% idle, the hdd spins down after 5 minutes, lcd is at 8/15, wifi on. And i get 13-14 watts power draw! I read that ppl get sub 7 watts draw on other forums! I killed every possible service, indexing, defrag, everything and can't get bellow 11 W during use, it drops to 9 sometimes when idle with the screen dimed but thats not good! the 4 cell with that power draw manages about 2h. So with some clever math i calculated that the 9 cell would last for about 6.5h, thats way lower that the reviews!

So please if you have any idea what am doing wrong please tell me i need all day computing and want the machine to work for atleast 9 h when i get the 9 cell.

Thx and sorry for the long post

[lordt8]

sorry for 2xpost but really need some info

[harrigan]

I don't use Windows so I can only offer a couple of suggestions.

Have you checked the BIOS->Config->Power settings?

Something else you could try is turn off the radios, remove the turbo memory and WWAN card. If any of those cause a significant drop in power consumption, it would at least give you somewhere to focus your efforts.

[Flintz]

Also try switching off the Areo-Effects by going to the "Basic" Theme (I think it's called that way).
I had Win7 64bit on my X200s before and I got a power consumption of ~6.5W in idle and lowest screen brightness.

[lordt8]

i got a new clue, the intel processor identifier says enhanced halt state is disable on my CPU... how to enable that?

[ThinkRob]

i got a new clue, the intel processor identifier says enhanced halt state is disable on my CPU... how to enable that?

Any chance you have a Celeron M rather than a Core 2 Duo?

[lordt8]

nope, proc id says core 2 duo, 6mb chace and so on

[ThinkRob]

nope, proc id says core 2 duo, 6mb chace and so on

Huh.

Perhaps the SLxx00 chips don't support it? The ULV ones (SUxx00) sure do -- my X200s dropped into C6, no problem -- but I have no idea about the LV models.

[harrigan]

With my SL9600 it appears to only drop into C6 while on battery.

[ThinkRob]

With my SL9600 it appears to only drop into C6 while on battery.

Hm. That might be it.

Also, check the BIOS settings -- I suspect (though cannot confirm) that selecting different AC/battery power/performance settings might have an effect on this sort of thing.