Best Power Output Results (include which Kernel)

dmdsoftware · **Posted:** Sun Mar 11, 2012 3:26 pm

I've been trying to tweak my Linux Mint Debian (LMDE) installation on one of my X61T as a side-project. I would really like to switch all my home systems to linux.

I attempted this switch back in 2005 as well, and the biggest hurdle then was power management. I only run laptops at home, and power management is one of the key ingredients in desktop Linux systems.

I'm re-attempting the move as I know there has been more emphasis on power management by linux developers. I'm currently trying to optimize a 3.0.0 kernel release.

With moderate backlight, with all tunable parameters switched to good in PowerTop (at least the ones that will switch), and running on 800mhz (cpufreq powersave) with moderate screen brightness + all non-used perhapherals turned off, I get about 15W power usage on my X61T with L7300 processor. This seems quite high for a low voltage cpu.

I did do some reading but a lot of the documentation and tips/tricks tend to be out of date. In particular the lesswatts.org which has a lot of great tips on patches etc, but most of the specific tweaks are for early 2.6 kernels with comments such as "this is not necessary on more recent distributions". Anyone have any more recent sources of information that I should follow.

I did try a kernel that enables undervolting, but it seemed it was getting higher power usage since the cpufreq wasn't working on it. I may attempt this again, but it's one of the lower items on my list since my laptop already uses a low-voltage cpu.

One of the areas I know I can shed at least a watt or two is on the intel wifi card. Apparently power management is completely turned off on all intel wifi cards post-2.6.30 kernels, with no clear intentions on bringing it back. There were a few posts I found of some folks compiling their own wifi drivers using compat-wireless and then manually stripping the patch that disable power management from the drivers, but I haven't been able to get the compat-wireless working on my kernel and the sources listing the patch change that I need a copy of to actually remove the patch from the driver has been down for months (and is likely not being revived anytime soon).

What are people doing for the Intel Power Management issue since these were the only cards available in almost all thinkpads. It seems everyone is just living with the higher usage?

ThinkRob · **Posted:** Mon Mar 19, 2012 9:24 pm

dmdsoftware wrote:

One of the areas I know I can shed at least a watt or two is on the intel wifi card. Apparently power management is completely turned off on all intel wifi cards post-2.6.30 kernels, with no clear intentions on bringing it back.

This is incorrect. It is disabled for the 4965 and 3945 (and possibly the 2x00, I haven't checked.) The 5100 and later have working power management.

Quote:

With moderate backlight, with all tunable parameters switched to good in PowerTop (at least the ones that will switch), and running on 800mhz (cpufreq powersave) with moderate screen brightness + all non-used perhapherals turned off, I get about 15W power usage on my X61T with L7300 processor. This seems quite high for a low voltage cpu.

It is quite high.

Here's what I'd recommend:

1) Stop messing with the cpu frequencies yourself. Set the governor to on-demand.

2) Use the latest stable mainline kernel.

3) If you have an Intel GPU, enable FBC and LVDS downclocking if your hardware can handle it.

4) Install laptop-mode tools and enable it.

5) Follow powertops recommendations with the exception of the writeback settings (LMT will actually set a higher interval than powertop recommends.)

6) Set the highest kernel ENERGY_PERF_BIAS setting that you're comfortable with if you're running on hardware that supports it.

7) Enable USB autosuspend.

8) Enable PCIe ASPM if your hardware can handle it.

I prefer a clean and stable distro, so I go with Debian instead of Mint, but the tips above should apply to both.

dmdsoftware · **Posted:** Tue Mar 20, 2012 8:06 am

ThinkRob wrote:

One of the areas I know I can shed at least a watt or two is on the intel wifi card. Apparently power management is completely turned off on all intel wifi cards post-2.6.30 kernels, with no clear intentions on bringing it back.

This is incorrect. It is disabled for the 4965 and 3945 (and possibly the 2x00, I haven't checked.) The 5100 and later have working power management.

Quote:

It is quite high.

Here's what I'd recommend:

1) Stop messing with the cpu frequencies yourself. Set the governor to on-demand.

2) Use the latest stable mainline kernel.

3) If you have an Intel GPU, enable FBC and LVDS downclocking if your hardware can handle it.

4) Install laptop-mode tools and enable it.

5) Follow powertops recommendations with the exception of the writeback settings (LMT will actually set a higher interval than powertop recommends.)

6) Set the highest kernel epb setting that you're comfortable with.

7) Enable USB autosuspend.

Enable PCIe ASPM if your hardware can handle it.

I prefer a clean and stable distro, so I go with Debian instead of Mint, but the tips above should apply to both.

Thanks. I'll try out these recommendations.

I'm actually using pm-powersave which uses a version of laptop-mode, but I recently made so additional tweaks to it since I noticed it wasn't running in the right mode (never going into laptop mode).

I'm using one of the latest stable 3.0.0 kernels. I have also one of the latest stable 3.2 kernels installed, however, not currently using it as I had problems with a lot of kernel modules (that would require some time to correct).

After I got pm-powersave using laptop mode correctly, 7 + 8 were being enabled by it successfully. I also went into the BIOS and disabled any hardware I wasn't using (I made the initial mistake of disabling Serial and realized shortly after that the tablet features use the Serial connection). Since fixing pm-powersave, I'm getting 13W.

I'm going to look at your points 3 + 6 next. I have recently come across an article (45 pages) discussing cpu frequencies and sleep states, etc, so I'm going to read through this to better understand how to proceed with this. I'm going to try using on-demand and monitor power consumption with it vs messing with the values myself.

ThinkRob · **Posted:** Tue Mar 20, 2012 10:43 am

dmdsoftware wrote:

I'm actually using pm-powersave which uses a version of laptop-mode, but I recently made so additional tweaks to it since I noticed it wasn't running in the right mode (never going into laptop mode).

I don't know much about pm-powersave, but I don't believe it's the same as Laptop Mode Tools. Are you sure you're not confusing LMT for the kernel's laptop-mode, which pm-utils does make use of? (Confusing naming, I know.)

Quote:

I'm using one of the latest stable 3.0.0 kernels. I have also one of the latest stable 3.2 kernels installed, however, not currently using it as I had problems with a lot of kernel modules (that would require some time to correct).

Well since you're using Ubuntu, make sure you don't build a 3.2 kernel the traditional way. Use make-kpkg instead. I'd also recommend against using any of the the various PPAs out there.

Quote:

After I got pm-powersave using laptop mode correctly, 7 + 8 were being enabled by it successfully. I also went into the BIOS and disabled any hardware I wasn't using (I made the initial mistake of disabling Serial and realized shortly after that the tablet features use the Serial connection). Since fixing pm-powersave, I'm getting 13W.

pm-powersave force-enables ASPM? I have a tough time believing that, actually, since AFAIK the only way to override the support detection is via a boot parameter. You might want to double-check that. (You can quickly get a sense of what the kernel's doing re: ASPM via 'dmesg | grep -i aspm')

Quote:

I'm going to look at your points 3 + 6 next. I have recently come across an article (45 pages) discussing cpu frequencies and sleep states, etc, so I'm going to read through this to better understand how to proceed with this. I'm going to try using on-demand and monitor power consumption with it vs messing with the values myself.

The theory behind my recommendation is simple, if somewhat counter-intuitive. While it's true that lower frequencies can result in lower power usage, it's often better to finish the work as fast as you can such that the CPU can return to a deep sleep state as quickly as possible.

Let's say I have two configuration A and B. At capacity, A can perform one unit of work per unit of power in a given period of time. B can perform three units of work per two units of power in a given period of time. Both idle at the same power level. A uses half the power of B, so at first glance it's the best choice, right? Yes, provided you're never doing anything that needs more than one unit of work in that time period. If you are, you now need to run A at full power for longer to complete your work; if you used B you could return to idle faster, despite using more power in the short run. Given the massive delta in power consumption between idle and load, this can be a big deal.

Let me know if you have any other questions. I'm also usually on Freenode with the same nick as here if you're a fan of IRC and want to debug your configuration in real time.

dmdsoftware · **Posted:** Tue Mar 20, 2012 12:06 pm

ThinkRob wrote:

I don't know much about pm-powersave, but I don't believe it's the same as Laptop Mode Tools. Are you sure you're not confusing LMT for the kernel's laptop-mode, which pm-utils does make use of? (Confusing naming, I know.)

I could be confusing the name. I believe pm-powersave is a subset of pm-utils. I typically install/use laptop-mode (installed via apt-get). On this Mint install, I followed suit, but realized shortly later that it wasn't working. Dug deeper and discovered that pm-powersave was installed and it contains a subset of laptop-mode and they cannot co-exist together or neither will work. I can invoke laptop mode by either running pm-powersave true or setting a non-zero integer (such as 5) in the laptop_mode file under /sys/...

Quote:

I'm using one of the latest stable 3.0.0 kernels. I have also one of the latest stable 3.2 kernels installed, however, not currently using it as I had problems with a lot of kernel modules (that would require some time to correct).

Quote:

Well since you're using Ubuntu, make sure you don't build a 3.2 kernel the traditional way. Use make-kpkg instead. I'd also recommend against using any of the the various PPAs out there.

I'm actually using Mint LMDE (Linux Mint Debian Edition). So it is based off Debian instead of Ubuntu. I tried both out and went with the LMDE as I've tired of Ubuntu over the past decade. In the 90s, I was a Debian-guy but switched to Ubuntu in 2003/4.

Quote:

pm-powersave force-enables ASPM? I have a tough time believing that, actually, since AFAIK the only way to override the support detection is via a boot parameter. You might want to double-check that. (You can quickly get a sense of what the kernel's doing re: ASPM via 'dmesg | grep -i aspm')

It's a parameter in the configuration. The default is false but I have modified the configuration to true to enable.

Quote:

The theory behind my recommendation is simple, if somewhat counter-intuitive. While it's true that lower frequencies can result in lower power usage, it's often better to finish the work as fast as you can such that the CPU can return to a deep sleep state as quickly as possible.

Let's say I have two configuration A and B. At capacity, A can perform one unit of work per unit of power in a given period of time. B can perform three units of work per two units of power in a given period of time. Both idle at the same power level. A uses half the power of B, so at first glance it's the best choice, right? Yes, provided you're never doing anything that needs more than one unit of work in that time period. If you are, you now need to run A at full power for longer to complete your work; if you used B you could return to idle faster, despite using more power in the short run. Given the massive delta in power consumption between idle and load, this can be a big deal.

A few friends have told me the same, so was always on the fence on the subject. I started reading through a bit more on the governance models, and they do recommend on-demand or conservative over powersave as the typical net result is better power savings.

I'm going to finish doing some more research and then try monitoring consumption under various situations.

A question about FBC and LVDS.... do these need to be enabled, and how? I was starting to dig up some information on these and found information suggesting they are enabled by default in the newer kernels.

ThinkRob · **Posted:** Tue Mar 20, 2012 2:14 pm

dmdsoftware wrote:

I could be confusing the name. I believe pm-powersave is a subset of pm-utils. I typically install/use laptop-mode (installed via apt-get). On this Mint install, I followed suit, but realized shortly later that it wasn't working. Dug deeper and discovered that pm-powersave was installed and it contains a subset of laptop-mode and they cannot co-exist together or neither will work. I can invoke laptop mode by either running pm-powersave true or setting a non-zero integer (such as 5) in the laptop_mode file under /sys/...

It still seems like you're confusing the kernel usage of the term "laptop mode" with the software package Laptop Mode Tools.

To the best of my knowledge, pm-utils does not "contain a subset of laptop mode", but rather uses the *kernel*'s laptop mode setting (in addition to other things) in an attempt to improve power usage.

For more information, you might want to check out LMT's home page.

Personally I use LMT. I have pm-utils installed, but I don't have it set up to do anything. LMT handles all of my power saving config just fine. I can help you configure your box in a similar manner and it will work, but if you don't want to use LMT I'm afraid that I'll be somewhat more limited in terms of the advice that I can offer.

Quote:

It's a parameter in the configuration. The default is false but I have modified the configuration to true to enable.

Are you referring to /sys/module/pcie_aspm/parameters/policy ? Because that only takes effect if ASPM is enabled -- and on most ThinkPads it's disabled by default due to BIOS quirks (hence the need for the command line options.) Again, I recommend you check what the ASPM support check shows in your kernel log and go from there.

Quote:

A few friends have told me the same, so was always on the fence on the subject. I started reading through a bit more on the governance models, and they do recommend on-demand or conservative over powersave as the typical net result is better power savings.

I'm going to finish doing some more research and then try monitoring consumption under various situations.

When last I looked at the implementation, the only difference between the conservative and on-demand governors was how aggressive they were about up-clocking under load. Conservative used a much more, well, conservative scaling approach whereas on-demand almost immediately ramps to the top frequency. In general, this means that on-demand gets the work done faster which (on modern chips) means a faster return to sleep states. This is what you want.

Quote:

A question about FBC and LVDS.... do these need to be enabled, and how? I was starting to dig up some information on these and found information suggesting they are enabled by default in the newer kernels.

Oh god, the i915 options... well it depends on the kernel version. It seems like the defaults for that are *always* changing thanks to the insane number of hardware bugs that the DRM devs discover.

As of 3.2.6 (what my current local kernel is based on):

Code:

root@kestrel:/etc/pm# modinfo i915
filename:       /lib/modules/3.2.6-grsec/kernel/drivers/gpu/drm/i915/i915.ko
license:        GPL and additional rights
description:    Intel Graphics
author:         Tungsten Graphics, Inc.
license:        GPL and additional rights
depends:        drm_kms_helper,video,button
intree:         Y
vermagic:       3.2.6-grsec SMP preempt mod_unload modversions KERNEXEC_BTS UDEREF REFCOUNT GRSEC 
parm:           modeset:Use kernel modesetting [KMS] (0=DRM_I915_KMS from .config, 1=on, -1=force vga console preference [default]) (int)
parm:           fbpercrtc:int
parm:           panel_ignore_lid:Override lid status (0=autodetect [default], 1=lid open, -1=lid closed) (int)
parm:           powersave:Enable powersavings, fbc, downclocking, etc. (default: true) (int)
parm:           semaphores:Use semaphores for inter-ring sync (default: -1 (use per-chip defaults)) (int)
parm:           i915_enable_rc6:Enable power-saving render C-state 6 (default: -1 (use per-chip default) (int)
parm:           i915_enable_fbc:Enable frame buffer compression for power savings (default: -1 (use per-chip default)) (int)
parm:           lvds_downclock:Use panel (LVDS/eDP) downclocking for power savings (default: false) (int)
parm:           lvds_use_ssc:Use Spread Spectrum Clock with panels [LVDS/eDP] (default: auto from VBT) (int)
parm:           vbt_sdvo_panel_type:Override selection of SDVO panel mode in the VBT (default: auto) (int)
parm:           reset:Attempt GPU resets (default: true) (bool)
parm:           enable_hangcheck:Periodically check GPU activity for detecting hangs. WARNING: Disabling this can cause system wide hangs. (default: true) (bool)

I'd recommend you check your kernel though, since your defaults may very well be different.

dmdsoftware · **Posted:** Tue Mar 20, 2012 4:25 pm

ThinkRob wrote:

I could be confusing the name. I believe pm-powersave is a subset of pm-utils. I typically install/use laptop-mode (installed via apt-get). On this Mint install, I followed suit, but realized shortly later that it wasn't working. Dug deeper and discovered that pm-powersave was installed and it contains a subset of laptop-mode and they cannot co-exist together or neither will work. I can invoke laptop mode by either running pm-powersave true or setting a non-zero integer (such as 5) in the laptop_mode file under /sys/...

This is what I'm referring to as pm-powersave
http://ubuntuforums.org/showthread.php?t=1450501

Like was pointed out in the thread, when I had pm-util installed and the took laptop-mode tools both installed, then neither worked. pm-utiil with powersave scripts were installed by default.

I could always try to disable pm-powersave by wiping out the laptop_mode script and then trying to use install and use laptop mode tools.

ThinkRob · **Posted:** Tue Mar 20, 2012 5:40 pm

dmdsoftware wrote:

This is what I'm referring to as pm-powersave
http://ubuntuforums.org/showthread.php?t=1450501

Like was pointed out in the thread, when I had pm-util installed and the took laptop-mode tools both installed, then neither worked. pm-utiil with powersave scripts were installed by default.

I could always try to disable pm-powersave by wiping out the laptop_mode script and then trying to use install and use laptop mode tools.

That's something different though!

pm-utils-powersave-policy is an Ubuntu package which conflicts with the laptop-mode-tools package.

At least under Debian there is no conflict between the pm-utils package (of which the pm-powersave command is a part) and laptop-mode-tools package.

dmdsoftware · **Posted:** Tue Mar 20, 2012 8:09 pm

ThinkRob wrote:

This is what I'm referring to as pm-powersave
http://ubuntuforums.org/showthread.php?t=1450501

Like was pointed out in the thread, when I had pm-util installed and the took laptop-mode tools both installed, then neither worked. pm-utiil with powersave scripts were installed by default.

I could always try to disable pm-powersave by wiping out the laptop_mode script and then trying to use install and use laptop mode tools.

That's something different though!

pm-utils-powersave-policy is an Ubuntu package which conflicts with the laptop-mode-tools package.

At least under Debian there is no conflict between the pm-utils package (of which the pm-powersave command is a part) and laptop-mode-tools package.

Thanks for the info. I reinstalled laptop-mode-tools
Laptop mode
enabled,
active [unchanged]

Drive power status:

/dev/sda:
drive state is: active/idle

(NOTE: drive settings affected by Laptop Mode cannot be retrieved.)

Readahead states:
/dev/sda1: 3072 kB
/dev/sda3: 3072 kB

Laptop Mode Tools is allowed to run: /var/run/laptop-mode-tools/enabled exists.

/proc/sys/vm/laptop_mode:
2

/proc/sys/vm/dirty_ratio:
60

/proc/sys/vm/dirty_background_ratio:
1

/proc/sys/vm/dirty_expire_centisecs:
60000

/proc/sys/vm/dirty_writeback_centisecs:
60000

/sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq:
800000

/sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_max_freq:
1601000

/sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_min_freq:
800000

/sys/devices/system/cpu/cpu1/cpufreq/cpuinfo_cur_freq:
800000

/sys/devices/system/cpu/cpu1/cpufreq/cpuinfo_max_freq:
1601000

/sys/devices/system/cpu/cpu1/cpufreq/cpuinfo_min_freq:
800000

Anyways, it's installed now.

dmdsoftware · **Posted:** Tue Mar 20, 2012 8:19 pm

ThinkRob wrote:

Are you referring to /sys/module/pcie_aspm/parameters/policy ? Because that only takes effect if ASPM is enabled -- and on most ThinkPads it's disabled by default due to BIOS quirks (hence the need for the command line options.) Again, I recommend you check what the ASPM support check shows in your kernel log and go from there.

You're right, it is getting disabled.

Mar 17 22:30:53 x61t-2 kernel: [ 0.170744] ACPI FADT declares the system doesn't support PCIe ASPM, so disable it

ThinkRob · **Posted:** Wed Mar 21, 2012 10:28 am

dmdsoftware wrote:

You're right, it is getting disabled.

Mar 17 22:30:53 x61t-2 kernel: [ 0.170744] ACPI FADT declares the system doesn't support PCIe ASPM, so disable it

Yep. That's why I recommended you check.

The reworked detection code hasn't been backported yet IIRC, so you either need to force it or use the latest mainline. If you want to forcibly enable it (with the knowledge that the hardware might be quirked, etc.) then you can do so via adding the 'pcie_aspm=force' parameter to your kernel boot options.

ThinkRob · **Posted:** Wed Mar 21, 2012 10:36 am

The sysctl posting above looks reasonable. I'd also check that all of the appropriate tweaks within /etc/laptop-mode/conf.d/ are configured correctly (particularly WOL, AHCI power saving, and any WLAN settings, but really any of them that you think you can do.) IIRC you need to enable some of those manually as they require a bit of config (usually device names, etc.)

dmdsoftware · **Posted:** Thu Mar 22, 2012 9:19 am

Thanks for all the tips.

I've decided I will proceed with upgrading to the 3.3 mainline kernel. I heed your advise not to use distribution provided kernel. I'll be using the method you suggested.

I also looked into the downlocking if the i915. How to set the different settings is pretty clear, and I will proceed with trying this out when I get the kernel installed.

Regarding ASPM, I went through the steps to deem if cards I use would benefit and indeed the wifi card is L0 capable, so I will look at this. I'm also still looking at the possibility of switching out the wireless card. I think the T410 that I have sitting in a drawer at work has a Centrino Advanced N wireless card in it, which is probably likely to be better than the existing card. I have Middleton's bios installed, so I'm not expecting whitelist issues with that respect.

ThinkRob · **Posted:** Thu Mar 22, 2012 11:44 am

dmdsoftware wrote:

Regarding ASPM, I went through the steps to deem if cards I use would benefit and indeed the wifi card is L0 capable, so I will look at this. I'm also still looking at the possibility of switching out the wireless card. I think the T410 that I have sitting in a drawer at work has a Centrino Advanced N wireless card in it, which is probably likely to be better than the existing card. I have Middleton's bios installed, so I'm not expecting whitelist issues with that respect.

That probably will drop the power consumption. I'm not sure what card you use now, but the latest Intel WLAN cards definitely have better driver support for power savings. You can also drop the Tx power too using iwconfig, which should help if you're near enough to your AP that you don't need full Tx power.

Quote:

I've decided I will proceed with upgrading to the 3.3 mainline kernel. I heed your advise not to use distribution provided kernel. I'll be using the method you suggested.

I haven't tried 3.3 yet, but I do build mainline kernels quite a bit, and once you get used to it make-kpkg is a *great* tool. I would highly recommend that you RTFM first, otherwise you'll wind up confused like me on my first time with a non-booting kernel.

Let me know if you have any further questions.

dmdsoftware · **Posted:** Thu Mar 22, 2012 1:15 pm

ThinkRob wrote:

Regarding ASPM, I went through the steps to deem if cards I use would benefit and indeed the wifi card is L0 capable, so I will look at this. I'm also still looking at the possibility of switching out the wireless card. I think the T410 that I have sitting in a drawer at work has a Centrino Advanced N wireless card in it, which is probably likely to be better than the existing card. I have Middleton's bios installed, so I'm not expecting whitelist issues with that respect.

That probably will drop the power consumption. I'm not sure what card you use now, but the latest Intel WLAN cards definitely have better driver support for power savings. You can also drop the Tx power too using iwconfig, which should help if you're near enough to your AP that you don't need full Tx power.

I've decided I will proceed with upgrading to the 3.3 mainline kernel. I heed your advise not to use distribution provided kernel. I'll be using the method you suggested.

I haven't tried 3.3 yet, but I do build mainline kernels quite a bit, and once you get used to it make-kpkg is a *great* tool. I would highly recommend that you RTFM first, otherwise you'll wind up confused like me on my first time with a non-booting kernel.

Let me know if you have any further questions.

Right now, the only power management I'm doing is dropping down the txpower to 7 when I'm accessing the AP from within the house. I also like to access the AP from out on the patio, and I've found 7 isn't adequate enough, so I usually tweak it to 10-12 which I find is stable enough.

I thought maybe you were already on 3.3. From what I was reading, ASPM is working in 3.3 (at least to get rid of the warning). I'm still expecting to have to set the bits to activate it on each card, but I figured it'd probably be best to also be on 3.3.

I'm thoroughly reading through the documentation on the tool, so hopefully the first attempt will produce something that boots.

ThinkRob · **Posted:** Fri Mar 23, 2012 12:59 pm

dmdsoftware wrote:

Right now, the only power management I'm doing is dropping down the txpower to 7 when I'm accessing the AP from within the house. I also like to access the AP from out on the patio, and I've found 7 isn't adequate enough, so I usually tweak it to 10-12 which I find is stable enough.

Yeah Tx power is different than power savings (go figure). Even the 3945 and 4965 support setting the Tx power manually, but their normal power saving support is broken.

Quote:

I thought maybe you were already on 3.3. From what I was reading, ASPM is working in 3.3 (at least to get rid of the warning). I'm still expecting to have to set the bits to activate it on each card, but I figured it'd probably be best to also be on 3.3.

Oh, ASPM works fine in all of 3.x. It's the detection of BIOS support that's the tricky bit. 3.3 basically changed the behavior to "act like Windows", so there are a bunch of boxes that will now be detected as supporting it that previously wouldn't. I have no idea whether most ThinkPads will be detected as supporting it, but I do know that it's worked on all of the ones that I've forced it on.

Quote:

I'm thoroughly reading through the documentation on the tool, so hopefully the first attempt will produce something that boots.

I didn't mean to imply it's impossible or anything -- it's just that people often forget to add the intrd support, or forget to either produce an unsigned package (or add their signing keys), or... etc. Not that I'm guilty of doing both of those and more...

dmdsoftware · **Posted:** Wed Mar 28, 2012 11:37 am

I was waiting to get some time, which was today, so today I got the latest 3.3 kernel installed. I enabled ASPM with the force command, but I'm still seeing the warning, so I assume ASPM is really getting forced enabled or not?

[ 0.000000] PCIe ASPM is forcibly enabled
[ 0.100473] ACPI FADT declares the system doesn't support PCIe ASPM, so disable it

I still have to do the setpci part for the wifi card, which I will work on next.

ThinkRob · **Posted:** Wed Mar 28, 2012 10:35 pm

dmdsoftware wrote:

I was waiting to get some time, which was today, so today I got the latest 3.3 kernel installed. I enabled ASPM with the force command, but I'm still seeing the warning, so I assume ASPM is really getting forced enabled or not?

[ 0.000000] PCIe ASPM is forcibly enabled
[ 0.100473] ACPI FADT declares the system doesn't support PCIe ASPM, so disable it

IIRC the second warning isn't correct if you've forced it on.

Quote:

I still have to do the setpci part for the wifi card, which I will work on next.

?

dmdsoftware · **Posted:** Thu Mar 29, 2012 7:29 am

I'll have to have another look at why the force isn't working on 3.3.

With regards to setpci, I still thought the following was necessary even when forcing on:

http://linuxwireless.org/en/users/Documentation/ASPM

I tried the above and was able to get ASPM active for the wifi.

I'll have another look at the kernel. Maybe I'll try the latest 3.2 instead.

BTW, the tweaks (downclocking,etc) on the intel GPU worked out fine on 3.3. I got idle power consumption down to 10.6W

What are you using to measure the wattage?

ThinkRob · **Posted:** Thu Mar 29, 2012 10:14 am

dmdsoftware wrote:

I'll have to have another look at why the force isn't working on 3.3.

With regards to setpci, I still thought the following was necessary even when forcing on:

http://linuxwireless.org/en/users/Documentation/ASPM

I tried the above and was able to get ASPM active for the wifi.

Ah, gotcha. Yeah, individual links may need their state to be set -- the boot parameter just allows ASPM support to be enabled regardless of what the BIOS claims.

Quote:

BTW, the tweaks (downclocking,etc) on the intel GPU worked out fine on 3.3. I got idle power consumption down to 10.6W

What are you using to measure the wattage?

powertop

10.6 is good, and might actually be close to what you can get with that hardware. When I had an X61s I was hitting 8 not infrequently, and that was using a much older 2.6.x kernel (albeit with some pretty heavy patching) and a *way* older userland.

What does the X61t usually get under Windows?

dmdsoftware · **Posted:** Thu Mar 29, 2012 1:00 pm

Nm, I realize now that if you use force, you don't need to setpci.

It seems rather my kernel isn't working.

I went to recompile it and I noticed maybe I missed this after I installed it:

Selecting previously deselected package linux-image-3.3.0.20120329.
(Reading database ... 239775 files and directories currently installed.)
Unpacking linux-image-3.3.0.20120329 (from linux-image-3.3.0.20120329_3.3.0.20120329-10.00.Custom_amd64.deb) ...
Done.
Setting up linux-image-3.3.0.20120329 (3.3.0.20120329-10.00.Custom) ...
Running depmod.
Examining /etc/kernel/postinst.d.
run-parts: executing /etc/kernel/postinst.d/dkms 3.3.0.20120329 /boot/vmlinuz-3.3.0.20120329
dkms.conf: Error! No 'BUILT_MODULE_NAME' directive specified for record #0.
Error! Bad conf file.
File:
does not represent a valid dkms.conf file.
run-parts: executing /etc/kernel/postinst.d/initramfs-tools 3.3.0.20120329 /boot/vmlinuz-3.3.0.20120329
run-parts: executing /etc/kernel/postinst.d/pm-utils 3.3.0.20120329 /boot/vmlinuz-3.3.0.20120329
run-parts: executing /etc/kernel/postinst.d/zz-update-grub 3.3.0.20120329 /boot/vmlinuz-3.3.0.20120329
Generating grub.cfg ...
Found background image: linuxmint.png
Found Debian background: linuxmint.png
Found linux image: /boot/vmlinuz-3.3.0.20120329
Found linux image: /boot/vmlinuz-3.3.0-rc6-amd64
Found initrd image: /boot/initrd.img-3.3.0-rc6-amd64
Found linux image: /boot/vmlinuz-3.2.0-1-amd64
Found initrd image: /boot/initrd.img-3.2.0-1-amd64
Found linux image: /boot/vmlinuz-3.0.0-1-amd64
Found initrd image: /boot/initrd.img-3.0.0-1-amd64
done

Looks like make-kpkg isn't generating a clean binary.

dmdsoftware · **Posted:** Thu Mar 29, 2012 5:06 pm

Ah, I've never seen the watts displayed in powertop before, in the years I've used it. I just looked at the one I have installed on this machine, 1.97-2, and don't see it. I've been checking the consumption by checking the ACPI power_now battery value and by using a kill-a-watt.

I have never actually installed Windows 7 on this machine to-date. I have benchmark values I recorded with the kill-a-watt for Windows 7 on a X61 and the SXGA+ X61T, but nothing for this particular machine. I'll look at comparing the two once I get all the linux tweaks in.

dmdsoftware · **Posted:** Thu Mar 29, 2012 9:35 pm

I got the error with the kernel make resolved. It was VirtualBox kernel modules causing an issue. Recompiled them and made the kernel again.

However, with the latest incarnation still getting
ACPI FADT declares the system doesn't support PCIe ASPM, so disable it

Even when I do a pcie_aspm=force, I no longer get
[ 0.000000] PCIe ASPM is forcibly enabled

which is what I was getting previously with 3.3.

Is there a particular kernel configuration parameter that needs to be toggled?

dmdsoftware · **Posted:** Sun Apr 01, 2012 12:16 pm

Some great progress in the last few days.

I got the X61T (with touch) -- the same model I've been disusing in this whole thread -- down to 8.7W (!) [screen brightness on lowest setting. I achieved 7.4W when I turned wireless off. Both scenarios, this was with Touch screen and digitizer turned on. So if I was in a need for a desperate battery savings situation, I could toggle these off and expect to save some more.

That is an amazing feat!

What I ended up doing was loading the liquorix kernel. I was encountering the ASPM warning even when forcing it on when I compiled my own 3.3 mainline kernel, so I assumed I must have been doing something wrong. I read that all the liquorix kernels have the proper ASPM setup in them, so that attracted me to it. Also, in a previous install when I was trying Linux Mint 12 (the ubuntu flavour -- before moving to LMDE), I had liked the idea of using a PHC kernel so that I could undervolt. I had started using undervolting on some of my Windows Thinkpads (in particular a X60) due to overheating issues. So I wanted to do the same in Linux. I could never get the PHC kernel working -- the PHC entries didn't show up on my CPU devices. Anyways, with liquorix, I read that it also included PHC, so it was a second reason why I wanted to try it.

So in using the new kernel (latest liquorix 3.2) with forcing ASPM on (by the way, I still get the ASPM warning) + i915 tweaks + undervolting with known stable VID values for my CPU, I was now achieving these new numbers with a quiet, stable and cool running Thinkpad.

I'll always continue tweaking, but I'm very happy with these results, so I think I can bring closure to this venture. There is still the issue of the wifi card that I'm going to replace with a newer (and one that supports proper PM), which I will try next week once I get access to my T410 to grab the card out of.

One of the things I did on Saturday was also move my HP machine to LMDE. There was a bundle hardware issues that I'm working my way through (the wifi wasn't supported, but got a 3rd party kernel module working, and intermittent issue with the ethernet). Once I stabilize that machine, I'll look at tweaking it's power management. It was a Window 7 machine that hosted VMs that I would connect to. It was always my plan to migrate the machine to Linux when I got the time. One thing I noticed is that I get a very different ASPM message in the dmeg when I used my compiled mainline 3.3 kernel (something about pre 1.1 PCI), which I'll spend some time researching about. I also noticed there is no proper working ACPI on the machine at the moment (so no way of monitoring power consumption through software), so when I do begin tweaking, I'll have to rely on the kill-a-watt reader.

I want to thank you/ ThinkRob for all your help, and I'm hoping this thread will assist other Thinkpad + Linux users (especially those with X60/X60s/X60T/X61/X61s/X61T/T60/T61 hardware) tweak there gear. I would have never got to where I am today without you sharing your experience. Your an asset to the Thinkpad and Linux communities! And I'm sure I'll be running into you on this and other forums in the near future.

ThinkRob · **Posted:** Sun Apr 01, 2012 6:29 pm

Great news! I'm glad you're happy with the results.

I generally have had rather uneven experiences with PHC, so I never really sought out patchsets that offered it.

My bet is that the difference in some of the behavior that you're describing is due to the differences in 3.2 vs. 3.3 rather than due to the liquorix patchset per se, but I don't really know. I generally only use patches that I'm intimately familiar with, have personally backported, or have very, very good reason to trust (e.g. grsecurity) -- and while I know that liquorix kernels are popular they don't really meet those criteria for me.

BTW, I neglected to confirm and post this, but up until the full re-work of that code, the "ACPI FADT declares the system doesn't support PCIe ASPM, so disable it" warning isn't actually true if you forcibly enabled ASPM. It always prints that -- but it only does the subsequent state clear if you didn't forcibly enable. Screwy, but there you have it.

dmdsoftware · **Posted:** Wed May 09, 2012 9:22 pm

Just a little update.

I needed a halfsize minipci bracket to try the replacement WLAN card from the T420. The bracket arrived this week. I then had issues trying to get the tight screws loose. I eventually stripped the screws. Instead of trying to make things worse by possibly damaging the unit by trying to drill the screws out, I decided to make use of the second minipci slot. I figured it'll give me better options anyways -- this way I can keep the original card and try out the new card.

I soon discovered that I needed to put a piece of tap on pin 20 on the wifi card. I was under the impression this was only needed if your machine was still whitelisting accepted WLAN cards. However, the pin causes the system to think the hardware wifi switch is turned off, which is exactly what happened. I was able to see both cards in linux, but both were toggled off due to hardware switch being detected off.

I borrowed one of the primary WLAN antenna connections and decided to snake in a 4th antenna.

Now both connections show up in linux. Further, if I want to disable the primary WLAN (the Intel 4965 -- with no modern power management support in linux) all I have to do is toggle the Wireless LAN toggle to disable in the BIOS under security. Likewise, I can control the secondary WLAN card by toggling the Wireless WLAN to disable under the same screen.

The end result is that I now have a better wireless card, one that produces less heat, consumes less power, and has proper power management in linux. I also appear to get better range. I've had trouble getting a connection with this particular laptop on the patio with the 4965, but I have no problem now maintaining a working reliable connection.

DRobinson · **Posted:** Thu May 10, 2012 12:18 am

About 11watts idle, 35watts full cpu load on a T20 w/900MHz P3.

Code:

acpitool -B
Battery #1     : present
    Remaining capacity : 14394 mWh, 99.82%, 01:19:24
    Design capacity    : 38880 mWh
    Last full capacity : 14420 mWh, 37.09% of design capacity
    Capacity loss      : 62.91%
    Present rate       : 10875 mW
    Charging state     : Discharging
    Battery type       : Li-ion 
    Model number       : ThinkPad

Code:

lshw
ibm-t20-01                
    description: Notebook
    vendor: IBM
    version: Not Available
    serial: Not Available
    width: 32 bits
    capabilities: smbios-2.3 dmi-2.3
    configuration: boot=normal chassis=notebook
  *-core
       description: Motherboard
       vendor: IBM
       physical id: 0
       version: Not Available
       serial: J1JUG18D2AY
     *-firmware
          description: BIOS
          vendor: IBM
          physical id: 0
          version: 16ET31WW (1.11 )
          date: 03/20/2003
          size: 144KiB
          capacity: 448KiB
          capabilities: isa pci pcmcia pnp apm upgrade shadowing escd cdboot bootselect pcmciaboot edd int13floppy720 int13floppy2880 int5printscreen int9keyboard int14serial int17printer int10video acpi agp ls120boot biosbootspecification
     *-cpu
          description: CPU
          product: Pentium III (Coppermine)
          vendor: Intel Corp.
          physical id: 6
          bus info: cpu@0
          version: 6.8.10
          slot: None
          size: 900MHz
          capacity: 900MHz
          width: 32 bits
          clock: 100MHz
          capabilities: fpu fpu_exception wp vme de pse tsc msr pae mce cx8 sep mtrr pge mca cmov pse36 mmx fxsr sse up
        *-cache:0
             description: L1 cache
             physical id: a
             slot: Internal L1 Cache
             size: 32KiB
             capacity: 32KiB
             capabilities: synchronous internal write-back
        *-cache:1
             description: L2 cache
             physical id: b
             slot: Internal L2 Cache
             size: 256KiB
             capacity: 256KiB
             capabilities: internal write-back unified
     *-memory
          description: System Memory
          physical id: 24
          slot: System board or motherboard
          size: 512MiB
          capacity: 1GiB
        *-bank:0
             description: SODIMM SDRAM Synchronous
             physical id: 0
             slot: DIMM 1
             size: 256MiB
             width: 64 bits
        *-bank:1
             description: SODIMM SDRAM Synchronous
             physical id: 1
             slot: DIMM 2
             size: 256MiB
             width: 64 bits
     *-pci
          description: Host bridge
          product: 440BX/ZX/DX - 82443BX/ZX/DX Host bridge
          vendor: Intel Corporation
          physical id: 100
          bus info: pci@0000:00:00.0
          version: 03
          width: 32 bits
          clock: 33MHz
          configuration: driver=agpgart-intel latency=64
          resources: irq:0 memory:f8000000-fbffffff
        *-pci
             description: PCI bridge
             product: 440BX/ZX/DX - 82443BX/ZX/DX AGP bridge
             vendor: Intel Corporation
             physical id: 1
             bus info: pci@0000:00:01.0
             version: 03
             width: 32 bits
             clock: 66MHz
             capabilities: pci normal_decode bus_master
             resources: memory:f0000000-f7ffffff memory:20000000-200fffff
           *-display UNCLAIMED
                description: VGA compatible controller
                product: 86C270-294 Savage/IX-MV
                vendor: S3 Inc.
                physical id: 0
                bus info: pci@0000:01:00.0
                version: 13
                width: 32 bits
                clock: 66MHz
                capabilities: pm agp agp-1.0 vga_controller bus_master cap_list
                configuration: latency=64 maxlatency=255 mingnt=4
                resources: memory:f0000000-f7ffffff memory:20000000-2000ffff
        *-pcmcia:0
             description: CardBus bridge
             product: PCI1450
             vendor: Texas Instruments
             physical id: 2
             bus info: pci@0000:00:02.0
             version: 03
             width: 32 bits
             clock: 33MHz
             capabilities: pcmcia bus_master cap_list
             configuration: driver=yenta_cardbus latency=176 maxlatency=5 mingnt=192
             resources: irq:11 memory:50000000-50000fff ioport:2400(size=256) ioport:2000(size=256) memory:30000000-33ffffff memory:2c000000-2fffffff
        *-pcmcia:1
             description: CardBus bridge
             product: PCI1450
             vendor: Texas Instruments
             physical id: 2.1
             bus info: pci@0000:00:02.1
             version: 03
             width: 32 bits
             clock: 33MHz
             capabilities: pcmcia bus_master cap_list
             configuration: driver=yenta_cardbus latency=176 maxlatency=5 mingnt=192
             resources: irq:11 memory:50100000-50100fff ioport:1c00(size=256) ioport:1400(size=256) memory:28000000-2bffffff memory:24000000-27ffffff
        *-network
             description: Wireless interface
             product: AR5413/AR5414 Wireless Network Adapter [AR5006X(S) 802.11abg]
             vendor: Atheros Communications Inc.
             physical id: 3
             bus info: pci@0000:00:03.0
             logical name: wlan1
             version: 01
             serial: 00:11:f5:e4:0c:06
             width: 32 bits
             clock: 33MHz
             capabilities: pm bus_master cap_list ethernet physical wireless
             configuration: broadcast=yes driver=ath5k driverversion=3.2.0-2-686-pae firmware=N/A ip=10.0.2.25 latency=168 link=yes maxlatency=28 mingnt=10 multicast=yes wireless=IEEE 802.11abg
             resources: irq:11 memory:e8000000-e800ffff
        *-multimedia
             description: Multimedia audio controller
             product: CS 4614/22/24/30 [CrystalClear SoundFusion Audio Accelerator]
             vendor: Cirrus Logic
             physical id: 5
             bus info: pci@0000:00:05.0
             version: 01
             width: 32 bits
             clock: 33MHz
             capabilities: pm bus_master cap_list
             configuration: driver=snd_cs46xx latency=64 maxlatency=24 mingnt=4
             resources: irq:11 memory:e8010000-e8010fff memory:e8100000-e81fffff
        *-bridge:0 UNCLAIMED
             description: Bridge
             product: 82371AB/EB/MB PIIX4 ISA
             vendor: Intel Corporation
             physical id: 7
             bus info: pci@0000:00:07.0
             version: 02
             width: 32 bits
             clock: 33MHz
             capabilities: bridge bus_master
             configuration: latency=0
        *-ide
             description: IDE interface
             product: 82371AB/EB/MB PIIX4 IDE
             vendor: Intel Corporation
             physical id: 7.1
             bus info: pci@0000:00:07.1
             logical name: scsi0
             version: 01
             width: 32 bits
             clock: 33MHz
             capabilities: ide bus_master emulated
             configuration: driver=ata_piix latency=64
             resources: irq:0 ioport:1f0(size=8) ioport:3f6 ioport:170(size=8) ioport:376 ioport:1800(size=16)
           *-disk:0
                description: ATA Disk
                product: CF Card
                physical id: 0.0.0
                bus info: scsi@0:0.0.0
                logical name: /dev/sda
                version: Ver6
                serial: 14EF0717050800757884
                size: 7623MiB (7994MB)
                capabilities: partitioned partitioned:dos
                configuration: ansiversion=5 signature=000aad4c
              *-volume:0
                   description: EXT4 volume
                   vendor: Linux
                   physical id: 1
                   bus info: scsi@0:0.0.0,1
                   logical name: /dev/sda1
                   logical name: /
                   version: 1.0
                   serial: 19fd9a0b-3427-4a56-b94a-1da6666cc332
                   size: 7261MiB
                   capacity: 7261MiB
                   capabilities: primary bootable journaled extended_attributes large_files huge_files dir_nlink recover extents ext4 ext2 initialized
                   configuration: created=2012-03-20 21:30:46 filesystem=ext4 lastmountpoint=/ modified=2012-04-15 21:59:15 mount.fstype=ext4 mount.options=rw,relatime,errors=remount-ro,user_xattr,barrier=1,data=ordered mounted=2012-05-09 07:14:53 state=mounted
              *-volume:1
                   description: Extended partition
                   physical id: 2
                   bus info: scsi@0:0.0.0,2
                   logical name: /dev/sda2
                   size: 360MiB
                   capacity: 360MiB
                   capabilities: primary extended partitioned partitioned:extended
                 *-logicalvolume
                      description: Linux swap / Solaris partition
                      physical id: 5
                      logical name: /dev/sda5
                      capacity: 360MiB
                      capabilities: nofs
           *-disk:1
                description: ATA Disk
                product: ELITE PRO CF CAR
                physical id: 0.1.0
                bus info: scsi@0:0.1.0
                logical name: /dev/sdb
                version: Ver6
                serial: 5B22028B000999BB
                size: 7623MiB (7994MB)
                capabilities: partitioned partitioned:dos
                configuration: ansiversion=5
              *-volume:0
                   description: Linux swap volume
                   physical id: 1
                   bus info: scsi@0:0.1.0,1
                   logical name: /dev/sdb1
                   version: 1
                   serial: 2c31e1be-7308-4f3c-bfe9-61c1662beb5f
                   size: 972MiB
                   capacity: 972MiB
                   capabilities: primary nofs swap initialized
                   configuration: filesystem=swap pagesize=4096
              *-volume:1
                   description: EXT4 volume
                   vendor: Linux
                   physical id: 2
                   bus info: scsi@0:0.1.0,2
                   logical name: /dev/sdb2
                   logical name: /home
                   version: 1.0
                   serial: 865fc5e1-6a27-4384-9305-a3e2daaeb23e
                   size: 6651MiB
                   capacity: 6651MiB
                   capabilities: primary journaled extended_attributes large_files huge_files dir_nlink recover extents ext4 ext2 initialized
                   configuration: created=2012-04-01 15:03:32 filesystem=ext4 lastmountpoint=/home modified=2012-05-09 07:14:56 mount.fstype=ext4 mount.options=rw,relatime,errors=remount-ro,user_xattr,commit=600,barrier=1,data=ordered mounted=2012-05-09 07:14:56 state=mounted
        *-usb
             description: USB controller
             product: 82371AB/EB/MB PIIX4 USB
             vendor: Intel Corporation
             physical id: 7.2
             bus info: pci@0000:00:07.2
             version: 01
             width: 32 bits
             clock: 33MHz
             capabilities: uhci
             configuration: driver=uhci_hcd latency=64
             resources: irq:11 ioport:1820(size=32)
        *-bridge:1 UNCLAIMED
             description: Bridge
             product: 82371AB/EB/MB PIIX4 ACPI
             vendor: Intel Corporation
             physical id: 7.3
             bus info: pci@0000:00:07.3
             version: 03
             width: 32 bits
             clock: 33MHz
             capabilities: bridge
             configuration: latency=0

Code:

dmesg
[    0.000000] Linux version 3.2.0-2-686-pae (Debian 3.2.15-1) (debian-kernel@lists.debian.org) (gcc version 4.6.3 (Debian 4.6.3-3) ) #1 SMP Sun Apr 15 17:56:31 UTC 2012

Soul_Est · **Posted:** Fri Jun 01, 2012 10:06 am

I'm currently running at 6.30W - 6.45W at idle and ~7W to ~9W when I'm actively working depending on what radios are on and active. This is on a X200s mind you so the CPU is a low-voltage variant (which still has some serious power).

I'm currently running the 3.3.7 kernel under Arch Linux and i3 as my window manager.

Best Power Output Results (include which Kernel)

Who is online