[ltp] Power drain workaround with radeonfb module?

Dan Sawyer linux-thinkpad@linux-thinkpad.org
Sun, 28 Oct 2007 08:45:33 -0700


I have been on a similar trail. This seems to be dependent on agp and 
drm configurations. In 2.6.20 kernels without drm comfigured radeonfb 
loaded as a module. I then moved up levels and configured drm - the 
result was radeonfb seemed to be preempted by another driver and would 
not load. I then moved up to 2.6.23.1 and configured radeonfb into the 
kernel. With this level and configuration suspend works with about 500 
mw (however glxgears is only running at about 1/3 speed) The difference 
shows up between 3d rendering (glxgears) and suspend power.  - Dan


Felix E. Klee wrote:
> On Thinkwiki it's stated [1] that doing "modprobe radeonfb" before
> starting X can solve the high power drain issue on certain Thinkpads.
>
> Can that work? I previously thought that radeonfb has to be built into
> the kernel, and that's what a recent test confirms with Ubuntu 7.10
> (Gutsy Gibbon) on a T41 (2373-3kg):
>
> 1. The system started up in text mode (no X).
>
> 2. I loaded the module:
>
>   # modprobe radeonfb
>   # dmesg
>     [...]
>     [  932.232000] radeonfb: Monitor 1 type LCD found
>     [  932.232000] radeonfb: Monitor 2 type CRT found
>     [  932.232000] radeonfb: panel ID string: 1024x768                
>     [  932.232000] radeonfb: detected LVDS panel size from BIOS:
>     1024x768
>     [  932.232000] radeondb: BIOS provided dividers will be used
>     [  932.352000] radeonfb: Dynamic Clock Power Management enabled
>     [  932.352000] radeonfb: IBM Thinkpad R50/R51/T40/T41 detected,
>     enabling workaround
>     [  932.352000] radeonfb (0000:01:00.0): ATI Radeon Lf 
>
> 3. I executed sleep.sh, from Thinkwiki [2]:
>
>   # ./sleep.sh; reboot
>   [...]
>   # tail /var/log/battery.log
>     [...]
>     Sun Oct 28 14:27:16 CET 2007
>     before: 12150 mWh
>     after: 11200 mWh
>     diff: -950 mWh
>     seconds: 3023 sec
>     result: -1131 mW
>     Your model seems to be affected.
>
> On the same hardware, with Slackware 10.2 and a 2.16.14.4 kernel with
> radeonfb built in, the power drain issue is solved.
>
> [1]
> http://www.thinkwiki.org/wiki/Problem_with_high_power_drain_in_ACPI_sleep
>
> [2] http://www.thinkwiki.org/wiki/ACPI_sleep_power_drain_test_script
>