[ltp] but i810 can't use DynamicClocks

[Václav Šte(pán]

Dan Jacobson napsal(a):
> If only I could get the graphics chip cooler, I wouldn't have to
> listen to the fan.  But i810 can't use DynamicClocks, now what?
> Besides, even without doing /etc/init.d/xdm start, the fan is on half
> the time. So what can be done on an 80x25 line console to lower the
> graphics chip temperature?  Any cases where X can use less power than
> non X during normal use?
>   
I have never noticed two fans running - are you sure it's a GPU fan and 
not the main one?
Anyway - in my case the fan is off most of the time if I let the air 
vents on the bottom and left
free and stick to the 600 MHz CPU frequency. So maybe something is wrong 
with your computer?
If it's the main fan - isn't there some CPU intensive process running in 
the background?
> http://www.mk-stuff.de/artikel-linux_on_an_ibm_thinkpad_r50e.php#freq
> finds "no savings while decreasing the CPU frequency. Reason enough for
> me to deactivate all CPU-Frequencing options in the Kernel". Therefore
> no need to bother with the CPU.
That has a different meaning in the original context - he finds out that 
decreasing frequency:
 - leads to about 1/4 decrease in power requirements while increasing 
time up to 8x
 - has no effect when there is no load

   To the first point - sure, when you lower the CPU frequency, you get 
lower power
consumption and lower performance.
   To the second - that should be true, because when there is no load 
and you run some
intelligent OS, the CPU is halted when idle. However, you may disable 
this for both
PCI and CPU using the ps2.exe configuration utility or using the 
settings in Windows.
   The values below show a small difference, but that may be because 
they were
obtained in Windows and there might have been something causing CPU load 
in background
(well - like Windows I mean).

Alternative values:
Using the ACPI provided information on my R50e I get the following 
values under Windows
(the same is available through ACPI under Linux) They seem maybe even 
too-low and there
is the difference, that here the system runs on battery - his 
measurements are on AC:

10.86 W at 600 MHz fixed, fan off, no load
12.60 W at 600 MHz fixed, fan off, full load
16.65 W at 1.7 GHz fixed, no load
23.32 W at 1.7 GHz fixed, full load

(the load included no disk activity, Level 0 LCD intensity, no CD intensity)

The point in using the kernel adaptive frequency code is that your CPU 
runs can run on full speed
when you need it and relax on the low speed when you something like 
typing a letter.
So as a result you get a more quiet computer (less heat, no fan) and 
better battery performance.

Best wishes

Vaclav Stepan