[ltp] Re: Linux-Thinkpad digest, Vol 1 #1303 - 15 msgs

[Guarded Identity]

On Thursday 29 June 2006 10:15 am, Paul RIVIER paul.rivier-at-eleve.emn.fr |
LinuxThinkpad| wrote:
> alex@clusterfs.com wrote:
> >t42, 14" lcd. the minimum I can get is 7.6W:
> >
> >1) 600MHz @ 0.7V
> >2) rovclock -c 80 -m 100
> >3) minimum brightness
> >4) wireless radio off
> >5) disabled dvd drive (via lt_hotswap)
> >6) very agressive hdd power-saving
> >
> >thanks, Alex
>
> hi,
>
> this power consumption seems very low. Isn't it with backlight off ?
> I have the same type of extrem tweaking, with hdd spin down, no fan,
> 512mo in 1 slot, and most of services stopped, I'm still at 9W doing
> nothing.
> It is a 15" lcd, but I believe there is no 1.5W of difference, so do you
> know what I am missing ?
>
> paul

I too am amazed at 7.6W.  Here's a few of my thoughts on the matter:

I think we're all using 3) and 4) to measure our lowest power consumption, so 
those aren't the issue.  

Is the 0.7 Volts in 1) undervolting?  I'm not sure what the normal voltage is.  
If so, this probably accounts for some improvements.  

With respect to 2), there was another thread of this list that I participated 
in recently where rovclock was discussed relative to dynamic clocking (with 
the generic X.org driver, say).  I think the consensus was that rovclocking 
is statically low, so it should perform better than dynamic clocking in some 
cases.  However, if the system is idle, it seems likely that dynamic clocking 
might perform as well as rovclock (unless the "-c 80 -m 100" switches are 
beyond the range of dynamic clocking).

About 4), I messed around with turning off the bay with lt_hotswap.  I didn't 
see any appreciable difference.  I also played around with just a simple 
power down of the bay using ibm_acpi's /proc/acpi/ibm/bay interface 
(writing "eject" to it).  That didn't do much either.  Is the bay actually 
consuming an appreciable amount of power?

Finally, ATI just released a new fglrx driver, which according to the release 
notes 
(https://a248.e.akamai.net/f/674/9206/0/www2.ati.com/drivers/linux/linux_8.26.18.html), 
seems to offer "thermal event power management."  Maybe there can be some 
addition loss in power consumption by correctly configuring this driver.

I look forward to any else's thoughts about these points.

- Sukant