[mythtv-users] Improving Power Consumption on ASUS M3N78 Pro NVIDIA 8300 Chipset

[Michael T. Dean]

On 08/03/2009 01:26 PM, Erik Hovland wrote:
>> Now my media centre PC is also running Fedora 10 and trunk version of
>> MythTv. I followed the same recommendations that I did on my server above
>> but the power consumption is a little high at idle for my liking. I do
>> enable suspend to disk, so that the machine is only running when recording
>> or playing back.
>>
>> Media Centre Specs:
>> ASUS M3N78 Pro onboard NVIDIA 8300 chipset
>> AMD 5200 CPU
>> 2Gb DDR2 (800MHz)
>> 1x 250Gb WD SATA HDD
>> 1x 100MB Ethernet
>>
>> Server Power Consumption: 90-100w Idle.
>>
>> Now I did a little bit of research into this and assume that it is the
>> video processor that is using the majority of current? I also noticed that
>> the NVIDIA drivers have something called PowerMizer that should reduce the
>> clock speed of the GFX core when it is less loaded/idle? However this only
>> seems to function for mobile GPUs? Is there anyway of enabling its
>> function on my chipset or any software which provides similar
>> functionality? I would greatly like to reduce the power consumption of the
>> box especially when it is only recording and not displaying any video.
>>     
> Relevant to this discussion is the Linux Symposium. There was a summit on
> power w/in the kernel. here is the LWN link to the notes taken at that meeting:
> http://lwn.net/Articles/345007/
>
> See how NVIDIA gets dinged for being behind the curve on power
> management when the notes turn to graphics drivers.
>
> Not surprised since NVIDIA has been kickin' tail regarding video
> acceleration.

Also note that the frontend is using a 65W TDP CPU versus the server 
with a 45W TDP CPU.  Now, that doesn't necessarily account for a full 
20W of the difference (as you're more likely to see large differences in 
power usage at load).  However, if you're running the CPU--especially if 
just the frontend CPU--without powersaving features (Cool 'n Quiet) 
enabled or with the Performance governor or ..., it could well account 
for a large difference.  Also, since each CPU's power requirements 
vary--even when comparing 2 4850e's or 2 X2 5200's--you may just have a 
really nice power-efficient 4850e (as compared to other 4850e's) and a 
not-so-power-efficent 5200 (as compared to other 5200's).

Mike