[mythtv-users] Cost of running devices 24/7

Michael T. Dean mtdean at thirdcontact.com
Thu Sep 10 19:14:28 UTC 2009


On 09/10/2009 01:49 PM, jansenj wrote:
> I was curious the other day and wanted to know how much it costs per 
> year for some of the electronic devices I have.  Maybe someone with a 
> kill-a-watt can tell me if my rule of thumb is remotely accurate.  I'm 
> just too cheap to go out and get a kill-a-watt, but I probably should 
> get one.

I /highly/ recommend getting one if you're really serious about saving 
money on power costs.  In the US, you can get them for $20 + $6-$8 S&H 
or for $26-$28 with free S&H when they're not on sale.  On sale, they 
tend to be $20 with free S&H.  It's easy to recoup that cost on power 
savings.

>   The numbers I ran assuming about an 11cent/kwh gave me the following 
> rule of thumb for devices running 24/7...like myth. 
>
> Replace the unit of watts with dollars and that how much it costs to 
> operate the device per year if the device uses that power rating 
> 24/7.  So if you have a device that CONSUMES 5W nominally, its costs 
> close to $5/year to operate that device 24/7. If your whole mythbox 
> uses a minimum of 50W, it costs at least $50/year to operate.
>
> I thought it was an interesting how the numbers came out to give me 
> that rule of thumb. 
>
> Obviously you can't rely on power supply wattage ratings of you 
> computer, but when comparing idle wattages between two products, the 
> rule of thumb may come in handy.

I did some very eye-opening tests which lead me to believe that the only 
measurement of power usage you can trust is a direct measurement.  
Basically, I found that just swapping out a power-supply unit (and 
leaving all other components of my system identical), I was able to save 
about 21W average draw (equating to about 500Wh/day or about $20/yr at 
your $0.11/kWh).  See 
http://www.gossamer-threads.com/lists/mythtv/users/371460#371460 for more.

TDP, ACP, and pretty much all the other power usage numbers tend to be 
way off compared to reality because they're meant for 
system/infrastructure designers, and not for actual power usage 
estimates, and they have to cover all the chips that come off the line 
and get put into a specific bin.  You can actually build a system that 
idles such that the /entire/ system power draw is less than the CPU's 
TDP rating.  And, even under max load, most CPU's draw significantly 
less power than their TDP rating would indicate (see 
http://www.lostcircuits.com/mambo//index.php?option=com_content&task=view&id=59&Itemid=42&limit=1&limitstart=7 
and note that the Athlon II X2 250 with the arrow pointing to its 30.8W 
max power consumption under load is a measurement of a 65W TDP 
processor--also check out the previous page of the article which shows 
idle usage).  The Intel chips are also way below TDP (though I don't 
know their chips as well, so I'll leave the researching to you), but AMD 
tends to "overestimate" their TDP's a bit more (see 
http://it.anandtech.com/IT/showdoc.aspx?i=3606&p=2 for more on Intel and 
AMD TDPs and AMD ACP).

Anyway, without direct measurement, anything is just a guess and may not 
have the effect you hope.

I will say, though, that the /best/ approach for saving money in Myth is 
to shut down systems when not in use.  In trunk (soon, 0.22), there's 
support for having the master backend shut down and wake up remote 
backends.  And, even in 0.21-fixes, non-backend (dedicated) frontend 
machines can shut down when idle using MythWelcome/mythshutdown, or the 
master backend can shut down and automatically wake up for the next 
recording using mythshutdown (and MythWelcome if it's a combined 
frontend/backend system).

Mike


More information about the mythtv-users mailing list