[mythtv-users] Power Measurement (Was Architecture Idea - Raspberry Pi, VM, XBMC)

[tortise]

On 15/01/2013 4:25 p.m., Michael Watson wrote:
> On 6/01/2013 5:28 AM, Raymond Wagner wrote:
>> On 1/5/2013 12:03, Roger Heflin wrote:
>>> On Sat, Jan 5, 2013 at 10:22 AM, Raymond Wagner
>>> <raymond at wagnerrp.com> wrote:
>>>> On 1/5/2013 09:37, Roger Heflin wrote:
>>>>>
>>>>> The baseline Atom board with a good power supply runs
>>>>> about 30-40W at idle
>>>>
>>>>
>>>> Thats awful! My 3.1GHz Ivy Bridge with a middling power supply only
>>>> runs
>>>> about 25W idle.
>>>
>>>
>>> It is a 3 year old machine measured at the outlet plug.
>>>
>>> What kind of meter are you using to measure that?
>>>
>>> My ivy bridge laptop runs 17W at idle with the screen off, and that
>>> is with
>>> a low power laptop cpu vs a high power desktop cpu and for the most
>>> part most
>>> of the desktop motherboards don't do that well on power savings at the
>>> mb level, where the laptop chips are setup for power savings.
>>
>> This is a 55W rated G2120 (Pentium, not i3) on a Foxconn H67 board,
>> and whatever not-80Plus rated 150W PSU Antec puts in their ISK300.
>> Idle is 25W measured at the wall with a Kill-A-Watt, and oddly enough
>> it runs that same value whether running full speed at 3.1GHz or
>> downclocked to 1.6GHz. Perhaps the CPU was doing that on its own
>> automatically when I didn't have the kernel modules loaded to allow
>> Linux control. ATSC playback with software decoding and OpenGL is 32W,
>> and full load on both cores is 45W.
> Or your Kill-A-Watt meter is not really that accurate.
>
> The sampling rate is very low compared to meters like the fluke, so it
> can miss quick spikes or surges in usage . When I used it to measure the
> power usage of a 51" hdtv at turn on, it constantly gave different
> readings. Range from 410 watts to 504 watts. Fluke 448 to 453 watts.
>
> PC, using a 750watt Ultra lsp model.
> Core2, running at 3.3ghz, 7900gtx , pc running gelato (nvidia software
> that uses the gpu as a fpu, while also using the host cpu), maxes both.
> Killawatt 316watts, fluke 329watts.
>

Hmmm time to chip in with my farthing.  As best I can tell accurate 
mains power measurement for switched mode power supplies is at best a 
challenge. The problem is that these do not have a power factor of 1, so 
simple power measuring devices get it wrong as they incorrectly assume a 
purely resistive load.  As Kill a watts were silent on this when I last 
looked I assume they accurately measure a resistive load e.g. an 
incandescent light bulb, however may well be quite inaccurate for fancy 
electronic kit.  An example CCF light bulb I have here has a power 
factor of 0.9.  Who knows what the switched mode power supplies are, and 
beyond that I expect it will change depending on the load....

I have an old Ferranti meter and "True RMS Multimeter" in series and 
they seem to give comparable readings, however whether they are also 
accurate - who knows?!

Best I can tell the issue is more the power factor than the sampling rate.

Example ref http://en.wikipedia.org/wiki/Power_factor