[BBLISA] PDU Power Ratings?

[John Orthoefer]

The wattage on a power supply is the Max that it is designed to draw.  And you should NEVER draw that much from it.   

As for 2 PS sharing load.  25-30% increase when one power supply is turned off, that sounds about right.. Remember that most of the power you dump into a computer is turned into heat of which a lot of it is generated by the power supply it’s self.  

It gets complicated, and depends on how things are wired and I’m not a Power Supply engineer.   But as I understand it the efficiency of the power supply is not linear,a power supply that is 75% efficient @ 300 watts (400watt draw, and 300 watt output) @600 watts I don’t think it’s going to be drawing 800watts.  

And a quick asking of the google machine shows just what I’m talking about…

http://www.silverstonetek.com/images/tech/WB10-005/10005-1.jpg

If you look at the graph, for say the 1000w one, you can see at 200W draw it’s 85% efficient, draw 500w and it’s it’s 88%.  So if you had 2 of these running at 250W for your computer and you loose a power supply you’d be converting about 3% better, and you’d lose 15% of your heat load in the chassis. 

Dewey I agree with it’s not simple, and this is a good place where it’s simpler to run a couple of experiments than it is to get out your notebook and calculator and figure it out. 


johno

On Apr 2, 2014, at 6:28 PM, Dewey Sasser <dewey at sasser.com> wrote:

> On 4/2/2014 4:51 PM, John P. Rouillard wrote:
>> ...If that isn't done what happens if the P1 breaker trips? The 1/2 of
>> the load being borne by P1 now is pulled from P2...
> 
> At one point I measured the power consumption of some HP DL380G5s at the plug (VM hosts that we kept around 75-90% CPU load but effectively no disk load).
> 
> Alas, I don't have the numbers any more but I recall that the power draw when running on a single supply was only incrementally more (25%?) than the power draw of each supply when both were operational. This results surprised me as I was expecting what John is expecting:  2x the load.
> 
> I also recall finding that the operational current -- even when operating around 75% CPU load on a single power supply -- was much less than the start-up current and IIRC the startup current would last for 10s of seconds at boot.  (And neither had any resemblance to the 1000W spec on the power supplies.)
> 
> My conclusion was that the issue John points out wasn't actually an issue for me but that if I ever brought down the whole system (which I believe happened once due to building power issues), the power-on current draw could be an problem and that if I started pushing the circuits, stagger-start would be the order of the day.
> 
> G5s are outdated and I'm sure your equipment is very different from what I measured 4 or so years ago, but the take-away is:  go measure something, don't just go on spec.
> 
> In theory there's no difference between theory and practice.
> 
> --
> Dewey
> 
> _______________________________________________
> bblisa mailing list
> bblisa at bblisa.org
> http://www.bblisa.org/mailman/listinfo/bblisa