Power consumption -- real vs actual

Tom Rockwell rockwell at pa.msu.edu
Tue Apr 6 10:28:04 CDT 2010


John,

To turn this around a bit, what question are you wanting to answer with 
the power consumption numbers?

Are you trying to size cooling systems?  Size UPS or PDUs?  Evaluating 
operating costs?

I've found the tools are dell.com/calc to be pretty good at reproducing 
what we measure using a power meter.  One caveat is that they give the 
same power for all bins of a XEON CPU --- same for all L, E, X clock 
speeds, which isn't really the case.  This results in overestimates for 
some CPUs.

-Tom




On 4/5/10 12:37 PM, John LLOYD wrote:
> We're coming up against power consumption issues on some servers, usually 10g (r900) and 11g (r710,r610) boxes.
>
> I'll use an R710 as an example.
>
> The nameplate rating, used for electrical safety, is 7 amps for an R710 (100V to 120V).
>
> The power supply rated capacity is 870W (for the high-output version).
>
> The enterprise power calculator dell.com/calc  says input power is 431 watts, maximum 578 watts.
>
> Open Manage reports power consumption (at idle) is 252W.
>
>
> Now, with this surplus of numbers to choose from, I have a few concerns.
>
> First, the enterprise calculator shows maximum (578) less than power supply capacity (870) which is a good thing.
>
> But, second, the power supply capacity is more than rated input which is 7A * 100V=700VA.  What is the basis for the current rating on the nameplate?
>
> (By nameplate I mean the black decal with numerous regulatory logos, input voltage/current ratings, model numbers, etc.)
>
> Third, while idle power is less than calculated (and I do understand why), why so much less?  Try as I might, I cannot get my 252W reported by OpenManage up near the 431W calculated, by exercising CPUs or disks.  So, does OpenManage report "input" power or "output" from the power supply, or, something else?  Or is the calculator based on highest-expected values rather than averages?
>
> Fourth, the Open Manage mechanism to measure power is based on whatever is built into the chassis.  How accurate (% of actual) could we expect this to be?  5% or 20%?  And is it "input" to the power supply or "input" to the baseboard?
>
>
>
> FYI, commands were
>   CPU heater:  dd if=/dev/zero of=/dev/null&     8 to 16 times
>   disk heater:  tar cf /dev/zero /path/to/big/disks/as/raid-5
>   measure power:  omreport chassis pwrmonitoring | grep Reading | head -1
> Config is
> # SPident
> CONCLUSION: System is up-to-date!
>    found    SLE-10-x86_64-SP2 + "online updates"
> Hardware is R710, two Intel(R) Xeon(R) CPU X5570  @ 2.93GHz, (quad-core)
> Six 600GB SAS disks.  One Perc 6/I controller.
> Open Manage is 6.1
>
>
> --John
>
>
>
>
>
>
>
> _______________________________________________
> Linux-PowerEdge mailing list
> Linux-PowerEdge at dell.com
> https://lists.us.dell.com/mailman/listinfo/linux-poweredge
> Please read the FAQ at http://lists.us.dell.com/faq



More information about the Linux-PowerEdge mailing list