Quoting Henning Brauer <lists-openbsd@bsws.de>:

quoted text
> * Nick <nick@holland-consulting.net> [2010-06-13 18:43]:
>> >> that might be (I am not convinced tho) with the electricity price in
>> >> the US, but certainly isn't universal.
>>
>> The calculations are.
>
> $/kWh isn't...
>
>> Cost of money (i.e., interest rate), watts saved (if any), cost of a
>> kWh, initial costs, etc.  Plug in your numbers, find out what the ROI
>> is.  Add in what your AC costs are (watts in have to be removed, and
>> that's more watts to pump them out).  Evaluate results.
>>
>> Going simpler, ignoring cost of money, IF your Atom machine draws 50%
>> of the power of my PIII, my quickie calc indicates you will save
>> 105kWh.  If you also have to pay for AC, maybe double that number.
>
> should be less than 50% actually, at much better performance. the
> atoms are surprisingly fast.
>
>> Granted, ROI (Return on Investment) isn't everything.
>
> i would not even remotely consider putting a PIII-era machine into
> service now. the cost of the hardware (in the case of smallish
> systems) is irrelevant in the big picture.
> PIII: old, rusty, reliability questionable, draws more power, adding up
> -> might have to invest in bigger A/C sooner
> atom: new, reliability way less questionable, has modern interfaces,
> saves power, is so cool that it'll survive forever even with all fans
> dead, way faster.
> heck, the supermicro atoms i buy aren't even cheap. not at all. but
> with server-class management, very low power consumption etc, they pay
> out quickly. they even would if they cost twice as much, easily.
>
> admittedly, the math is different for home hobby use.
>
>> Low power rack mount equipment is hard to find now
>
> huh? it is easier than ever.
>
>> analog clamp-on ammeter at the time, but they appeared to draw under 60W.
>
> pretty sure my average for new smallish (you know, 1U, reaosnable
> amount of ram, 1 disk, that style) machines is below that. not idle,
> but with typical workload.
>
>> If you want to talk about "power savings", get a wattmeter and quit
>> reading glossy sheets of one tiny part of the entire computer system.
>
> err, besides a stupid useless wattmeter I have dozens, if not hundreds,
> of points in my power distribution infratructure where power draw is
> measured. live, not artificial test runs.
>
>
>> The numbers will surprise you.  (Fans ALONE on one Dell 1U system
>> draw over 50W at full speed.  Hopefully, they aren't at full speed
>> very often.
>
> either your measurement is screwed or dell screwed up big time.
>
>> My PIII system will pump a LOT of data.
>
> and still lose compared to a reasonable atom.
>
> and for giggles, the dmesg. i forgot the exact power draw of that
> system, it was very very low.
>
> OpenBSD 4.7 (GENERIC.MP) #0: Mon Apr  5 08:50:54 CEST 2010
>     henning@terak.bsws.de:/usr/src/sys/arch/i386/compile/GENERIC.MP
> cpu0: Intel(R) Atom(TM) CPU 330 @ 1.60GHz ("GenuineIntel" 686-class) 1.61
GHz
quoted text
> cpu0:

Just one comment on all this.  It is very rare for me to have a difference
of opinion with you Henning, but I have to comment on P3 equipment.

Dell made some incredible Optiplex models that were white, using P3's
from 450MHz to about 1.2Ghz.  I have several at work in production
service, and some of them are 10 years old.  The disks aren't, but the
machine proper is.  They draw more power than an Atom, thats for
certain, but they are rock solid, and built FAR better than most things
today.  Me, I'm the IT department where I work.  The calculus of spending
more on electricty for systems so stable that they are more likely to
die when the power dies is pretty obvious to me. ;-)

I've watched everything get bigger, faster and cheaper, but usually at
the cost of quality.  This includes my ThinkPads, sigh.  I've had several
conversations where it was admitted that fewer smoothig capicators
were used because a bean-counter saw they could save money by
using fewer.  Boards aren't cleaned any more--I have some great
fingerprint samples of several techs from China.

For applications were speed really matters my little Dell's lose.  But
in many respects they are the best servers I've ever had.

--STeve Andre'