[Beowulf] power usage, Intel 5160 vs. AMD 2216

[Robert G. Brown]

On Fri, 13 Jul 2007, Greg Lindahl wrote:

> On Fri, Jul 13, 2007 at 08:48:02AM -0400, Robert G. Brown wrote:
>
>> Anyway, if a DIMM draws and average power of (say) 10W and is expected
>> to be on for 3 years, that means it costs roughly $30 over its lifetime.
>
> A few years ago we were trying to estimate power consumption by
> DIIMMs, and found that the measured number was much lower. Apparently
> they aren't all "on" all the time. So you can't estimate it the way
> you're all doing it.

Oh, I believe you.  They don't have a heat sink, after all, and if a
DIMM were burning 10W in 8 chips I think it would get pretty hot.  I was
just stating a number for the sake of argument.  If they really burn
only 3-4 watts, well, that cuts the marginal advantage down to only a
few dollars per DIMM per year.

>
> FBDIMMs seem to be "on" quite a bit more.
>
> I'd love for someone to whip out a kill-a-watt and start comparing the
> same machine at different dimm counts.

I honestly think that this is the only way to get a good measure.  At
different (memory) loads, too -- I wouldn't be surprised if energy
consumption in RW mode is higher than in refresh-only mode.  However,
my KaW is in use on a machine I don't want to turn off right now and I'm
pretty busy doing other things.  Anybody rigged to do the experiment?

     rgb

>
> -- greg
>
>
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
>

-- 
Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu