[Beowulf] Re: Beowulf of bare motherboards

Jim Lux james.p.lux at jpl.nasa.gov
Tue Sep 28 10:09:35 EDT 2004


----- Original Message -----
From: "Andrew Piskorski" <atp at piskorski.com>
To: "Glen Gardner" <Glen.Gardner at verizon.net>
Cc: <beowulf at beowulf.org>; "Jack Wathey" <wathey at salk.edu>; "Jean-Christophe
Ducom" <jducom at nd.edu>
Sent: Monday, September 27, 2004 9:00 PM
Subject: Re: [Beowulf] Re: Beowulf of bare motherboards


> On Mon, Sep 27, 2004 at 11:16:36PM -0400, Glen Gardner wrote:
>
> > I'd like to sequence the turnon of the nodes such that the inrush
> > currents at turnon are lower, allowing the 12V supply to come up to a
> > stable 12VDC a bit faster.
>
<snip>
> I've read that maximum current is usually drawn at boot up, so at
> least in theory, it could be better, and might allow running more
> nodes per supply, if the nodes on each supply DIDN'T all try to boot
> at exactly the same time.
>


There are several sorts of "in-rush" or "plug-in surge" currents:
1) When you first connect the supply to the line AC, there are some fairly
large capacitors that have to charge.  They typically charge in a few half
cycles, though.
2) When you turn on the computer, all the bypass capacitors on the board
have to charge (but I can't imagine that the stored energy is all that
huge.. they only store enough energy to ride through a few CPU or bus clock
cycles, really.. so even at 200W, you don't need much energy to ride through
a few microseconds).
3) The disk drives are spinning up... This is the biggest power draw, and is
definitely related to startup. Once spinning at rated speed, the drive
doesn't draw anywhere near as much power, but while accelerating...

You might want to blow the $30 and buy a Kill-A-Watt or similar to actually
look at the peak and average power draw.  (Or use a suitable oscilloscope
probe and a scope camera or digital scope, or a PC sound card)

Jim Lux
JPL

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



More information about the Beowulf mailing list