[nmglug] Spare P2s all running from one big power supply and irritating Flash Player

[Tim Emerick]

What can I say but.....WOW!

That is about as comprehensive as answers get.

I didn't think about the 2way communications with the ps and the mb.  I would
imagine that would present a problem.  Sounds like if I REALLY wanted to set
them all up and just run BOINC it would be easiest to just stack them up in a
corner and let'em run.

Hmmmm....$30-$60/month for the electric bill seems a bit steep for the amount
of processing power that the P2-400's would produce.  I think I'll set them
up and see what the next months bill looks like.  I don't know that much
about electricity but I sure will notice a spike in the bill.  LOL

The area of your explanation regarding power usage during intensive tasks
(calculating weather/flashplayer) was very interesting to me.  I thought that
the CPU and motherboard components drew a steady stream of power regardless
of what they were doing.  Very very interesting.  I recently installed
BOINC/Seti at home on my two main machines at the house.  They are set to
process after 5 minutes of user inactivity.  Now I'm curious to see if there
will be a spike in my electric bill with just those two computers (P3-500 &
P3-700).

Thanks again for the splendid reply

Tim Emerick


--- WA7BSZ <wa7bsz at yahoo.com> wrote:

> Well, there are a couple things that might happen.
> 
> Sometimes the motherboards like to be grounded to the cases, and may
> not work well without thaty connection.  However, some other
> motherboards do better if they are not connected to the case in
> multiple places.  One of those EMI conditions that is sometimes better
> one way and sometimes better another.
> 
> Another issue is that your power supply connection will be higher
> impedance, and noise from one board will show up on the power supply of
> the others.  So, it might work ok to use one big power supply if you
> hvae nice large conductors going to each motherboard so there isn't too
> much power conductor impedance.  It might not be a problem, but it just
> seems like there might be a power line noise/voltge drop problem if you
> have to put many extender power supply cables in to power all the
> boards.  How would you do this anyway?  Usually there is only one power
> connector set coming out of a power supply.  Plug it into one board and
> you are done.  You could perhaps find a source of those connectors and
> make some multi-connector cables.  
> 
> The more I think about it, It would be a problem.  The computer talks
> back to the power supply, so if you put more than one motherboard on a
> power supply there would be the equivalent of bus-contention.  If there
> were two output lines on the same wire, and one says "its a one" and
> the other says "its a zero" then one of them might get burned out.  
> 
> So, to do that you would probably have to get a bundh of connectors,
> make your own extra cables, and figure out all the signals and how to
> plug them in together without anything burning out.  Maybe you would
> hve to make a little extra circuitry so they could all exist
> peacefully.  
> 
> In order for the P2 to be able to shut down controlled by software, it
> has to have some control lines that control the power supply, so it can
> shut the computer off and even wake-on-lan.  That circuitry expects
> just one master.  If you understood the signaling, maybe you would just
> have one motherboard be the master, and leave that line out on the
> other motherboards.  
> 
> That would be a project in itself.  It could probably be done.
> 
> Let me talk a little on this rating of these power supplies.  Their
> ratings must be for how much power they could deliver for about one
> minute before smoking furiously.  See, they tell you that you need a
> 350W power supply, but the actual power they use is maybe 120W.  That
> is for a pretty hot computer.  If your computer was really using 350
> watts, that would be like having a small heater on all the time.  If
> you had three computers on at the same time all taking 350W, that is
> like having a little kilowatt heater on all the time.  Just fo get your
> electric space heater, say a 600W unit, and put that in your 8' x 10'
> computer room, and turn it on all day and all night, especially in
> August, and see what just 2 computers dissipating 350W each would
> actually do to the room temperature.  
> 
> My 350W supply computer actually draws about .8A or less (I have a
> little current meter).  So, I have to have a 350W rated supply so that
> the computer has good voltage to give to the processor regulator and
> other circuits on the motherboard, so that there aren't any voltage
> dips that cause data problems and cause the computer to lose its mind. 
> But the power supply is usually delivering and dissipating about 100W
> or less.  
> 
> So, if you need say, a 250W supply for each motherboard, and you have
> 8, then you should get a power supply rated for 2000W.  (Might need 220
> for that).  Not because it really needs to supply 2000W, but because
> motherboards are sensitive to some slight variations in their input
> voltages, so they need to over rate the supplies so that they provide
> the voltages in a more stable manner.  Or maybe their stated wattage is
> a selling point.  You don't really get a 350 Watt supply when you buy a
> 350W supply.  Maybe it is like Magahertz and Gigahertz.  The more the
> better, even if it makes little difference really, because hard drives
> still are not much faster than they were 5 years ago.  
> 
> Anyway, it looks like a project that has a chance of some unexpected
> problems in getting all the power at the correct source impedance to
> all the motherboards.  Just thinking out loud.
> 
> Well, you would need to know how much power the CDs and hard drives
> take.  Probably a hard drive takes maybe 5 to 10 Watts.  So you might
> save 50W taking out the drives.  The CDs don't take much while idle.  
> 
> I think it would be pushing it to get two motherboards operating from
> one supply.  I believe you would have to make up your own special
> connectors and patch cords and solve the signal problems.
> 
> My estimate of how much power 8 P2s would probably take.  Well, if they
> actually only take about 60W each, that would be about 500W, only one
> monitor used.  500W for a month is about 360 KWH. If it was 8 cents a
> KWH, that would be about $30.  If it really did take 2000W, that would
> be $120 a month if you kept it on all the time, plus, if you have an
> air conditioner, you would be paying more to cool the area (probably
> not in the garage).  If you could get the air into the house in the
> winter you could use the heat from the power you are paying for.  
> 
> To figure out what one computer uses for power?  If you had an AC
> current meter you could get a worst case estimate.  Find the current,
> and this would be a average over time, and multiply that times your
> voltge which is probably about 118V.  As the computer does different
> things it takes different amounts of current.  On my meter it might go
> up to 1.1A for a second or two, then back down to .6A or so.  Now if I
> had it calculating some weather program, then it would go up to the
> 1.1A all the time and I might want to put a few more fans in it.  So,
> for calculation intensive, always on situation, it would be dissipating
> about 130W.  This is a 2GHz processor machine, and your 400 MHz guys
> would be lower power.  130W for a month at 8 cents/KWH would be 94 KWH
> a month and about $7.50 for the electricity.  This estimate might be a
> little high because of power factor, the current might not be in phase
> with the voltage or the current may be in pulses, not sine waves.  
> 
> Now I am curious to see if there is a good way to figure this out
> without true RMS power meters.  If you have a lot of the same computer,
> like you do, you could probably get a good estimate by turning them all
> on and everything else in the house off, and see how fast the meter
> spins.  Then shut off everything except the dryer, and see how fast it
> spins for that, assuming you know what current that is drawing at what
> voltage.  You might get 10% accuracy, I don't know.  A power meter
> reads differently than a current meter.  
> 
> A way I would trust to figure the RMS power being taken by a computer
> would be to get a 1% tolerance 1 ohm resistor, and put it in series
> with the neutral side of the power going into the computer, then read
> the voltage across it with a scope.  With a dual trace scope you could
> see the voltage and the current and the waveform of the current, and
> you could hopefully calculate the RMS power the computer is taking. 
> Make the computer do calculations.  Run Spice simulations that take 2
> days while taking the measurement.  Then you would hvae a good estimate
> of the power one computer takes while calculating.  
> 
> Computers take a lot different amounts of power depending on what they
> are doing.  Sitting idle, they are at low power.  Boy, go to the wrong
> websites with their flash presentations and you will have 100%
> processor usage and will be drawing the current.  How about that? 
> Flash Player might be increasing the power used by computers by 100%,
> and that is a lot of money considering all computers.  This happens, as
> I watch the computer processor usage, and on the wrong sites the little
> flashplayer advertisements sometimes really get going.  So then, you
> might be paying $5 a month for flash advertising.  A company with 1000
> computers might be paying $5000 a month or more for flash advertising,
> if people surf to the sites and leave their browsers up on those sites
> all night too.  How interesting.  (Kind of like getting junk mail on
> your fax, you have to pay for the paper to get the ad you don't want.)
> Even Yahoo.com has flashplayer ads.  The faster your processor, the
> more power it will use on flash player, showing you those cure little
> irritating animated ads that you can't shut off.  You can go to lower
> quality, you can zoom in.  You can click on the play control, but it
> keeps playing anyway.  "Sorry buddy, we paid for this ad, and its going
> to play on your computer, so get used to it."  That is why I don't
> install flash player.  It comes on some distros, so I have it to some
> extent, but I don't like it much.
> 
> One of my computers was running pretty slow, slow response to user
> requests, so I looked at the processo usage, and it was up around 95%
> most the time (from top command).  I found one website that seemed to
> hve a lot of flash (this was on a distro with flash already installed,
> Xandros I believe) and I shut off that window, and viola, 50% processor
> usage.  Then I found another couple flash sites, and killed them and
> then the processor was at 20%.  It responded a lot better.  So flash
> player can be a real processor hog, and it is on a lot of tech sites
> heavily.
> 
> Anyway, I would say just use one supply per motherboard, and 8 400 MHz
> computers on all the time and calculating all the time might cost $30 a
> month for electricity if they only take a combined 500W. If they take a
> combined KW, it will be $60/month.  
> 
> I wonder if anyone will read all that.  It is a lot. 
> 
> 
> 
> 
> 
> 
> >    Spare P2's and boinc 
> > 
> > 
> >
> ----------------------------------------------------------------------
> > 
> > Message: 1
> > Date: Tue, 8 Mar 2005 19:35:55 -0800 (PST)
> > From: Tim Emerick <timothyemerick at yahoo.com>
> > Subject: [nmglug] Spare P2's and boinc
> > To: nmlug nmlug <nmlug at nmlug.org>, nmglug at nmglug.org
> > Message-ID: <20050309033555.51348.qmail at web40912.mail.yahoo.com>
> > Content-Type: text/plain; charset=us-ascii
> > 
> > I have about 8 p2-400's that I'm looking to turn into some sort of
> > project. 
> > I was going to try to do a LTSP type project but never got around to
> > it.  Now
> > I've decided I want to set them up in the garage to crunch boinc/seti
> > stuff. 
> > Couple of questions for the knowledgeable ones.
> > 
> > 1. I'm curious about clustering.  Would there be much of a
> > performance gain
> > by clustering them together in order to run BOINC or just run them
> > individually.
> > 
> > 2. I was thinking of removing the motherboards from the case and
> > mounting
> > them on a sheet of plywood to be mounted on a wall on my garage.  In
> > addition
> > to saving space/power it would create an interesting conversation
> > piece. 
> > What size power supply would/could I use.  Any problems with all
> > those
> > motherboards running off a single power supply or two?
> > 
> > 3. If I just run them in their individual cases, how would I
> > calculate the
> > amount of power used (no monitors of course).  I would hate to see a
> > $100
> > increase in my electric bill at the end of the month.
> > 
> > 4. In the vein of saving electricity.  What about booting from a
> > single
> > machine in order to save power from multiple cd and hard drives?
> > 
> > I look forward to hearing some ideas.
> > 
> > Tim Emerick
> > 
> > 
> 
> 
> 
> 		
> __________________________________ 
> Do you Yahoo!? 
> Yahoo! Small Business - Try our new resources site!
> http://smallbusiness.yahoo.com/resources/ 
> 
> _______________________________________________
> nmglug mailing list
> nmglug at nmglug.org
> http://www.nmglug.org/mailman/listinfo/nmglug
> 



		
__________________________________ 
Do you Yahoo!? 
Yahoo! Small Business - Try our new resources site!
http://smallbusiness.yahoo.com/resources/