disadvantages of linux cluster - admin

Peter Kjellström cap at nsc.liu.se
Fri Nov 8 11:53:27 EST 2002


[snip]
> Also, the way some people throw around how "Athlons use a lot of power and 
> make heat" or "Tyans use a lot of power" are simply propagating urban 
> legends (a legend that Intel does everything it can to fuel).

As I type this I realise that this topic ranks quite high on
flame-top-10, but I'll take my chances ;-)

First, your facts below are, to put it mildly, incorrect. Power max as
given by AMD and Intel for their cpus are unfortunately not 'measured' the
same way. To get some hard figures I grabbed my clamp meter and went into
the computer room. Time for a reality check:

dual MP2000+     idle: 0.5-0.6A (kept changing, 0.55)
dual Xeon 2.4GHz idle: 0.2A
dual MP2000+     load: 0.7A
dual Xeon 2.4GHz load: 0.5-0.6A (kept changing, ~0.55)

Obviously my clamp meter isn't very accurate but the relative changes
should hold. If the numbers seem awfully low keep in mind that we are
running on 230V here.

The 'load' in this case was a kernel 'make -j3' plus a 'badblocks' so both
disk IO and cpu was used. 

Both systems were 'cluster-assembled', that is contained not much more
then MB, cpu, nics, mem and one disk.

Conclusion 1) it's unfair to the xeon, I should have picked a 2.0GHz...
Conclusion 2) idle Xeon ~ loaded Athlon
Conclusion 3) the diff between idle and load is greater on the Xeon system

> 
> Fact: AMD design max power per CPU is around 45W
> 
> Fact #2: Intel P4 at 2GHz has a design consumption peak of 76W. Of course 
> getting the P4 to actually use that much is sometimes elusive as they speed 
> step the moment you approach that.
> Very safe, if all you wanted as 1GHz CPUs!

And you are talking about urban legends?! Our Xeon cpu's are cooled only
by the airflow through the box (which in our case is very good). No
specific cpu fans. Have I seen any termo-safes go off? No. Not a single
cpu has ever gone into termo protection mode without reason, not even when
running really intensive tasks and sucking in air 5-10 deg C warmer then
normal.

Hoping for a world with less urban legends,
 Peter

> 
> Fact #3: A XEON P4 2.2GHz eats up 87W of power.
> 
> These have all been tested here, by the simple method of testing with a 
> digital Ammeter, adding up the consumption on all lines, when the CPUs were 
> at 99% while generating Distributed Net keys.
> 
> A dual XEON box makes almost DOUBLE the heat of a dual AthlonMP.
> 
> As for admin time, it has little to do with hardware, once you get properly 
> built boxes.
> Sometimes you might have to pay a little more up front for that, but if you 
> don't you pay it later.
> Lastly, the power consumption translates into costs that often outstrip the 
> machine costs in a pretty short time period.
> 
> Power costs and cooling costs will generally exceed machine costs within a 
> year. And that is not even counting the capital investment in the cooling 
> equipment, or the nearly doubled initial investment cost for a properly 
> built dual XEON machine.
> 
> 
> 
> With our best regards,
> 
> Maurice W. Hilarius       Telephone: 01-780-456-9771
> Hard Data Ltd.               FAX:       01-780-456-9772
> 11060 - 166 Avenue        mailto:maurice at harddata.com
> Edmonton, AB, Canada      http://www.harddata.com/
>     T5X 1Y3
> 
> Ask me about NAS and near-line storage
> 
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
> 

-- 
------------------------------------------------------------
  Peter Kjellström               | E-mail: cap at nsc.liu.se
  National Supercomputer Centre  |
  Sweden                         | http://www.nsc.liu.se


_______________________________________________
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