What do you guys thing about the P4

Phillip D. Matz matz at wsunix.wsu.edu
Sat Apr 14 20:28:42 PDT 2001


> or heat dissipation, it does so because it's too hot. Bottom line, with
good

This would make sense to me, but from what I have read so far the specs are
for heat dissipation and not operating temperature.  The spec is reported in
such a way that it appears that the P4 1.5GHz is throttled regardless of the
operating temperature if the power exceeds 54.7 watts.  I would like to
believe this to mean that the throttling occurs at some cut-off temperature,
but temperature is not directly proportional to power (you get to play with
heat capacities and thermal diffusivities first).  I can only assume you are
correct though, a temperature cut-off is the only thing that *makes sense* -
why would Intel go to all the trouble of having a wattage limit?

Thanks for the input!

Phil Matz

----- Original Message -----
From: "Peter Kjellström" <cap at nsc.liu.se>
To: "Phillip D. Matz" <matz at wsunix.wsu.edu>
Cc: <beowulf at beowulf.org>
Sent: Saturday, April 14, 2001 3:13 PM
Subject: Re: What do you guys thing about the P4


>
>
> ----- Original Message -----
> From: "Phillip D. Matz" <matz at wsunix.wsu.edu>
> To: "Eddie Lehman" <elehman at sunflower.com>; <beowulf at beowulf.org>
> Sent: Saturday, April 14, 2001 11:03 PM
> Subject: Re: What do you guys thing about the P4
>
>
> > Since most people building Beowulfs are doing so to achieve some set of
> > performance goals - what does the cpu throttling do to a P4-equipped
> > cluster's performance?
>
> If one or more nodes in your cluster 'drops down' you'll probably see some
> unpleasant behaviour. However, the P4 doesn't throttle down because of
load
> or heat dissipation, it does so because it's too hot. Bottom line, with
good
> cooling (may it bee quite alot more then intels 52W) the cpu doesn't get
> overheated and therefore keeps working at full speed.
> So, if you intend to build something yourselfs plan for some good cooling.
> I'm quite sure that most OEM solutions will be well cooled.
> On the clusters we have built so far (T-bird) I run lm-sensors to keep an
> eye on the temperatur. If I had to deal with a P4 cluster the only
diffrence
> that I can see would be the effect of hot cpus.
> hot AMD t-bird -> *hot* cpu = bad for the cpu
> hot P4 -> slow cpu = bad for performance
> Either way, something is wrong with the cooling, may it be room, box, or
> cpu...
>
> just my 2 c.
>
> /Peter
>
>
> --
> ------------------------------------------------------------
>    Peter Kjellström
>    National Supercomputer Center
>    Linköping University
>    S-581 83 Linköping
>    Sweden
>
>    Email:cap at nsc.liu.se
>    Telephone: +46 13 281492 (office) +46 13 282535  (fax)
>    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