[Beowulf] Benchmark between Dell Poweredge 1950 And 1435

Robert G. Brown rgb at phy.duke.edu
Fri Mar 9 08:40:53 PST 2007


On Fri, 9 Mar 2007, Mike Davis wrote:

> As usual, excellent information RGB.
>
> The only thing that I might add which you alluded to is that even given the 
> same processor, and memory an application may or may not be faster on one 
> machine as opposed to another. A quality MB can offer you increased 
> performance.

This is one of the unstated reasons to consider (and if possible test)j multiple vendors, in
particular vendors who actually serve the HPC market.  Dell is more HA.
Penguin is more HPC.  I don't know if this makes a difference in HPC
performance -- it could be that both use high quality mobos and
chipsets.  OTOH, it would hardly surprise me if it DID make a
difference.

> Then there's the housekeeping matter of keeping it cool. Can the fans move 
> enough air to keep a rack full of X running efficiently? If you have to fill 
> racks only half way or create 6 to 10 ft space buffers, you make things more 
> difficult from a practical standpoint when scaling to larger numbers of 
> racks. Failing to consider these matters can lead to excess failed nodes down 
> the road.

Both good points.  And those of us who have been around a while recall
the "thermal throttle" that Intel in particular had installed on their
CPUs for a while.  You'd buy what was it a 1.6 GHz CPU at great expense,
then run it flat out and as it got hot, it would suddenly underclock to
800 MHz?  It isn't clear that this sort of behavior is impossible even
now -- an Intel roadmap presentation I went to several years ago still
referenced an expected high burst rate for certain tasks with a much
lower throughput for ongoing tasks as they throttled down to what the
silicon could actually take at the given clock.  Think of what a pain
this would be to determine -- you run a benchmark for a two minutes and
get speed A, but if you run the SAME benchmark for twenty minutes you
get speed B...

So there are many virtues to keeping your server room cool...:-)

      rgb

>
>
> Mike Davis
>

-- 
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





More information about the Beowulf mailing list