[Beowulf] Register article on Opteron - disagree

Philippe Blaise philippe.blaise at cea.fr
Mon Nov 22 00:59:16 PST 2004


Robert G. Brown wrote:

> ...
>
>
>  a) It lists identical hardware configurations as many times as they
>are submitted.  Thus "Geotrace" lists positions 109 through 114 with
>identical hardware.  If one ran uniq on the list, it would probably end
>up being the top 200.  At most.  
>
Do you mean it is 6 times the same cluster ?
If not, I don't see the problem.

>
>  b) It focusses on a single benchmark useful only to a (small) fraction
>of all HPC applications.  I mean, c'mon.  Linpack?  Have we learned
>  
>
>>>nothing<< from the last twenty years of developing benchmarks?  
>>>
...

>If they want me to take the top500 list seriously, they could start by
>de-commercializing it, running a pretty stringent unique-ing process on
>the submissions and accepting only the first of a given design or
>architecture, especially for clusters that are more or less turnkey and
>mass produced.  Then they could run a SERIOUS suite of benchmarkS (note
>plural) on the clusters, one which (like SPEC) attempts to provide
>useful information about things like latency, bandwidth, interconnect
>saturation for various communications patterns, speed for a variety of
>actual applications including several with very different
>computation/communication patterns (ranging from embarrassingly parallel
>to fine grained synchronous).  Scaling CURVES (rather than a single
>silly number) would be really useful.
>
>  
>
I agree with you. The linpack had some meanings 10 years ago, before
the (linux) massively parallel machines venue.
Nowadays, in the top500 list, most of the computers have some hundreds 
or thousands of cpus,
but most of the programs that people are using are unable to scale 
beyond, let's say 32-64 cpus !
It's definitly not the case with the "super famous dongara's linpack", 
it always scales, and the top500 list
has become  the list of gigabit ethernet clusters only usefull for 
parametric studies. I wonder when we
will see some grids in the top500, may be in the next few months ?

For us, a "HPC-spec" benchmark should be the right answer, but it seems 
there is no
serious alternative right now, probably because it costs to much efforts 
to do that.
So, the only thing that we can do is to criticize this stupid list >:o

 Phil.



More information about the Beowulf mailing list