[Beowulf] Register article on Opteron - disagree

Jeff Layton jeffrey.b.layton at lmco.com
Mon Nov 22 06:15:41 PST 2004


Patrick Geoffray wrote:

> Hi Jeff,
>
> Jeff Layton wrote:
>
>> Did anyone else notice that once the Columbia numbers were publically
>> announced and within a day or two of the Top500 announcement,
>> BlueGene/L suddenly appeared on the list at the top? I thought there was
>> some kind of deadline, but evidently, if you are a sponsor of the Top500
>> site, then deadlines mean little.
>
>
> As a matter of fact, the BluGene number was ready way before the 
> Columbia number. And as all of the major players are sponsor of the 
> Top500 site, your assumption that it can buy you an advantage is not 
> quite logical. 


I didn't know that. Thanks for the update about the IBM numbers.
However, this is counter to what people told me. I usually
discount rumors, but the sources were very reliable.

However, I still stick by the idea that a sponsor of the site has a
little more leverage than others.

> The ultimate deadline for everybody was set for October 25th. As soon 
> as they got their number, SGI/Intel turned on the PR machine and 
> flooded the headlines, vigorously claiming being #1. Of course without 
> knowing for sure. The IBM strategy was to stay quiet up to the 
> official release of the list, and I guess disgrace SGI/Intel at the 
> last minute. Unfortunately, the BluGene number was leaked and 
> published before the official release. 


It's interesting how PR works and how companies position themselves.
It might make an interesting study in human pyschology.

>> However, I guess it's understandable since, as rgb points out, the 
>> Top500
>> is an advertising thing anyway. Doug Eadline has a great quote about
>
>
> RGB is wrong. I will repeat it one more time: the top500 is a 
> statistical study. It is an advertising thing only in the eyes of the 
> believer. If you don't like what the Top500 is trying to do, just 
> ignore it. 


I watch the Top500 for the pure joy of watching a benchmark number
over time. Myself and others have written that one very good thing
the Top500 has going for it is that it has a very long history. This allows
you to watch HPC development over time. As for an absolute benchmark
I agree with rgb on this. It's a beauty contest pure and simple. Besides
there are faster parallel codes for solving LU problems anyway.

>> the SuperComputer show and the Top500. I can't repeat the quote here,
>> but it has something to do with "size." After experiencing the show, the
>> observation is ABSOLUTELY right on target.
>
>
> Do you know any effort involving benchmarks that do not ultimately 
> look and smell like a penis contest ? It's human nature :-) 


Sure I know of several. Larry McVoy's lmbench that rgb has spoken about
is one of the better ones. It's a bit dated, but still useful. I also 
think the new
ClusterWorld Benchmarking project, while still in the forming stages, will
be a good benchmark. However, it's not really a benchmarking project in the
sense of "my machine is faster than yours." It has a much better idea of 
what
benchmarking should be used for.

The way to help out the benchmarking "thing" is to explain to people
how benchmarks should be used and how to always look at benchmark
results with a skeptical eye.

>> Also, the CPUs in BlueGene/L are specialized. I haven't heard anything
>> from IBM about marketing those CPUs in a real system. In talking to
>> IBM they are even reluctant to talk to us about their 970 or Power5
>> systems, nevermind testing them. At least Cray created a product from
>> Red Storm.
>
>
> Ask for the JS-20 BladeCenter if you want to test their PPC970. I know 
> a few customers that have been testing it. You can even buy it from 
> the Web. 


We begged IBM to let us test on a 970 cluster and they wouldn't do it.
We're not sure if it was a marketing ploy or what. We are now working
with them to test the Power5 stuff.

> Has Cray deployed Red Storm machines except the one at Sandia ? At 
> this time, BluGene and Red Storm are both a product, or both not one. 


Have any BlueGene's been announced or sold anywhere else? You can
definitely find orders for the Cray XT3. I don't know of any deliveries
yet, but I haven't looked too closely. My point is that BlueGene at this
point appears to be a one-off kind of thing whereas Red Storm begot a
real product with orders and everything.

Jeff

-- 
Dr. Jeff Layton
Aerodynamics and CFD
Lockheed-Martin Aeronautical Company - Marietta





More information about the Beowulf mailing list