[tech] (fwd) Re: Another SKC, plus a book review, posted on the Acersoft site

[mustang at ucc.gu.uwa.edu.au_no.spam.please_]

This is what happens when you start a CPU DSW with physicists.
This one gets hella-wacky.

Cheers
/dave


-- forwarded message --
From: dsiebert at icaen.uiowa.edu (Doug Siebert)
Newsgroups: comp.os.vms,comp.sys.dec,vmsnet.alpha,comp.org.decus,comp.arch,comp.sys.intel,comp.unix.osf.osf1,vmsnet.sysmgt
Subject: Re: Another SKC, plus a book review, posted on the Acersoft site
Date: 27 Dec 1999 15:50:22 GMT
Organization: Iowa Computer Aided Engineering Network, University of Iowa
Lines: 46
References: <Fn0oDq.83H at world.std.com> <385F9C47.2F33D5B7 at austin.ibm.com> <38602BB0.51433E4E at igs.net> <rz7vh5liqcv.fsf at corton.inria.fr>

Robert Harley <harley at corton.inria.fr> writes:


>Paul DeMone <pdemone at igs.net> writes:
>>    It is easier pulling a heavy load harnessing 32 clydesdales than
>>    128 goats.

>"If you were plowing a field, which would you rather use?  Two strong oxen
> or 1024 chickens?"

>-- Seymour Cray


Aren't all these sayings pretty pointless?  It is easy to come up with
counterexamples...  If you wish to remove everything growing from a
field as quickly as possible, which would you rather use?  Two strong
oxen pulling a plow, or a half million locusts in a swarm?

For the types of problems Seymour Cray was interested in, what he said
was true.  But the fact that the special purpose ultra high end CPUs
that Cray built pretty much no longer exist has kind of shown there
isn't a very big market for two strong oxen.  Now we are all using
various types of chickens for our workloads, and there's really only a
2x difference between our best and worst chickens (the Alpha chickens
and MIPS chickens, if you will :) )

You can claim it is economics, or end of the Cold War, or whatever, but
if workloads like SAP worked better with two oxen rather than a bunch of
chickens, you can be damn well sure there would still be a market for
Cray.  But my experience has been that 2x CPUs of speed y is always
better than x CPUs of speed 2y for the SAP-type workloads that are
currently driving the high end market.  For certain scientific stuff that
is single threaded, that's not true, and it is too bad for those folks
that there was never a Cray 4 CPU or followon.  They are the ones
plowing a field with a bunch of chickens, and I imagine it is of little
consolation to them if their chickens have "Alpha" stamped on their
heads, because they are still just chickens.  I guess since these
chickens are as strong pulling the plow as their oxen were 10 years ago
or whatever, they don't feel that bad off, but they have no idea how
strong today's oxen would be if they were still being bred...

-- 
Douglas Siebert                Director of Computing Facilities
douglas-siebert at uiowa.edu      Division of Mathematical Sciences, U of Iowa

I'm not too interested in caller ID.  But caller IQ, I'll pay a lot for that!
-- end of forwarded message --

-- 
/  David Manchester <mustang at ucc.gu.uwa.edu.au>  [TDH]  Netware/UNIX droid. \
Tell someone who's interested, tell someone who can keep their lunch digested
Tell someone who wants your conversation, tell someone who doesn't regard you
\as an argument for compulsory sterilisation." [TISM], `How Do I Love Thee?'/