[Beowulf] Register article on Opteron - disagree

Mark Hahn hahn at physics.mcmaster.ca
Mon Nov 22 11:37:10 PST 2004

> I spotted this article on The Register.
> http://www.theregister.co.uk/2004/11/19/amd_top500_loss/

arrg!  as irritating as the article is (in it's misleadingness),
it *is* true that AMD is not doing a good job of marketing the Opteron.

for instance, the mark-up for 8xx chips is nothing short of breathtaking.
sure, 8xx's are used in higher-end systems, where traditionally generous
margins can hide a few grand in price inflation.  but AMD doesn't seem to 
grasp that their continued existence depends on *breaking* this traditional
pricing model!  people buy Opterons in order to break the usual model
of obscenely gold-plated "Enterprise Server" hardware.

further, it certainly seems like AMD hasn't caught on to the idea of giving
really good prices to successful resellers.  that's one of the main ways 
that Intel has so completely smushed AMD into a grease-spot in the past.
I seem to be able to buy a functional dual-em64t node from Dell for around
$1K, but can I get a dual-opteron for that price?  cannot AMD see that the 
only difference in hardware is the CPU?  (OK, e1000 chips vs tg3's, BFD!)

> What's more the article confuses 'supercomputer' with '(super)computer in
> the Top 500'

it's sort of an understatement to point out that top500 is a biased sample ;)

> The fact that there are fewer Opteron based systems in the Top 500 is
> irrefutable (I didn't know this) but it makes me uneasy to extrapolate
> this to the impending death of a CPU.

indeed, if AMD was interested in top500, they'd add a fused-mul-add 
instruction to SSE*.

More information about the Beowulf mailing list