SCIENTIFIC-LINUX-USERS Archives

October 2004

SCIENTIFIC-LINUX-USERS@LISTSERV.FNAL.GOV

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Troy Dawson <[log in to unmask]>
Reply To:
Troy Dawson <[log in to unmask]>
Date:
Mon, 25 Oct 2004 10:25:49 -0500
Content-Type:
text/plain
Parts/Attachments:
text/plain (42 lines)
Ken Teh wrote:
> Hello all,
>
> I'm looking to buy some hardware to set up a little farm for doing data
> analysis and numerical computation.  We've been conservative and used mostly
> Pentium-4 desktop workstations.  But, I'm curious about the SMP Opterons.  I
> hear mainly good things about them in the press.  I'd like to hear what you
> have to say.  Recommendations, gotchas, that sort of thing...  Also, your
> favorite vendors.
>
> Thanks!
>
> Ken

Hi Ken,
I'm a big fan of both Opteron as well as the AMD 64 (single cpu).  When doing
recompiles on them, it just amazes me at how fast they are.
I've also ran standard real life benchmarks on the dual opteron's versus the
dual Xeon am64t (or whatever the name is).  Despite the opteron being a whole
gigahurtz slower (2.2 GHz Opteron vs 3.2 GHz Xeon) the opteron beat the xeon
in every test I ran.
I have not done any formal heat load testing, but the hand test showed that
the Opteron was at least as cool at the xeon.  Neither of them felt to overyly
hot.
Vendors ... I needed good 24x7 support, and only the big names do that for us.
  So I have only tested HP and Sun.  Both of which provided good machines.
I have tested both i386 and x86_64 on the opteron's, and both ran all the
tests equally fast, but then none of my tests were optimized for 64 bit at all.
Gotcha's - both of these are for x86_64 based distro's.
aklog (for AFS) doesn't work out of the box on x86_64 S.L.
Sometimes adding and removing various i386 and x86_64 libraries can mess up
which libraries get's used.  That's being worked on but the bugs aren't worked
out yet.

Hopw that helps
Troy
--
__________________________________________________
Troy Dawson  [log in to unmask]  (630)840-6468
Fermilab  ComputingDivision/CSS  CSI Group
__________________________________________________

ATOM RSS1 RSS2