Carlos -
Looks to me like the fix data you are looking at for bohr is for 2
processors, whereas the fix data you are looking at for unc/duke is for 1
processor. The performance ratios for pmemd on fix and jac are consistent
with earlier results on the pentium. This is all eyeballing, instead of
ratios. You know that everything here except for GB (and G5 I think) is
pmemd, right? By the way, the "UNC/duke" rubrik is for "benchmarks done
by
Bob Duke who is at UNC, machine is in his home office". This is probably
more clear for the whole pmemd xeon benchmarks page I wrote up and Dave
added to the web site. I did not see much improvement for hyperthreading
in
this stuff; Dave Konerding has preliminary results showing a worthwhile
speedup however by the time you have several dual xeons talking over a GB
ethernet. I am waiting for it to be reproduced (and may also get around
to
doing a 2 dual boxes test myself, one of these days).
Regards - Bob
----- Original Message -----
From: "Carlos Simmerling" <carlos.ilion.bio.sunysb.edu>
To: <amber-developers.scripps.edu>
Sent: Tuesday, April 27, 2004 5:24 PM
Subject: amber-developers: UNC JAC benchmark?
> is that JAC timing of 202ps/day for single 3.2ghz Xeon at UNC/duke
> correct? (on the benchmarks page)
>
> if yes, then why is the "fix" number for UNC/duke much closer to the
bohr
> data?
>
> carlos
>
>
Received on Wed Apr 05 2006 - 23:50:03 PDT
