"Beemer Biker" <antispam@sbcglobal.net> wrote in message
news:10cgtq6s6jhsjed@corp.supernews.com...

I had an opportunity to run the linux version of setiathome on a dual
hyperthreaded xeon (2 x 2.8GHz)that will be available for a few weeks

where

I work.  I first loaded two seti images and noticed they were assigned to
the same cpu (2,3) and that only one of the cpu's ran warm.  I loaded the
other two images and this is what the "top" command shows on the Intel
SE7501HG2 motherboard.  The system was booted from a 1.0gb compact flash
type memory stick with RedHat 9 linux 2.4.20-8Bigmem and there was no hard
drive.

=========FROM TOP COMMAND=====
07:53:01  up 3 min,  1 user,  load average: 3.93, 2.06, 0.81
42 processes: 37 sleeping, 5 running, 0 zombie, 0 stopped
CPU0 states:  99.0% user   0.0% system   99.0% nice   0.0% iowait   0.1%
idle
CPU1 states:  99.0% user   0.0% system   99.0% nice   0.0% iowait   0.1%
idle
CPU2 states: 100.0% user   0.0% system  100.0% nice   0.0% iowait   0.0%
idle
CPU3 states:  99.0% user   0.1% system   99.0% nice   0.0% iowait   0.0%
idle
Mem:  4124724k av,  109844k used, 4014880k free,       0k shrd,    7640k
buff
                     83472k actv,     176k in_d,     256k in_c
Swap:       0k av,       0k used,       0k free                   17660k
cached

PID USER     PRI  NI  SIZE  RSS SHARE STAT %CPU %MEM   TIME CPU COMMAND
 2268 root      39  17 14284  13M   680 R N  99.9  0.3   2:46   2

setiathome

 2265 root      39  17 14284  13M   680 R N  99.4  0.3   2:48   3

setiathome

 2267 root      39  17 14480  14M   680 R N  99.4  0.3   2:37   0

setiathome

 2269 root      39  17 15312  14M   680 R N  99.4  0.3   2:47   1

setiathome

 2324 root      20   0  1024 1024   848 R     0.9  0.0   0:00   3 top

However the statistics are not too impressive compared to a dual 2.1ghz

AMD

MP system as shown below...
http:\\pages.sbcglobal.net/jstateson/wvane/top_sq.jpg
The .43 ip address is the dual hyperthread xeon and it took 4h22m to
complete a work unit.  The .71 ip address is the Asus A7M266-D with two XP
2100 athlons bridged to make them MP capable and they complete a work unit
in 3h31m.  On non-hyperthreaded xeon that I had tested previously, a WU

took

about 3h exactly so there is about a 1h22m loss due to running a 2nd

thread

(if that is the term) on the same cpu.  There seems to be no way to assign
one image of seti to run on one cpu and another on the second.  Linux
assigned the threads two at a time on each cpu.

Here is a picture of a "rack" of dual athlons if anyone has a wish list

this

would probably be on it.  They are not used for seti however.
http://f2.pg.photos.yahoo.com/ph/jysgeek/detail?.dir=/5967&.dnm=b33f.jpg

This is a very common, easily made, production error when running SETI on HT
machines.

The Xeon HT'd machine is indeed a better producer, since while the HT'd
machine takes longer to crunch a WU, it crunches 2X as many WU's in that
time.

ie.,

on avg, in 3h:31 the Athlon machine should crunch 2 WU's.

however,

on avg, in 4h:22 the XEON machine will crunch 4 WU's.

In round figures:

AMD = 2 WU's/3h33 = 1WU/1h:47:

XEON = 4 WU's/4h:22 = 1WU/1h:06

The XEON machine's TOTAL avg production time per WU is ~41 mins less than
the Athlon machine's TOTAL avg production time per WU.

This can further be extrapolated to:

Athlon Machine = 13.5 WU's/Day
XEON Machine = 21.8 WU's/Day

Even though the XEON is a physical dual cpu system, the times posted, and
those shown in your monitoring s/w, are for a 4 logical cpu system, running
four clients (WU's) concurrently.

If you monitor both machines on a output per day basis, the total
production should look similar to my final daily output figures, based on
the SetiQeue log & info you provided in your post.

Strat