[daip] AIPSmark(93) results

[Eric Greisen]

Thanks you.

Daniel C. Homan writes:
 > 
 > We recently installed AIPS 31DEC2000 on a DELL Precision Workstation
 > 620.  I'd like to report our AIPSmark(93) from running the Dirty Dozen
 > tests.  The test completed in a little over a minute and a half for an
 > AIPSmark of 42.2.  This is with the speed parameter set to 40, although
 > we got virtually identical results without setting the speed parameter
 > (I used SETPAR to set it, but I was not sure if there were any
 > additional steps).

     This concerns me - I just tried DDT on a Dell 1.3 GHz machine and
found a time of 153 seconds with the speed parameter not set - a very
noticable delay at the end of every task before AIPS resumes.  The
time dropped to 85 seconds (~47) with the parameter set.  You have to
exit AIPS to RUN SETPAR and then start a new AIPS - it is that program
that really uses the speed parameter and it reads it only once as it
starts.

 > 
 > This machine is a dual Pentium III 933 MHz/Xeon system, although only
 > one processor seemed to be used by the test (I assume this is because
 > the AIPS tasks were run sequentially and are single

       Yes - the 2nd processor does the I/O however making the net
real time rather better than you would get with a single processor.  I

 > threaded -- but if there is a way to utilize both processors for the
 > test, please let me know).  The machine came pre-installed with RedHat
 > 7.0, and after upgrading the recommended packages, I installed
 > gcc-2.95.3 as the default compiler on the system before building AIPS. 

       Thanks you - so many sites have tried to use the 2.96 compiler!

 > We have 512 MB of PC800 RDRAM installed and four scsi disks (1x18 GB and
 > 3x36 GB).  The 3 x 36 GB disks are configured in a raid 5 configuration
 > yielding 72 GB of space for user areas and AIPS data disks.
 > 
 >  - Dan Homan, Brandeis University
 > 
 > P.S. It is interesting to note that this machine replaces an old Dec
 > ALPHA 3000 model 600 with 128 Megabytes of RAM.  This old machine scored
 > an AIPSmark of 3.9 back in March of 1994, and I believe was the record
 > holder at the time.  I find it amazing that for less than a third of the
 > cost, our new box can now support two simulatenous AIPS sessions at ten
 > times this speed.
 > 

It is interesting to run 2 DDTs at the same time under different user
numbers to see how well the machine holds up.

It sounds like a marvelous machine...

Eric Greisen