Hello,
Slightly off topic, but I didn't know where else to ask this question
(if
someone does, please let me know). We are about to buy a new machine
which will be used exclusively for heavy calculations. We are of
course looking for the best performance/price ratio. Colleagues from
us told us that their Dell-PC is from 20-50% faster than their
AlphaStation (depending on the application and compiler used).
Dell-PC:
Dell Precision Workstation 610
Pentium II Xeon, 450 Mhz, 256 Mb Ram, 9.0 Gbyte SCSI, 151000 BF
Red Hat Linux release 5.2 (Apollo)
Kernel 2.0.36
AlphaStation:
Alpha 500/500, 1 Gbyte Ram,
Digital UNIX 4.0B, Digital Fortran 5.1 -O5
This seems strange given the big difference in price and the following
specfp95 numbers:
Dell 610 (450MHz) 15.2
AlphaStation 500/500 20.4
Compaq XP100 (ev6 alpha) 52.2
I know spec numbers are to be handled with care, but I think the
difference between 15
and 20 should be considerd as being significant (or am I wrong on this?)
The specfp95 for the Dell was obtained with NT as OS. Using UNIX instead
of NT will of course also have its impact on the performance, but I
wouldn't
expect the difference to be that large, would it?
We asked them if they were sure that they tested cpu-performance and not
for
example I/O performance, but they assured us that the test jobs they ran
did very
little I/O.
If we go ahead and by a new Alpha it will be a Compaq XP100 (ev6 alpha).
Given a specfp95 of 52 this machine will be faster then the Dell
machine
(at least that is what you expect from the specs) but it will also cost
a lot
more money.
So the question is how can we determine whether we should buy a new
alpha-based machine or one or more intel-based machines?
Thanks in advance,
Bart.
--
Bart Rousseau, Ph.D. student
University of Antwerp - Dep. of Chemistry
Structural Chemistry - Quantum Chemistry
http://sch-www.uia.ac.be/struct/quantum/
Received on Thu Mar 11 1999 - 17:54:55 NZDT