vishy@catinhat.Princeton.EDU (V. Visweswaran) (06/13/91)
We are looking for a compute-server for our lab of 5-6 users. Right now, we have narrowed our
options to two choices :
(1) An HP 9000/730 with 48-64 Mb RAM, 1Gbyte disk space.
(2) A Silicon Graphics 3D/310 server. The salesman for SGI has assured us that this would
be upgraded to the MIPS R4000 chip when it is released. Again, the configuration would
be 48-64 MB RAM and about 1 Gbyte of disk.
(Both configurations cost roughly the same)
The server (HP or SGI) would be to drive 4-5 X terminals and also act as a file server for a couple
of workstations from other vendors (MIPS/DEC). Since our lab is essentially a chemical engineering
design lab, we would be running a lot of optimization algorithms, which tend to be very floating-point
intensive. At any time, we expect some 4-5 big jobs of this type to be running in the background, so
a primary factor in our choice is that the performance of the server for driving the terminals
should not degrade too much even when these jobs are running in the background.
Another factor that we need to consider is the expandability in terms of upgrades to faster CPUs.
In the case of the SGI machines, these are inherently built for a multiprocessor machine, so it seems
like these would be more expandable. In the case of HP, we are not sure how expensive/easy it would
be to upgrade to any newer (faster?) chips that they might come up with.
I have looked at most of the articles posted in this newsgroup regarding comparisions of the
new HP machines with other vendors' products, but have never seen any comparison with the
SGI 3D-or 4D-series servers. Also, there seems to have been little discussion about how the HPs
would compare to the R4000 based machines (when they come out ). The stats for the two would
be roughly the same ( ~75 MIPS, 20MFLOPS etc etc ).
Unfortunately, we have not been able to persuade the HP salesman to give us a demo machine,
so there is no way for us to evaluate the machine ourselves. I am curious as to whether someone has
actually tested these machines in a multi-user environment, and if so, which one has the better
performance.
Thanks in advance for any help in this regard.
--
V. Visweswaran
--
Bitnet: viswswrn@pucc | Department of Chemical Engineering
Internet: vishy@catinhat.princeton.edu | Princeton University
Tel: (609) 258-6754 | Princeton, NJ 08544