I recently posted a request for peoples' opinions on which
Unix workstations are good for crystallographic number-crunching.
First, MANY THANKS to all those who shared their ideas with
me--I found them very useful. Second, I've compiled a brief synopsis
of the comments I've received, which is appended here:
1. Re graphics--The overwhelming majority are pro-Silicon Graphics.
Period. One or two dissenting voices lamented the
inferior line quality of the SGI's vs., eg., a
PS390, but the only realistic alternative to
the SGI's appears to be an Alpha with Kubota
graphics, which evidently costs BIG bucks (just
how big I don't know).
2. Re number-crunching--Here two clear winners emerged--The DEC Alpha
and the various SGI machines. Other vendors (Sun,
HP, IBM) are not contenders, either because of
inferior performance or because they haven't
achieved a significant portion of this specific
market, and so even though they might be as good
as the SGI or Alpha, there are problems with
availability of software, networking, etc..
Comments on Alphas & SGI boxes ranged from
"I love Alpha it's the best" to "I love SGI it's
the best" to "We have one of each and they're
equally good." After digesting all the comments,
my personal interpretation is that Alphas
and SGI machines ARE roughly equivalent, both
in terms of raw computing power and in terms
of bang per buck.
A recurrent theme was mixed vs. homogeneous
environments. Although respondents did have
environments as diverse as mixed VMS/Unix
networks with machines from 4 or 5 vendors
("...a pain in the butt, but...we have done [it]."),
the general consensus seemed to be that
the advantages of homogeneity were significant
(the advantages being increased ease of system
management, as well as time and space saved
since one needs to compile things once only).
To me, this would give the edge to SGI, since
one will presumably already have SGI graphics
So thanks again to all those who took the time to help
educate me in my efforts to break free of VMS. Further comments
Univ. of Chicago
loll at biovax.uchicago.edu