Is Sun moving away from the technical market?

[David B. Serafini]

I see three primary considerations. 1) compute speed  2) development
environment  3) available software

1) It certainly appears that Sun is lagging in compute speed.  The IBM's and HP's
are much faster on floating point benchmarks.  A SGI R4400 Iris is faster than
any Sun I've ever used (I haven't used a Sparc 10 yet), and is affordable as a
desktop machine (which a Sparc 10 might not be).  SGI has put some effort into
marketing their multiprocessors as scientific compute servers (the gigaflop+
CFD benchmark they did last summer) whereas I can't recall ever having heard
Sun doing anything like this for their multiprocessors.  I don't think the 
Sparc multi-procs do very well on the Linpack benchmarks, either.  All this
contributes to the perception that Sun isn't interested in performance.

2) man -k FORTRAN on our sun server brings up f77,fpr,fsplit and ratfor.
Hardly an extensive suite of development tools.  There's sourcebrowser,
which I haven't used but I guess is better than nothing.  f77 2.0 is certainly
better, features-wise, than 1.x, though it can't compare to the Cray compiler
system.  Clock resolution of .01 seconds doesn't make life easier.

3) no complaints here.  There's lots of available software for Suns, but given
that Sun has the largest installed base, it's unclear whether this means
anything to this issue.

-David Serafini
(former research consultant in computational fluid dynamics to NASA)