[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: sparc v. dec 5k

Todd_Kaufmann@ANKARA.MT.CS.CMU.EDU writes:
> What kind of 5000 is lisp-pmax2?  We were going to get '120s, but DEC is
> announcing a new machine next Tuesday, and are discontinuing that line.


> By "feel", your gut feeling when you're on the console hacking (lisp, maybe
> reading mail/news) -- does it really seem fast?  My sparc seems slogging
> generally, but I probably run too many things on it (`ps ax | wc -l` = 62).
> If slow, does it seem fine as a server?

I never sit at the console of the decstations.  I hate the keyboard.
I'd rather use an RT.  Well, maybe it's not that bad.  If I had the
option, I would run X on the RT and telnet to the decstation.  In
fact, I almost never run anything serious on the same machine I'm
sitting at.  But if I had to guess how our machines perform as servers
for big compile jobs, I would rate them:

1.0	lisp-pmax2	DECstation 5000/200, running MACH
0.85	lisp-sun1	SPARCstation 1+, running SunOS w/ Seagate drive
0.4	ardath		SPARCstation 1+, running MACH w/ Hitachi drive

[They all have 24 meg of physical memory.]  I haven't done any
side-by-sides involving lisp-pmax1 (our 3100), so I can't really rank
it.  But it would probably be somewhere between ardath and lisp-sun1.


ps: lisp-pmax2 didn't work worth shit until Gripe fixed the +5 on the
power supply to actually deliver +5.  It kept hanging under heavy disk
load.  If you are buying a bunch of machines, you probably want to
factor reliability in.  My impression is that Sun is a lot closers to
having their act together than DEC, but I don't have any real data to
back that up.

pps: You could always look into some of MIPS' stuff.  If you want to
run CMUCL, a port to RISCos (MIPS' unix varient) would be a lot easier
than an Ultrix port.  I have a friend who works for MIPS, so I figured
I should at least mention their existance :-).  If you want to contact
him, send mail to zalman@mips.com.