DECstation 2100/3100 window speeds and differences?

[Daniel S. Riley]

In article <6923 at muvms3.bitnet> mcguffey at muvms3.bitnet (Michael McGuffey) writes:
>We are in the process of evaluating Digital workstations for use as 
>DECwindows terminals (i.e. use them for the standard DECwindows apps and
>some minimal local processing, but mostly as Xterminals with DECterms and
>major applications running on a VAX 11-785/8700/6000-430 VAXcluster (VMS).

I have a mono 8 Meg DS3100 on my disk, and it does work.  It's true that
I can tell when mail arrives by the sound the disk makes as it swaps
sendmail in, and having more than a few local DECwindows programs active 
at once tends to result in a lot of swapping, but I haven't found it all 
that painful.  I regularly run programs on our VAX/VMS cluster (8600/6360) 
along with a few local DECwindows programs, and it works fine.  Running
remote applications doesn't take much local memory, so memory isn't as
critical if you're using a DS3100 primarily as an Xterminal and just running
a few local DECwindows programs.

We don't have any color DECstations, so I don't know how much worse 
the story would be with color.

In a different article thompson at tcgould.tn.cornell.edu (Steve Thompson) writes:
>We have three DS3100's that all started off life with 8MB. Two of them
>now have 12MB, and the third has 16MB. Definitely makes a lot of
>difference! On the 16MB DS3100, we have a f77 program of about 2000
>lines that used to take 5.5 hours to compile when the machine had 8MB
>(yes, that's right, HOURS). LOTS of disk activity. This program now
>compiles on both 12MB and 16MB machines in 3.5 minutes. An interesting
>note here is that the same program compiles on a VS3100 under VMS in
>28 seconds.

Not really a good comparison for Michael's purposes--doing development 
work isn't much like using one as an Xterminal.  If you're doing
development work, extra memory helps, of course.  I have compiled some big
packages (TeX, X11R3, GNU emacs, MIT Scheme, etc.) on my 8 Meg DS3100, and 
the machine isn't good for a lot else while it's compiling.  The compiles
do usually go reasonably quickly--the only exception was web2c, which took
forever with "-O" and a few minutes without it.

-Dan Riley (riley at tcgould.tn.cornell.edu, cornell!batcomputer!riley)
-Wilson Lab, Cornell U.