Chronic DEC hardware problems

[utzoo!decvax!ucbvax!unix-wizards]

>From decvax!duke!bcw at Berkeley Sat Dec 26 21:25:35 1981
The various VAX hardware problems which have been discussed in the
net are not limited to UNIX installations.  We are running an 11/780
with VAX/VMS (no flames, please;  although we feel both systems have
good and bad points, VAX/VMS runs some software [like RSX compatibility
mode and PL/I] which we feel we need;  UNIX lookalikes as subprocesses
under VAX/VMS have so far satisfied our UNIX users although they are
not as nice as straight UNIX), and DEC has had a great deal of trouble
getting it to run.

When the system was first delivered, it was blowing up power supplies
and memory like mad (we went through about 6 power supplies in two
months);  typically, none were in the local office and would have to
be sent out to Charlotte or Atlanta to obtain some.  The total turn-
around time was often 3 or 4 days;  this would mean that the system
would remain down for the remainder of the working week.  It was so
unuseable that we almost sent it back and replaced it with a PDP-11/70
or an IBM-4300 (UGH!!!) machine.  DEC blamed power problems, grounding
problems, and everything they could think of;  we rewired much of the
power and had them redo much of the power stuff in the machine;  we
have still had problems with blown power supplies (although not quite
as much).  Background on our power supply:  it was used to power a
XEROX Sigma-V machine back many moons ago and consequently is drawn
directly from the transformer at the power station with 00 (or is it
000?) cable.  The machine room is also from the Sigma-V days;  the
grounding and air conditioning (both blamed by DEC for our troubles)
also date from that time.  The Sigma-V, being 1965 technology, drew
MUCH more power and produced MUCH more heat than the VAX does, so we
suspect that the real problem has to do with their switching power
supplies (rather than the more reliable transformer-based power supplies
on older equipment).

We are not the only installation at Duke to have problems:  the Physics
Department (who also run VAX/VMS) have had similar problems with their
11/780.  Sometimes it seems to help DEC to be able to see the error
log, but that does not seem to be the usual case.  I suspect that thi~
is just another case of "I don't know what's going wrong so I'll blame
anything I can" which is typical of DEC field service (in our exper-
ience, which has lasted about 8 years now).

				Bruce Wright @ Duke University