ALR CPU's

[Greg Wettstein]

This article is being composed from an ALR 386/220 running XENIX 2.3.1.
Actually as you can tell from the mail header the mail is not coming from my
machine here but rather I am in a terminal emulator running under VPIX
connected to a mainframe.

I spent about two months trying to get XENIX up and running on the ALR and
I hope these comments will be of use.

I started out and installed 2.3.1 on a stock 386/220 with a 70MB micropolis
drive and 2 meg of ram on the motherboard.  Everything installed just fine
and ran like it was supposed to including VPIX.  Two meg of memory is somewhat
limited so I purchased the 32 bit memory expansion system which allows the
machine to be expanded to 6 meg (2 on motherboard, 4 on expansion).  On this
machine the memory expansion system consists of two identical boards which
plug into two special 16 bits slots which are designated as 32 bit expansion
slots.  Evidently the CPU and memory controllers know how to go through the
necessary gymnastics to gain 32 bit accesses to these boards.  I stuck the
boards into the machine, re-configured the CMOS and kicked up the operating
system.  XENIX came up, I did a ps -e and the kernel panicked.  I worked on
the machine for about two days, RTFM'ed extensively, tested memory with
various DOS utilities to no avail, the system still panicked.

I called ALR, described the system and was put through eventually to customer
support, (my local dealer is helpful but with no knowledge whatsoever of
things beyond DOS and some LAN stuff).  It should be noted that ALR is not
real eager to put just common folk through to their techie types, it took
some polite, forceful complaining to indicate to the people that there was no
competent local help available.

Their technical person was very knowledgable about the machine but offered what
I considered to be somewhat odd advice.  I had 80 nanosecond ram on the MB and
the expansion boards had a mix of 80 and 70 nanosecond ram on it.  The support
person immediately indicated that my problem was due to the mis-matched speed
of the chips on the expansion chassis.  I have stuffed a lot of boards in the
last five years and have had trouble with adding chips slower than indicated
but never trouble with faster chips.  I decided that they must know what they
are doing and replaced all the 70 nanosecond parts with 80's and fired things
up, kernel still panics.......

I got ahold of technical support and convinced them that the mix of 70 and 80
second parts was not it.  He then made the off-hand remark that they had some
reports of trouble with the memory expansion cards.  So I called my vendor,
he shipped me two new sets of boards with new memory which he had also put in
a 386/220 and run for a weekend with no problems.  Installed new boards,
kernel still panics.....

At this point I began figuring that there was a fatal flaw somewhere with the
motherboard and the addressing of the 32 bit memory on the expansion chassis.
The other thing that bothered me was that the CPU had a large heat sink
firmly cemented to the top of the chip.  The 386/220 I had was of about 1988
vintage when 20 MHZ machines were first starting to appear.  The thought kept
running through my mind, 'I wonder if they stuck 16 MHZ parts in these early
machines and over-clocked them.......'.  What was especially troubling was
that the heat-sink kept me from learning very much about the processor.

At this point I had spent about two months monkeying around and our project
was hopelessly behind schedule.  We put a lot of pressure on our local vendor
and he came up with a brand new 386/220.  This was just a box with a MB, no
memory, boards etc.  I transplanted everything out of my machine into the new
machine, kicked up XENIX and .....  no problems, everything including VPIX
ran flawlessly and has continued to run flawlessly 24 hours/day since that
time (2 months).

Since we had purchased the machine less than a year ago our dealer to his
credit left the new chassis with us and took back my old machine.  The last I
heard he was negotiating with ALR about a replacement motherboard.  From what
I understood ALR did not want to warranty the board since although the machine
was in use for less than a year our dealer had taken delivery on the machine
about two months prior to the time he had sold it to us and ALR claimed that
the machine was effectively out of warranty.  I guess that issue belongs in
the realm of dealyer <-> supplier negotiations.

The story does not end at this point however.  I learned from my dealer's
technical support person that they spent a lot of time dealing with ALR's
technical support department about this motherboard.  At some time during the
process it was acknowledged that some (perhaps early board runs) MB's had
severe DMA problems which produced problems with XENIX and/or NOVELL
applications.  It would have been nice if the technical support people had
bothered to tell me this the first time I contacted them.....  As it turns out
there is probably a pretty good chance that the first set of memory boards were
probably not troublesome although our vendor did have some difficulties with a
few of those boards.

I guess the upshot of all this is that our ALR 386/220 in its present
incantation effectively and efficiently runs XENIX.  It was unfortunate that it
took about 2 1/2 months to get to that point.  It is also interesting to note
that the CPU on my new motherboard does not have a heat sink cemented to the
top of the chip and the chip proudly proclaims itself as an 80386-20........

I should also point out that SCO technical support was very helpful throughout
the process.  They ran through all the hoops verifying that I did not have a
bad version of the kernel or of VPIX.  Considering that it was a total
hardware problem I couldn't ask for much more.  I would also like to thank
Karl Denninger whom I met through the net for spending a fair amount of time
talking to me about the problem even though I did not purchase any of the
hardware or software from him.  I would also like to thank the net for
providing a forum where a lot of my problems get answered before I need to
call the companies who sometimes seem much less knowledgable than the net
gurus.

                                      As always,
                                      Dr. Greg Wettstein
                                      NU013809 at NDSUVM1

P.S.: If I had to do it again what would I do??  Tough question, my group
      is committed to a number of XENIX installations.  Considering the
      multiplicity of MB's, display adapters, drive controllers etc I am not
      sure how you get systems that will definitely work.  I am beginning to
      think that the thing to do is to find a VAR or a dealer who specializes
      in XENIX/UNIX hardware and software and buy your system from them.  I
      am sure that those guys have probably spent a lot of time trying to
      piece stuff together that they know works.