HP 9000 520

From: Frank McConnell <fmc_at_reanimators.org>
Date: Sat Apr 7 12:18:18 2001

Joe <rigdonj_at_intellistar.net> wrote:
> FWIW I do consider the 9000 520 an odd machine. Perhaps unusual would be
> a better term. I've been collecting HP computers for a good number of years
> and I've never seen one before. I found (and bought) one a couple of weeks
> ago. Before that the only person that I know of that actually had one was
> Frank McConnell. My research indicates that there were a lot 520s used by
> the US navy and some of it's subcontractors but few elsewhere. Even my
> contacts at HP tell me that the 520 was problematic and was very late
> getting into production and was a commercail failure. Despite the large
> number of them used by the US Navy, very few managed to get into the
> surplus market. I couldn't find a single surplus dealer that had one or
> that even knew what they were.

The 9000/500 may have been a commercial failure, but HP learned a lot
from it -- its I/O bus (later known as CIO I think) was adopted by the
PA-RISC 9000/800 (and 3000/900) machines. I expect this meant the
machines had trade-in value for their I/O cards (as well as the gold
and copper in the CPU, IOP, and memory cards, which are all
surprisingly heavy from the copper core and shiny from the
characteristic HP gold plating).

The packaging is interesting, especially on the CPU/IOP/memory cards:
the boards don't have sockets for the ICs. Instead, there is a cavity
in the non-conductive board material, exposing the copper sheet at the
core of the board. The IC substrate is bound to this copper core, and
its bond-out wires are soldered to the traces on the board. Then a
lid is fitted over the cavity(ies) to protect the ICs and bond-out
wires. This was done to carry heat away from the ICs, which ran from
a 16MHz clock and did a pretty good job of turning electricity into
heat.

It was also HP's first entry into the Unix market. The 520 could also
be a BASIC workstation (a high-powered multiprocessor replacement for
the 9845!) but I have not seen any configured that way.

It was also HP's last big stack machine, or at least the last one
that made it to market. (Maybe Lee Courtney can tell us more about
Vision.)

HP contracted with The Wollongong Group to provide TCP/IP transport
and application software for HP-UX v5 on these in the mid-to-late
1980s. I started working at TWG in 1989, by which time engineering
support for the product was over, and soon found that from their point
of view, 9000s and 3000s were all the same and so I got to support
that product as well. That was my first introduction to it.

Then I found the FOCUS machine instruction set manual in TWG's
library. Oh my. It's a 32-bit stack machine, sort of a big 3000.
Oh my.

Then I got to investigating a remsh crash that took me deep into the
kernel. That was how I became a TCP/IP stack internals kind of guy.

My memories are dim and distant (TWG ended support for the 9000/500s
in early 1994 I think, due to no sales and no contract renewals), but
most of the customers I talked to were either US military or
contractors.

Its Unix port has some oddities. It was ported on top of another HP
OS called SUNOS that provided low-level generic services to both BASIC
and HP-UX, and the AT&T filesystem was layered on top of an HP
filesystem structure called Structured Directory Format or SDF. SDF
makes itself visible by the absence of . and .. in directories -- ls
and open() will fake them as needed, but if your code opens the
directory and reads it they aren't there!

I don't think it has the concept of partitioning a disc into multiple
filesystems: one disc volume is one filesystem, no way around it.
Swap space is taken as needed from whatever free space is available on
at least the / volume, maybe others too. Strange but nice in its own
way.

-Frank McConnell
Received on Sat Apr 07 2001 - 12:18:18 BST