My UNIX Workstation Collection

Hewlett-Packard Apollo 9000/735 (c. 1992)
serial number 6338A00031

I almost passed up this machine when it was being scrapped by my employer in San Diego, late 1999. Although I had seen it sitting high up on a lab shelf for about a year, I didn't even recognize it as a computer, thinking it was some sort of signal analyzer with all the BNC's sticking out. However, one of the other engineers eventually did drag it down and fired it up for a laugh before it hit the dumpster. The monitor was too dark to be usable and the hard disk appeared to be on the way out. Despite my knowing nothing about its history or heritage, and being distracted by my newly-aquired Sun IPC - I felt obligated to at least take it apart.

After pulling a few cards out I could see that this was a fairly high-performance and expensively-made machine. I was able to get a longtime employee to remember the root password and fiddled with the monitor's internal adjustments until the picture was restored. Once I could see it worked I wasted no time in "making off with it."

The computer turned out to be a Hewlett-Packard Apollo 9000-series model 730, a UNIX Workstation from about 1992, originally costing in the region of US$30k. It sports the first version of the HP PA-RISC processor running at 66 MHz with 32 MB memory and a 425 MB SCSI hard disk. The HPUX 8.07 operating system installed when I took possession of the machine turned out to be non-Y2K compliant, even though fortunately the hardware was. With a double stroke of luck, I located a 1GB hard disk of the same Quantum SCSI product family at work and obtained a new OS (HPUX 10.20) directly from HP via a free Y2K upgrade program.

HPUX 10.20 has the CDE "Common Desktop Environment" window manager on the X-Windows graphical user interface. The 19" monochrome monitor runs at 1280 x 1024 with four levels of grayscale including black and white. Text and GUI windows are pin-sharp but images cannot be rendered with such few gray levels. Presumably it was a pretty smart performer in its time for floating-point work, as Gamma-Metrics had used it for radiation calculations along with the Sun IPC.

The main proprietary system bus (called SGC DIO-II) runs at 33 MHz and has auxiliary support for one EISA slot, which if you recall was the oddball 32-bit industry alternative to IBM's proprietary Microchannel bus. Most interesting is the machine's high-quality construction. Six plug-in modules at the enclosure rear contain all essential systems: power supply, hard disk, processor, communications, video, and EISA slot. The quality of PCB's and connectors is without compromise.

Although the enclosure's industrial design is a bit boring, the mouse and keyboard are classics. Check out the pics below.

A Year Later ...

There's more to this story - the power supply in the 730 failed in Dec '01 but I was lucky enough to spot a complete model 735 at the next Santee (Southern California) swap meet - which I picked up for a mere $10. The people selling these things usually have no idea what they are and are happy just to avoid having to load it back in their truck. The 735 was introduced in November 1992 and seemed to have the exact-same power supply. It also turns out the machine has an identical backplane and enclosure. These two models were designed in the "old-school" style of modular and easily-replaceble components rather than the single board integrated design of the later HP 9000 range and just about any microcomputer of newer vintage. This "mid-range" model 735 sold for an incredible US$37,400 when new - talk about depreciation!

Although the outside case of the 735 was very rough, the interior was original and untouched, yielding not only the working power supply I needed, but also a faster (99 MHz) processor card with the second-generation (7100) PA-RISC design, an additional 144 MB of the same type of special RAM (wow!,) a color frame buffer, and a communications card with both narrow and wide fast-SCSI-2 ports, sound, and Ethernet in AUI form. I did notice that the RAM had been incorrectly installed, preventing the machine from booting. That, combined with the fact that a minor design flaw made it impossible to remove the CPU/RAM board without complete disassembly of the case may have lead previous owners to believe that the machine had completely failed.

To take full advantage of this incredible find, I swapped over all the internal parts from the 735 to the 730 case and re-installed the OS. Now with 176 MB RAM, a 'new' 2 GB SCSI-2 Seagate drive (from a Mac) the machine seemed much faster than the 730. Next I tried installing HPUX 10.20 using the Journaling File System option. This yielded another noticeable improvement in speed - a surprise to me as there is more overhead.

I transferred the badge and serial number so the unit is officially a 735, circa late-1992.

Not only do these early HP Apollo machines impress me as being the "Rolls-Royce" of micro computers, but while tearing these down I noticed the metal inner case of the 735 was signed and dated by the factory assembler just like a Rolls-Royce engine! The 730 case just had a date, 3-13-92.

As an update, now in December 2006 I've found a new Micropolis 2 GB half-height hard disk which matches the time period (in form factor, not so much in capacity,) and the correct external AMP SCSI terminator.

I've also found out that the CPU is running quite hot. I had transferred a smaller heat sink from the 730 to alleviate the CPU Card removal issue described earlier. I didn't realize that the particular chip affected was the actual CPU (see black heat sink in photo.) A related odd issue is that the 12V fans are connected to a dedicated 9.25V source on the power supply - which could have been intended to reduce fan noise. Experiments with a cardboard airflow duct channeling air specifically to the heat sink in conjunction with a proper 12V fan supply have shown a reduction in temperature from 60° C to 50° C. I've now fabricated a permanent aluminum duct, but may have to live with the low supply voltage for now as I'm reluctant to micky-mouse anything on my most prized possesion.

Despite the overall solid design I think the enclosure's cooling is marginal, especially in the forward hard disk bay which receives no significant air flow at all. The scientist who used the 730 at Gamma-Metrics had mentioned to me that it had an appetite for hard disks.