The Poofygoof Home of Orphaned Technology

pictures and benchmark/amperage graph coming eventually... heckle me.

It all started when a friend of mine offered to rescue a MicroVAX-II from another friend's garage in San Diego and drive it up to Portland... Around that time, a professor I had worked with at Oregon Graduate Institute presented me with a surplus DECStation 5000/240, thinking I might be able to get it going and do something useful with it. That VAX and DECStation formed a critical mass, and before I knew it, a reaction was taking place, and DEC hardware started appearing right before my very eyes. Reed College had replaced all its DECStations with Alphas, and while I was able to help them go to good homes, most of them ended up with me. A year later, OGI was getting rid of its early alphas, and two of them followed me home.

While I currently don't have all these machines online, and don't really have any specific projects for them, I figure it's better for them to sit in my basement waiting for me than waiting for nobody in a landfill.

Someone pointed out the Computer Shelter to me, which seeks to provide a database for museum ``curators''.

Most of these machines run NetBSD, and a flickr group for NetBSD has been created, where you can see other crazy people like me showing off pictures of their old gear.

The Orphans

Most of the orphans' names come from Tolkien's middle earth series, with a given machine's "race" generally following its architecture, (with a few exceptions,) as follows:

i386 Humans The i386 architecture is quite a hack, much like human evolution. They come in all shapes and sizes.
pmax Nature Spirits I inadvertently started the whole architecture/race association when I named my first two pmaxen "goldberry" and "bombadil." Other pmaxen seem to get names of Ents ("treebeard") and plants ("clover").
alpha Elves The Alpha processor is tall (64-bits), fast, elegant, and full of magic, much like elves.
VAX Dwarves VAXen are large, physically tough, and are most at home in cool underground caverns, much like dwarves.
SGI ? SGIs are flashy and artistic, related to the pmaxen by virtue of their MIPS processors.
sparc Orcs These workhorses from Sun have since been obsoleted by the 64-bit ultrasparc. They were the unix workstation of the industry in the early 90s. If I ever get any ultrasparcs, they will be uruk-hais. :)

And now for the machines themselves...

Digital Equipment Corporation Logo
Architecture Name Acquisition Date and Source Model CPU RAM OS Description
DECStation ("pmax") goldberry 1997, OGI 5000/240 MIPS R3000/3010 @ 40MHz 112MB ECC RAM NetBSD 1.4.3A Goldberry has been relegated to secondary service duty (DNS, NIS) for, which it performs reliably. The peripherals, on the other hand, are showing their age.
bombadil 1998, Reed College 5000/200 MIPS R3000/3010 @ 25MHz 480MB ECC RAM NetBSD 3 Bombadil is brought up occationally with the idea of building an updated NetBSD world for goldberry, including needed packages, but something always seems to pull it off track, bombadil gets powered down and moved. It took 10 days to build the NetBSD-3 branch last time I attempted it. With 480MB of memory, it never swapped.
2*unnamed 2001, Reed via Kaben Nanlohy and Mark Jefferys 3100 MIPS R2000/2010 @ 16MHz 24MB parity RAM   The DECStation 3100 was the fastest workstation available when it was introduced in 1989. It seems rediculously slow by todays standards, but back then it was kicking the pants off of the 25MHz 80486, and still does a respectable job with floating-point operations compared to machines of its vintage.
2*unnamed 1998, OGI and Reed 2100 MIPS R2000/2010 @ 12.5MHz The 2100 was a slower low-cost version of the 3100.
VAX gimli 1997, Nick Rahaghi µVAX-II VAX 78032 (0.9VUPS) 16MB parity RAM BA123 case.
gloin 1999, Lake Oswego 8MB parity RAM BA23 rackmount.
unnamed 2001, Chuck McManis VAXStation 3100/M76 VAX KA-43A (7.6VUPS) 16MB parity RAM BA42a enclosure.
unnamed 2001, Chuck McManis VAXStation 3100(/M40) VAX KA-42A (2.8VUPS) 8MB parity RAM BA42b enclosure.
Alpha lothlorien 1999, OGI 3000/400 Alpha 21064 @ 133 MHz 96MB ECC RAM NetBSD 3 These old 3000-class alphas have classic DEC reliability, but are quite the space heaters, and thus are not in continual service. 96MB of memory is also quite tight for Alpha; gcc in particular is very memory-hungry. NetBSD-3 built in a little over four days.
elrond Elrond sports a PMAGB-B graphics adapter with 1280x1024 8bpp display.
unnamed 2000, Chris Chen Multia Alpha 21066 @ 166 MHz 64MB parity FPM RAM (used as ECC)   This machine was given to me after it started suffering from "multia heat death"(tm). I replaced the fan, and thanks to a nimble-fingered co-worker who does surface-mount assembly for a living, was able to also replace the 74F623 on the bottom of the board. After removing the NCR53C810 stub-of-a-SCSI controller and installing some SIMMs, it came back to life. It will run self-tests all day long, but I haven't gotten it to boot yet.
unnamed 2002, OGI Alphastation 200 4/166 Alpha 21064 @ 166MHz   Another leftover from OGI. It will be interesting to compare against the slightly slower 3000/400s since architectually they are quite different machines.
arwen 2005, Reed AlphaServer 1000A 5/400 Alpha 21164A-2 @ 400MHz 256MB FPM RAM (used as ECC) NetBSD 2-0 This machine was previously known as and replaced a pair of MIPS-based decservers as the main academic server for Reed college in 1995 until ~2001. My wife and I both TeX'd our theses on this box. It has taken over main server duties from goldberry.
Silicon Graphics
sgimips 5*unnamed 2001, Reed College Challenge `S' 200MHz R4400SC 128MB parity FPM RAM   Headless SGIs designed as compute and file servers. A few of these landed in the hands of NetBSD/sgimips hackers, who improved NetBSD support on them. I have brought them up occationally for sanity checking, and unfortunately a couple of them are useful only as parts donators due to bad power supplies and other components. The stock drives are unfortunately VERY noisy.
unnamed Indy 200MHz R4400SC 128MB parity FPM RAM The original 133MHz CPU was replaced with one from a Challenge, making this a faster box for the desktop. Unfortunately SGI wants $600 per license, so I won't be IRIXing any time soon.
Sun Microsystems
sparc shagrat 2001, Reed College SparcStation 2 (4/75) microsparcII (CY7C601) @ 40MHz 64MB parity FPM RAM NetBSD -current (1.5ZA) My first NetBSD-running sparc. It has the same clock rate as goldberry, so is interesting for benchmark comparisons.
muzgash 32MB parity FPM RAM NetBSD 1.6.1 provides routing and NAT duties. An SBUS AMD Lance le(4) provides a second ethernet.
3*unnamed     These sparcs used to belong to the math department of Reed College and were named after the cardinal numbers. (one, two, three, etc.) They were replaced with faster sun blades.
gorbag SparcStation 5 microsparcII @ 110MHz 64MB parity RAM NetBSD 1.6.1 This machine was brought up to do some tests with compilation under NetBSD/sparc, but it also does a passable job as an X workstation. It actually has enough CPU to decode MP3s and ogg vorbis files in real time, as well as being one of the few suns in my collection which can produce 16-bit 44.1kHz sound.
2*unnamed microsparcII @ 70MHz none   These poor shell machines have been stripped for memory.
radbug SparcStation 20 2* supersparc @ 50MHz 512MB ECC NetBSD 2-0 Thanks to John Heasley for the dual 50MHz CPUs. Radbug handles incoming mail for, including spam filtering.
unnamed 2005, Ben Shelton Axil 245 microsparcII @ ??     Technically this isn't a Sun machine, but it is a fairly faithful clone of the Sparc 5.
2*unnamed 2002, OGI via TCFKAM SparcStation IPX Fujitsu MB86903 / Weitek W8601 @ 40MHz 64MB FPM parity RAM One of these was originally going to be my new routing box, but unfortunately its NVRAM has conked out along with all my other IPXes, and it suffers from the same problem as lagduf.
lagduf I still haven't quite recovered this box from its dead NVRAM. I replaced the sun NVRAM with an ST M48T02, but apparently it's just different enough that I can't get the TOD clock to permanently start, so it always halts after power on.
Miscellaneous PC-makers
i386 aragorn 2005 VIA VIA C3 Samuel 2 @ 800MHz 256MB 133MHz SDRAM NetBSD 2-0 The original slot-A athlon incantation of this machine died in early 2006, and I scrambled over the parts pile to find a suitable replacement from a number of machines left to be my a departed friend. Aragorn serves as a backup and storage front-end to a DLT library, small RAID, and large IDE drives.
unnamed 1999 Soltek AMD K6 @ 200MHz 64MB EDO RAM Windows '98 Although it is housed in a beat-up IBM case, this machine's motherboard and CPU were purchased new by Kris. She literally woke up one morning, decided she wanted a new computer, and that evening this machine was running. The only thing I seem to turn back to it for is running Outpost II which for some reason remains engaging to me, probably because I haven't been able to master it, even after all this time.
gollum 1993(?), Frye family IBM OEM TI 486SLC @ 25MHz 8MB FPM parity RAM Redhat Linux 5.2 This poofy packet pusher used to handle NAT and firewall duties for This particular machine was given to Kristina by her parents when she entered college. We resurrected it her senior year with Linux so she could \LaTeX from home. Later it replaced metro's routing duties, and has since been usurped by a marginally faster sparc 2. Its memory cannot be expanded beyond 8MB, and its processor is bested by even a 12.5MHz DECStation 2100. It remains a technical curiosity and I believe a good example of early 1990's high-integration PC state-of-the-art.
smeagol 1998, Frye Family   TI 486SLC @ 25MHz 8MB FPM RAM NetBSD 1.5.1 When I originally started hacking on this decrepit machine, it had only 4MB of RAM and an 80MB western digital ``tidbit'' drive. I was able to shoehorn a copy of slackware on it, and networked it to gollum via PLIP, which is how it got its name. My senior year I hand-soldered individual surface-mount DRAM chips to expand it to 8MB, and armed with a 340MB drive and an svgamode postscript viewer, I wrote a significant amount of my thesis on it. After not having touched it for years (thesis flashbacks), it now runs NetBSD and can be brought up on the network via SLIP.
metro 1997, Doug Mandell AMD 5x86 @ 133MHz 40MB FPM parity RAM RedHat Linux 6 This machine is the second incantation of metro, which was originally a 386DX/40. Before learning more about NetBSD and switching my main unix environment over to the DECStations, metro handled server duties as well as NAT, via a ``dedicated'' dial-up PPP line. It's been offline since 2000. I have had problems getting NetBSD installed on it, so it's still running Linux. It serves as an archive of my late 1990s larval stage.
saruman 2000 intel 486DX @ 33MHz 16MB FPM parity RAM NetBSD 1.5 Saruman sits in a black rack-mount case wishing for better days. It came into being when I realized I had no NetBSD/i386 machines at home, and wanted to play around a little with IPv6, routing protocols, and RAIDs built from old SCSI disks running through ISA bus SCSI controllers. But of course once you let one i386 machine in, all the rest come, and so saruman has been passed up.
unnamed 2002, Value Village Intel 486SX @ 25MHz     This poor excuse of a machine was purchased for the gigantic price of $20 at a thrift store by a co-worker who misunderstood me, and assumed I actually wanted it. It does appear to be in fine working order, but it's not a terribly exciting or interesting machine, but at least I've gleaned some useful parts from it. I keep it out of pity more than anything else.
unnamed 2002, Frye Electronics Intel 80386/80387 @ 25MHz 8MB FPM parity An experimentors box, used for benchmarks and answering the question ``will that even work on an original 80386?'' One of these days I'd like to build the NetBSD world on it, just to see how long it takes and if it's even still possible.
boromir 2004, Ben Shelton VIA AMD Athlon XP 1700+ "Palomino" @ 1.467GHz 256MB 266MHz DRAM NetBSD 3 The motherboard and RAM were gifted to me after my friend updated his system. It still (in 2006) seems incredibly fast, and can cross-compile a full NetBSD distribution for any of my other architectures in a few hours.
arathorn 2005, Ben Shelton SuperMicro 2* Intel Pentium III @ 650MHz from Free Geek 1GB PC133 (@100MHz) SDRAM Linux Fedora 5 As much as Linux is not my first choice for a unix environment, I realized that if I was going to do any kernel hacking with my Linux-running friends, I'd need a Linux box. After some wrestling, this box came together in April 2006. I attempted to get some 1GHz processors in it due to the excessive kernel compile times, but it appears to have overheated, and I have not attempted to bring it back to life.

There are plenty of bits around as well, and this doesn't cover the two laptops my wife has, or the studio machine. the PCs keep multiplying as machines are deprecated at work, by friends, etc. So long as they still work, I have a tough time seeing them go to the trash bin. So they rot away in my basement. Perhaps one day I'll get them all running and make my own Stone Soupercomputer which demonstrates how much more efficient modern hardware is. Oh well.

TCFKAM = The Company Formerly Known as Measurecast. It's called Select Metrics now since the name was sold off.

Back to Aaron's Home Page...

Last updated $Date: 2007/06/21 17:40:32 $
Copyright Aaron J. Grier

This page doesn't conform to any real HTML standards due to the color changes in the table. One of these days I'll learn how to do content style sheets and do it right.