• Please review our updated Terms and Rules here

Micro-minis? Whatever happened to them?

Chuck(G)

25k Member
Joined
Jan 11, 2007
Messages
44,225
Location
Pacific Northwest, USA
Back in the early 80's, a number of outfits had the idea to port Unix (usually System III or V7) to any of the 16-bit CPUs with largish memory spaces (Z8000, 68000, etc.), stick them in a big box or rack and sell them as multi-user systems to compete with the minicomputers. Names like Plexus and Onyx come to mind; there were probably many others.

Of course, the bit problem was that these things seldom turned in decent I/O performance and System III on a Plexus 68K box running at 12 MHz was downright ugly with more than one or two users on it. Pretty much a case of someone's eyes being bigger than one's CPU.

Whatever became of these things? Do any survive?
 
No idea, probably cost vs performance and the emerging PC market killed them off. That time period was when people were ditching central computers for personal computers as prices were cheaper for commodity desktops.

Somebody got the idea of turning an IBM desktop into a multi user machine in the late 80's. I have a long 16 bit ISA card from a company that provided 2 sets of video/mouse/keyboard ports so that 2 remote users could run software on one host machine per card (not sure how many cards they could gang together). No idea what OS they used, I havn't tried it out because I am missing the dongles and software.
 
At first I thought this was a "hot weather / small skirt" related thread!

I've had half an eye out for a Plexus for ages, not even seen one in a museum. It was the last machine I got trained up on at CFM/Granada (3rd party computer mainenance place). that was about 20 years ago, and we charged about £1000 a board for repairs (and apparently it was worth doing so)!
AFAIK "Harry" has an OSM Zeus 4, which was a multiuser Z80 based machine.
 
A company that I was associated with had been using a very nice VAX 11/750 and determined that they needed more resources, but not quite as much as another 11/750 would have given them. I'd suggested an 11/730, but one of the directors had the hots for Plexus and bought a loaded P35 (12MHz, 68K). It wasn't long before the employees started making jokes about the thing (as in "the US in Plexus stands for "Unbearably Slow""). Since the firm had purchased it, the system was almost impossible to sell a frustrating year later when it was replaced--with an 11/730.
 
Cromemco had multi-user systems running Cromix, a sort-of-Unix, pretty early in the game, first on a Z80 and then on a 68000/Z80 combo, but even with separate Z80 I/O coprocessors they were pretty slow especially since most of them ran single-user CP/M software that was not really optimized for time sharing.

But their later systems with a 16MHz 68020, accelerator card, math coprocessor, 4-8MB of memory, a couple of relatively fast ST-412 and SCSI hard disks and a number of eight-port I/O coprocessors actually turned in pretty good performance figures running Cromix and/or 'real' UNIX, comparable to or better than equivalent DEC offerings of the time.
 
So the later Cromemco boxes ran, what, System V?

It's kind of difficult to understand how a ST412 could outpace an SMD drive like a Fuji Eagle. I suspect a lot depended on the load.
 
So the later Cromemco boxes ran, what, System V?

It's kind of difficult to understand how a ST412 could outpace an SMD drive like a Fuji Eagle. I suspect a lot depended on the load.
System V, and later System V.2; Cromix was often faster though, and most systems were supplied with UNIX on one partition and Cromix on another, with a common data partition.

The standard configuration for around $40,000 was one or two 150MB ST412 drives, depending on model, from Hitachi, Priam, Micropolis et al. An SMD interface card was available and we did try an SMD drive for a while, but IIRC from a cost/performance PoV SCSI came out ahead (and you didn't need to spend another $1000 or so for a separate card for the SCSI backup tape drive). There was also a 9-track tape drive interface card as well as a network card (all souped-up S100 of course) but I never saw any of those.

Seems that not much is known about Cromemco's systems in the later days; most people just associate Cromemco with an ancient black box Z-2 with a 4MHz Z80 and 64K memory or a System 3 with those funky fast but flaky dual-slot voice coil 8" Persci drives...

cromix.jpgCS400sp.jpg
 
Last edited:
I'm not sure this will be the right target of response but every once and a while you'll still see a company coming out with thin-client computing for offices. I've honestly never seen it get purchased or used but I guess the concept didn't die, it's just a tough sell when the cost of the full blown computer is so cheap and very few folks live in a compartmentalized world unlike the office having lots of terminals and plugging into the mainframe.

A few years ago at a hamfest I ended up picking up an AT&T 7300 system with manuals from a gentleman who realized what it was and wanted it not to just end up on ebay. He said his wife office was using it until about a year before that.. I have no idea what they did or what they ran on it but wow. I guess if the application doesn't need much updating and does the job it needs to they can still work for what they need to.

I think a lot of the "need" for computers on every desk may be a bit more hype/warm fuzzy based. IMHO it became more of a status symbol to have a company computer and be on that new bandwagon.
 
Thin-client is very much alive and well and not a tough sell at all; the hardware costs are indeed somewhat lower, but the main selling factors are much lower maintenance and support costs and greater security when all the clients are identical and can just be hot-swapped, all the updates and security patches etc. only have to be done once in one place, and the desktops can be locked down fairly securely to prevent stealing company data, inappropriate net access, etc.
 
I think most thin client installs are in stores that are tied into the main store stock database which gets tied into the corperate inventory system.
 
Also, isn't MS Terminal Services/RDP just a re-hash of Citrix' thin client implementation?

I think Cromemco's late big systems were a casualty of the ascent of the workstation (Sun, Apollo, Daisy, HP, etc.). For the same bucks, you could get a more flexibile platform. When it came down to everyone running some sort of Unix port, value-added software and system architecture were serious consideration. Cromemco was probably still operating off of the late 70's paradigm. By the time you got to Intel 386 systems, you could get multiprocessor configurations for a fraction of what you'd have spent fiver years earlier.
 
I think most thin client installs are in stores that are tied into the main store stock database which gets tied into the corperate inventory system.
Well yes, but I think you'd be surprised at how many medium to large corporate offices use thin clients on everyone's desk, for the reasons I mention above.
 
Also, isn't MS Terminal Services/RDP just a re-hash of Citrix' thin client implementation?
Yeah, I kinda lost track of who does what these days; most (actually all) of the systems I deal with are still Citrix though.

I think Cromemco's late big systems were a casualty of the ascent of the workstation (Sun, Apollo, Daisy, HP, etc.). For the same bucks, you could get a more flexibile platform. When it came down to everyone running some sort of Unix port, value-added software and system architecture were serious consideration. Cromemco was probably still operating off of the late 70's paradigm. By the time you got to Intel 386 systems, you could get multiprocessor configurations for a fraction of what you'd have spent fiver years earlier.
Well, one factor was that they stuck with the S-100 bus to the bitter end (appropriately, since they'd named it in the first place) and although they managed to push it way farther than the original spec, it was getting pretty tough to squeeze any more performance out of it, even with high-speed "sub"-busses between some cards using ribbon cables on the opposite edge (not to mention the relatively high cost).

And although they always did produce "premium" systems, mostly for government/air force and institutional/university labs, their necessarily higher prices just weren't justifiable any more in the face of the developments that you mention.

There were rumours (and some internet mentions) of a 32-processor "colossus" developed near or after the end by the still operating German division...
 
The micro minis got squeezed out between the micros and the minis is my recall. They were neither fish nor fowl. Too expensive for workstations, not well enough supported for sales into the mini market. An HP 300 series, Apollo D series, or Sun Sparcstation made a better deal--Suns and Apollos on the desktop, HPs in the data center (or IBMs if you could stomach "AIX and Pains".). Masscomp was as close as I ever got to working with one of the off-brands. It was an 8MHz 68K as I recall (maybe a bit faster), Masscomp committed the double sin of bad system and bad support, IMO. And they were more on the side of a mini company than a lot of the others, they had legit onsite support, even if they weren't very skilled.

When I evaluated systems other than those above, the companies seemed to be looking for mini-type margins on micro-style sales and support. Their idea of support was "Don't let the door hit you on the way out!"

I almost picked up an AT&T for personal use, but at the price I balked. There were cheaper ways of getting System V in my home. And at work, I was using SunOS, BSD and HP-UX (a bastard child of System V and BSD at the time), so pure AT&T Unix had diminished appeal for me. Plus, I seem to recall the base system was a bit crippled, by the time it was fitted out the way I'd want it (enough memory and disk), I was in workstation price territory.

I ended up getting a nice little Apollo 68K system with a decent 8-bit graphics board about the time HP bought them. It helped that I got the fixed-frequency monitor for the price of shipping because a company that had leased them got them for free at the end of their lease (the lease agency wanted the workstations back, but not the monitors), and they just wanted to get rid of them.

Aside from the AT&T I wasn't much tempted by the other mini-micros, when their prices were so close to a real workstation or a low-end server like an HP 9000/300.
 
Last edited:
Heh. Back when we got the VAX 11/750, we tried to buy an AT&T 3B2-series mini also. (ca. 1984). It was impossible--the AT&TCS marketing people were idiots--they didn't know what we were talking about and kept trying to sell us a bunch of 6300s. Eventually, we gave up. We figured if the marketing people didn't know anything, that technical support wouldn't be far behind. It probably was a good decision.

Given that Unix belonged to AT&T, you'd figure that they'd cash in on the fact, selling their own iron. Apparently, this didn't dawn on the Computer Systems division.
 
I'm not sure how closely this fits to the micro-mini paradigm, but Ohio Scientific produced OS-65U, which was intended as a timesharing OS for their higher end Challenger III machines. I've got a copy, along with two Challenger IIIs I'm in the process of restoring. Even with the 29 MB 14" hard disk, I'm not sure how fast of a timeshare system it would have made, being as how it relied on the 6502 processor!
 
I'm not sure how closely this fits to the micro-mini paradigm, but Ohio Scientific produced OS-65U, which was intended as a timesharing OS for their higher end Challenger III machines. I've got a copy, along with two Challenger IIIs I'm in the process of restoring. Even with the 29 MB 14" hard disk, I'm not sure how fast of a timeshare system it would have made, being as how it relied on the 6502 processor!

I don't know that a 6502 had sufficient bells and whistles to offer support as a general-purpose mini. No memory protection, limited addressing, and very slow, compared with even a VAX 11/730. But it reminds me of various systems that came out during the 70's trying to cash in on microprocessors. So, for example, you had the Molecular systems, with one Z80 CPU per user (and a separate supervisor CPU).

But the whole micro-mini thing came in big after microprocessors capable of running Unix came out. The 68K was probably the most popular; the Z8000 was also a contender. To get virtual memory support, I believe that Daisy used two 68K CPUs, running one a half-clock ahead of the other to pick up memory references (several 68K instructions were not restartable). I don't know if the NS32K series ever made an appearance in that market or if they were just too late.
 
I'm not sure how closely this fits to the micro-mini paradigm, but Ohio Scientific produced OS-65U, which was intended as a timesharing OS for their higher end Challenger III machines. I've got a copy, along with two Challenger IIIs I'm in the process of restoring. Even with the 29 MB 14" hard disk, I'm not sure how fast of a timeshare system it would have made, being as how it relied on the 6502 processor!

I thought (see Prof. Mark Csele's OSI page, about half-way down the page), that the Challenger IIIs used 3 CPUs - 6502, Z80 and 6800 - and the 6502 only handled the I/O, which was probably well within its capabilities.
 
I thought (see Prof. Mark Csele's OSI page, about half-way down the page), that the Challenger IIIs used 3 CPUs - 6502, Z80 and 6800 - and the 6502 only handled the I/O, which was probably well within its capabilities.

The Challenger did have three processors, at least if you use the 510 processor board. Both of ours have the above mentioned processor compliment. However, only one processor can be active at a time. I don't know about the 6502 handling I/O when one of the other processors was enabled, but the 6502 was definitely the main processor for the system. OS-65D and OS-65U (Ohio Scientific's single- and multi-user OSes, respectively) were both written for the 6502. Additionally, with the software-select option on the 510 board, the system booted into 6502-mode when powered on. There's a ROM-based 6502 monitor, as well as a Z-80 monitor (it's either ROM or loaded off of floppy...don't remember).

Interestingly, the Challenger III with 510 board will actually run CP/M. This process requires a special "escort disk" that boots the machine in 6502 mode, loads the CP/M relevant bootloader into RAM, then software-switches to the Z-80 and starts executing the bootloader. I've got one of the Challenger IIIs to the point that it will execute the 6502 bootloader, but CP/M on the C3 requires an additional 8K of ram at $D000. When I get some spare time, I'm going to build a RAM board for it to provide this extra 8K.
 
The Challenger did have three processors, at least if you use the 510 processor board. Both of ours have the above mentioned processor compliment. However, only one processor can be active at a time. I don't know about the 6502 handling I/O when one of the other processors was enabled, but the 6502 was definitely the main processor for the system. OS-65D and OS-65U (Ohio Scientific's single- and multi-user OSes, respectively) were both written for the 6502. Additionally, with the software-select option on the 510 board, the system booted into 6502-mode when powered on. There's a ROM-based 6502 monitor, as well as a Z-80 monitor (it's either ROM or loaded off of floppy...don't remember).

Thanks for the info, I stand corrected. So, was it true multi-user, or just supporting background processes for one user, like the early versions of MS Windows? I know that OS-65D swaps page zero and the stack to memory buffers when switching between BASIC and the OS, but surely this would be too slow for real-time context switching, and slower still if the hard disk was used instead of RAM? Lunix (Linux-alike for 6502) supports multiple processes, but it divides up the stack between them, which severely limits both the number of processes and the depth of subroutine nesting each can support.

Interestingly, the Challenger III with 510 board will actually run CP/M. This process requires a special "escort disk" that boots the machine in 6502 mode, loads the CP/M relevant bootloader into RAM, then software-switches to the Z-80 and starts executing the bootloader. I've got one of the Challenger IIIs to the point that it will execute the 6502 bootloader, but CP/M on the C3 requires an additional 8K of ram at $D000. When I get some spare time, I'm going to build a RAM board for it to provide this extra 8K.

I'd really like to hear if you manage to get this to fly, as a Z80 and CP/M support are on the long-term wishlist for my Challenger 1P emulation (see my blog).
 
Back
Top