12

I have read that the first version of Unix was created for the PDP-7, and later versions were created for the PDP-11.

But I am wondering, what was the first Unix version to run on a microcomputer?

  • 11
    This question is at least as blurry as the definition what a microcomputer is - isn't it? – Raffzahn Mar 29 at 11:55
  • 3
    Define "microcomputer". The first 64K RAM desktop with a framebuffer running UNIX was Terak in 1976 or 1977. – Leo B. Mar 29 at 20:48
  • 1
    Define "Unix". Linux, for example, famously isn't Unix because it has never been certified as meeting the specification. – Mark Mar 29 at 21:39
  • 1
    @Mark en.wikipedia.org/wiki/Inspur_K-UX – Leo B. Mar 30 at 0:47
  • 1
    @SolarMike Mark said "it has never been certified as meeting the specification" and I proved that many distros have been certified as meeting the specification. What's wrong? Have you read the comment? In fact it's not one but two – phuclv Mar 31 at 9:55
30

In 1981, Microsoft released Xenix which could run on the x86, the 68000, and others. As you know, Microsoft targets microcomputers above other markets; that was especially true in the 80s. So this was perhaps the first unix, or at least among the first commercial unices, to be intended to run on high-end micros.

But maybe, the term "micro" isn't so useful. After all, Version 6 UNIX from 1975 was written in C and ran blithely on the PDP-11. The PDP-11 architecture was later implemented on a microchip; A J-11 or something like that would have run V6 just as happily. And in Soviet Russia, V6 was indeed modified to run on local PDP-11-compatible microcomputers. It's arguable whether MNOS and V6 are the same UNIX; if they are, then it predates Microsoft's offering by around five years.

BYTE magazine, October 1983 talks about early unices on microcomputers

  • 5
    I agree this doesn’t deserve its downvote. The release date for Xenix is somewhat inaccurate (it was announced in 1980, but only shipped in 1981, although there are claims of a 3Com release in 1980) but that’s not all that important! – Stephen Kitt Mar 29 at 12:24
  • 3
    I always think it is fun to remind Unix bigots/Microsoft haters to remember that for several years Microsoft's Xenix had more systems out there than any other Unix! – davidbak Mar 29 at 17:09
  • 1
    "Microsoft targets microcomputers above other markets; that was especially true in the 80s." This sounds like you're saying this was more true in the 80's than it is now, which is definitely incorrect, no? Or, do you mean it was especially true in the 80's compared to the 70's? – DarthFennec Mar 29 at 20:14
  • 1
    I was under the impression that smartphones and modern servers are microcomputers. There's not much difference internally between a phone and a laptop, or a server and a desktop. Or am I misunderstanding what a microcomputer is? – DarthFennec Mar 29 at 21:00
  • 1
    Eh, it's a matter of what definition of "micro" you like I guess – Wilson Mar 29 at 21:04
12

It all comes down to the definition of micro. After all, already the PDP-11/03, aka LSI-11, of 1975, is based on a micro chipset. That's less than a year after Unix became known outside AT&T. An 'official' AT&T port to the LSI-11 was published in 1978. Not much later, the PDP-11/24 (1979) used the F11 chips follow us where J11 and T11 (single chip PDP). Clearly micros (*1).

When it comes to more common (*2) CPUs, Intel's 8086 got its first port, by AT&T in 1978, shortly followed by Microsoft's XENIX, based on the AT&T port (*3).


*1 - Or do the case form make the difference? Then I must point to some 6502, Z80 or even 2650 based systems in 19"

*2 - not that I would call the LSI-11 uncommon

*3 - Sans the custom MMU that is.

  • Wow, I had read the porting paper a while ago and had completely forgotten about the AT&T 8086 port! Was Xenix 8086 really based on that though? I thought Microsoft had started with PDP-11 Xenix, then ported it to the Z8001 and only after that to the 8086 (with help from SCO). – Stephen Kitt Mar 29 at 12:20
  • @StephenKitt Not sure. It's the way I learned about it back in the 1980s - but I never verified it, so it may be unfounded coffee machine talk. Then again, isn't the AT&T itself based on the PDP-11 code? Interesting maybe in addition is that the Siemens PC-MX, a 8086 based Xenix (Sinix) system, had a custom MMU, somewhat like the AT&T one - on a ..lets say intel inspired ... CPU board. Similar for the later 80186 based PC-X workstation. – Raffzahn Mar 29 at 12:30
  • 1
    Raffzahn, What did you mean in Note *1? Are you saying there were Unix systems for 6502 and z80? Or that there were microcomputers based on these 8-bit CPUs? I'm not challenging your answer, I just don't understand this note. – RichF Mar 29 at 15:48
  • @RichF No, *1 is related to what qualifies a microcomputer. It's meant to mark that the form factor can't be used as reasoning. – Raffzahn Mar 29 at 17:46
9

According to the chronology of personal computers, the first micro-computer running Unix was “the Onyx C8002 microcomputer. It features a Zilog Z8000 microprocessor, 256 kB RAM, tape backup, hard disk, serial ports for eight users, and running UNIX, for US$20,000.” It was introduced in June 1980. (The quote is inaccurate in at least one respect: the C8002, as its name suggests, used a Zilog Z8002 CPU — thanks Raffzahn!)

This was followed shortly by Xenix, Microsoft’s licensed port of Unix, which first shipped in January 1981 on a Z8001-based Central Data Corporation system (not to be confused with the more famous Control Data Corporation). 8086 systems running Xenix started shipping in 1982. Xenix was also available on PDP-11 computers (see the OEM list for details), and there were micro-computer-class PDP-11 systems, including DEC’s own Professional range which was available with 2.9BSD.

  • I do not get it. How could 256kb RAM brick of metal cost 20k? It is new, super silicon ninja tech, but why it was used so it could cost 20k? – Croll Mar 29 at 12:27
  • 5
    Erm, @Croll, you did note, that this is the price for the whole system, including CPU, I/O, (hard) disks, tape and so on? 20k doesn't sound much to me for a 16 bit multi user system in 1980 - rather resonable I say. – Raffzahn Mar 29 at 12:37
  • 1
    Terak was not "personal" enough? – Leo B. Mar 30 at 1:11
  • 1
    @Leo you’d have to ask Ken Polsson ;-). Was it “micro” enough? – Stephen Kitt Mar 30 at 8:43
  • From Wikipedia it seems that Z8000 was a family including (at least) two members - Z8001 with 24 MB addressable memory and Z8002 with 64 KB. So calling this processor Z8000 or Z8002 is both correct. – Bulat Mar 31 at 19:01
5

With the other answers, it is obvious the Sun-1 was not the first. However, due to Sun Microsystem's relevance in the microcomputing world, I think this May, 1982 entry is worth mentioning here. The system was based on a 68000 CPU along with a Sun-designed MMU. (Apparently the Motorola MMU at the time was not sufficiently reliable.)

  1. Sun-1, 1982, 68000, Sun MMU
  2. Sun-2, 1983, 68010, Sun MMU
  3. Sun-3, 1985, 68020, Sun MMU, Motorola fp math coprocessor
  4. Sun-3x, 1989, 68030, inherent MMU, Motorola fp math, system coincident with SparcStation 1

No more 680x0 models past that, and later Sun models would either use the Sparc or x86 CPUs (both 32-bit and later 64-bit).

  • 1
    Why did Intel not use Motorola chips when they asked Microsoft to provide a disk operating system for their new microcomputer... We'd have an 8 core 68080 ... sane assembler ... the world would be a better place ;-). – Peter - Reinstate Monica Mar 29 at 18:13
  • 1
    @PeterA.Schneider: If you mean IBM, rather than Intel, using the 68000 was seriously considered fir the IBM PC. It wasn't done largely because the 8088 could use cheap 8-bit peripheral chips: remember that the IBM PC was not intended to become a dominant computer architecture at all. – John Dallman Mar 29 at 18:47
  • @JohnDallman IBM indeed, sorry. – Peter - Reinstate Monica Mar 29 at 18:56
  • 1
    I don't think there was actually anything wrong with the Motorola MMU other than it was slow. This made it unpopular, HP also created their own MMU for their 68000 based workstations. The problem was with the 68000 CPU itself, a design flaw meant it couldn't reliably restart execution after a page fault. This meant that it wasn't possible to implement virtual memory, but it was possible to use the MMU to implement process isolation. VM support had to wait until the Sun-2 which used a 68010 which fixed the 68000 flaw. – Ross Ridge Mar 31 at 3:01
  • @RossRidge Thank you for the info. Do you know why Sun stuck with their proprietary MMU until the 68030? My guess would be momentum. They were used to it and had no compelling reason to change. – RichF Mar 31 at 3:05
4

In the early 1980s, Torch Unix was available for the BBC Micro.

The catch was that you had to have the Torch 68000 second processor and hard drive. This meant that it fell between two stools. It was far too expensive for use by the usual BBC Micro hobbyists and office/professional users found it too unusual to compete with more mainstream offerings from other manufacturers. So it soon died out.

I did use this at work for a short time. I remember it had to be installed onto the hard drive from 50 5.25" floppies, this took me pretty much an entire working day.

  • I remember seeing Torch Unix running at one of the Acorn User shows at Alexandra Palace. As you say, too weird to survive - unfortunately. – Chenmunka Apr 3 at 9:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.