RIP Z80

[coppice]

I worked on a Z8000 based multiuser system with dumb terminals running Pick (anyone remember that?) around 1987 thru 1989, including offering Mac based terminal emulator for it so copy could be keyed into the dumb terminals, then imported into the Mac for desktop publishing.

A friend of mine developed his own Eurocard based systems based around the Z8000 around this time, but as others have indicated, Intel was the way everything was going.

Are you sure you have those dates right, or did you mean the Z80000? The Z8000 was pretty much dead before 1987. I wasn't sure if the Z80000 ever shipped, but if it did, it should have been around 1987.

[woofy]

Yes, Z8000 was first, but was not Z80 compatible. The Z800 was Z80 compatible.. The Z280 was essentially an update the Z800 design.

[brucehoult]

I worked on a Z8000 based multiuser system with dumb terminals running Pick (anyone remember that?) around 1987 thru 1989

An ex university classmate was using Pick at their city council work and showed me around it, but I think it was running on a Pr1me machine (which I wrote a small system in COBOL on in the 82/83 summer holidays at another council, but that was running regular PrimeOS).

My Z8000 experience was a System 8000 Unix machine at university in 1984. It was the first real *nix I used and first exposure to vi. It was closer to a souped-up PDP-11 than to a VAX. I think we would have preferred a 32016 or 32032 machine but they weren't available yet, or too expensive or something. We had a project to support several programming languages (Modula 2 and some local student languages such as Lawrence D'Oliviero's "Peano" on several different ISAs (VAX, Z8k, M6809), and develop an intermediate representation, optimiser etc that could work with all combinations. Not an original idea but no one had made it work well at the time. Of course GCC and LLVM and others (Microsoft CIL) do today.

[SiliconWizard]

Yes; the Olivetti machine. Never sold AFAIK. Had a unix OS.

Nope. Market lifetime was about 2 years, and they sold about 50,000 in the first year, according to the Wikipedia article. Sales plummeted after that, so I don't know exactly how many they sold in total. Probably under 100k. That's not what they expected, but not ridiculous either.

And the OS was not Unix at all - it was some kind of proprietary OS (single user, single task) which was part of the issue. Had they ported CP/M right from the start, the machine may have had a lot greater success. They did provide a CP/M emulator later on, but it was too late.

https://en.wikipedia.org/wiki/Olivetti_M20

its successor was just some PC "compatible" (M24).
It's easy to claim to know, in restrospect, what they should or shouldn't have done, or that the IBM PC would become a de-facto standard, but it was absolutely not in the late seventies, when the M20 (and many other machines) was designed, and the IBM PC wasn't even a thing yet.

I wonder where that firm gets the Z280 chips? Must be from some used stock source.

Not sure. I couldn't find any myself through the usual channels.
The Z180 is still available at Digikey though.

[coppice]

I think we would have preferred a 32016 or 32032 machine but they weren't available yet, or too expensive or something.

You might have preferred at 32032, but only masochists preferred the 32016. It was very buggy. By the time they got the 32032 out and cleaned things up they were too far behind in performance the get anywhere. The 68020 had already got into most of the workstations.

[coppice]

I worked on a Z8000 based multiuser system with dumb terminals running Pick (anyone remember that?)

There was a time when Pick seemed to be everywhere for modest sized database needs. It seemed to be doing great, until the relational databases started to roll in.

[peter-h]

I still make a Z180 product, though not sold any for about a year.

Remember the Z180 was sold in two versions: a "mask 160" or some such, and the latest one which has a broken UART hardware handshake. When the UART bug was found, Zilog did not have the resources to do another mask so they continued to sell the older one under that funny name

I never used the Hitachi 64180 in production, though I did start with a huge ceramic version with an EPROM window. Hitachi sold a lot of the 64180, including lots of the EPROM version.

Good chips, with the IAR C Large Model, straight up to 1MB codespace. You just had to choose the bank size to be bigger than your largest function. In those days (1980s) the banking overhead was negligible, since IAR C was so crap We developed a box converting between IBM coax and twinax, and RS232/Centronics for printer emulation. But other (printer buffer) products sold more, and the Z280 box, with 14 configurable ports, sold so fast we could not make it fast enough. Good days, when a reasonably bright individual could design and sell a product to a general end user application

Olivetti M24 and M28 were my first "PCs". Never had an IBM one.

[gatk555]

Surely, any discussion of the Z80's successors should include Captain Zilog: https://archive.computerhistory.org/resources/text/Zilog/Zilog.Z8000.1979.102646293.pdf. (Link button does not work for me.) I recall seeing the paper comic in a small ICL development office in Ealing, London around 1980.

[coppice]

Remember the Z180 was sold in two versions: a "mask 160" or some such, and the latest one which has a broken UART hardware handshake. When the UART bug was found, Zilog did not have the resources to do another mask so they continued to sell the older one under that funny name

In 1987 Zilog was still launching major things like the Z80000. In 1993 or 94 I remember a friend joining Zilog and the rest of us asking the same question. Why? Zilog was already pretty much dead in the water, and even our friend couldn't give a convincing answer about his move. Such a major downfall, from which they never seemed to recover.

[brucehoult]

In 1993 or 94 I remember a friend joining Zilog and the rest of us asking the same question. Why? Zilog was already pretty much dead in the water, and even our friend couldn't give a convincing answer about his move. Such a major downfall, from which they never seemed to recover.

I hope you mean Zilog, not your friend!

I was astounded in 2005 when my cousin left Coca Cola to join NZ Yellow Pages -- a company that clear as day was going to get steam-rolled by the internet. Maybe he figured it was just a short term CV builder, and anyway after three years he went to a nice safe museum director position he's now had 15+ years.

[peter-h]

Mask 1960 was the one with the non-broken UART

In 1987 Zilog was still launching major things like the Z80000. In 1993 or 94 I remember a friend joining Zilog and the rest of us asking the same question. Why? Zilog was already pretty much dead in the water,

The Z80k was a great chip, and 1987 was before the Intel domination with the x86. But sentiment is like a tidal wave...

I still have somewhere the Z80k C compiler. I believe it was generated entirely with YACC MUFOM tool kit.

Maybe Zilog's mistake was to not make microcontrollers. A smaller company must address a smaller and more specialised market - true for any business. And look at today's uC market, with so many players  co-existing. People would have gone for a Z80 based uC very readily in say 1990. And it was easy... they had the PIO SIO SCC (85C30) CTC DMA. Just throw them on the same chip. Product designers would have almost no work to do. And the custom CMOS ASIC business was well developed in 1990; I was prototyping ASICs on Xilinx FPGAs, using schematic entry on Viewlogic 4, routing with XACT5, so for Zilog to knock up a uC with all this would have been dead easy. A Z80 is only a few k gates, and would fit in a mid level 1990 FPGA like a XC3090.

1976 was hard work for a new uC (the chips were laid out by hand) but 1990 was dead easy.

[mikeselectricstuff]

A couple of decades after that first one, Zilog was pushing a Z80-based microcontroller, and had a giveaway for a design contest. I got a kit. Then I wrote to them that it had a dead short across the power supply, and they sent a replacement.  That one didn't work either.  And that was my last Zilog experience.

I think that was the Z86<something> - if you turned it upside-down, the pinout was compatible with the PIC16C54. ISTR it was mostly used by people making universal remotes as Zilog had licensed a big library of IR codes & made it available free to anyone using their chip.

 

[coppice]

The Z80k was a great chip, and 1987 was before the Intel domination with the x86. But sentiment is like a tidal wave...

Perhaps, but the market the Z80k was targetting was already dominated by the 68k, with people looking towards things like SPARC to replace it. It had no compelling benefits.

Maybe Zilog's mistake was to not make microcontrollers. A smaller company must address a smaller and more specialised market - true for any business. And look at today's uC market, with so many players  co-existing. People would have gone for a Z80 based uC very readily in say 1990.

That's true.

And it was easy... they had the PIO SIO SCC (85C30) CTC DMA. Just throw them on the same chip. Product designers would have almost no work to do.

It would probably have been uncompetitive in 1990 to throw that stuff in. Gate count was still important, and most MCUs were fine tuned, with every SKU being its own die. Successful peripherals on MCUs didn't usually look that similar to successful peripherals in their own package. Today it would be a no-brainer to throw in the kitchen sink.

1976 was hard work for a new uC (the chips were laid out by hand) but 1990 was dead easy.

Wow. I see you never worked in MCUs in the early 90s.

[peter-h]

No but I was doing ASICs of the right sort of gate count. The ASIC business was highly competitive, with sub $1 pricing for 10k+ (yes only 10k!), say 10k gates.

I don't know about today but putting a CPU in an FPGA was never competitive.

Has a Z80 schematic (or VHDL) ever been published? Various 3rd party ones exist e.g. https://opencores.org/projects/a-z80

[CatalinaWOW]

CPUs on a gate array makes no sense in new designs or where chip cost drives program cost.  But if there is a large software investment, or other large one time costs involved it can be a huge cost saving.

[RoGeorge]

Found this, pointed by Hackaday, an open source Z80 in hardware:
https://github.com/rejunity/z80-open-silicon
https://hackaday.com/2024/04/28/the-z80-is-dead-long-live-the-free-z80/

[peter-h]

Many years ago I was asked to do an FPGA plug-in equivalent for some TMS9900 peripherals. These seemed to have a funny silicon problem in that they all died after some years.

The size of the FPGA needed was surprising - something like a (then) Xilinx 3090. Not cheap at about 100 quid.

It was also a huge job. I quoted the guy about 20k for the design. He didn't go for it.

[SiliconWizard]

Found this, pointed by Hackaday, an open source Z80 in hardware:
https://github.com/rejunity/z80-open-silicon
https://hackaday.com/2024/04/28/the-z80-is-dead-long-live-the-free-z80/

Yes, not sure what the point really is compared to a small-FPGA solution (you only need a pretty entry-level FPGA for implementing a Z80), even with added level shifters if you absolutely need a replacement for existing, vintage products.
Or even a software emulation on some moderately powerful MCU can be a solution - with its PIO, I guess that a Z80 could be emulated on a RP2040. You may run into GPIO limitation though. They should definitely release an updated version of it with more GPIOs, an updated PIO and maybe M4-based. That would be a killer.

[bson]

I think we would have preferred a 32016 or 32032 machine but they weren't available yet, or too expensive or something.

You might have preferred at 32032, but only masochists preferred the 32016. It was very buggy. By the time they got the 32032 out and cleaned things up they were too far behind in performance the get anywhere. The 68020 had already got into most of the workstations.

Yeah, the 32032 was way too late to market.  The 32332 was a significant step up and saw some success though.

[peter-h]

you only need a pretty entry-level FPGA for implementing a Z80

Has anyone tried it?

[RoGeorge]

There are quite a few Z80 designs for FPGAs already, e.g.:  https://opencores.org/projects/a-z80

[peter-h]

Aha I did see that project but on a quick look didn't realise they actually put it in an FPGA.

The EP2C20F484C7 is GBP 102 and they are using 11% of it. The Spartan-6 6x1s16csg324-2 I could not find. The devkits all list as obsolete.

Yes; this is what I found years ago when looking at FPGA versions of the TMS9900 UARTs. Very expensive back then. The UART used up a chunk of a GBP 200 Xilinx 4k FPGA.

So to be doing this, you would probably be working on a project which needs a big FPGA anyway, and there aren't many of those. I see things like $5000 ARINC429 interface cards using 3 digit priced FPGAs...

It is not efficient to use FPGA logic structure for a design which was totally hard-wired.

[RoGeorge]

Since most of the projects are giving the HDL sources, it means it can be recompiled/ported to any big enough FPGA, older or newer.  Z80 is a small CPU, so it should fit in any small/average FPGA.  I'll say the design should fit in most $50-$100 FPGA devboards, though I didn't check how big it is exactly.

Also, ignore the astronomical prices for FPGAs at Digikey and such, nobody buys them at that price.

[peter-h]

A couple of rhetorical questions:

Can you fit the Z80 into a $5 (100+) FPGA?

Since you will still need an EPROM and an SRAM (for 64k address space you won't be using a 8GB DRAM module ) you still have a couple of tricky chips to source. EPROMs do exist, just about, FLASH chips definitely exist (but you need to design the PCB for in-circuit programming of a parallel FLASH chip - not trivial, lots of test points for a spring-loaded jig) and SRAMs also exist although only a few vendors nowadays. Can you get a cheap FPGA with say 16k RAM? I had a quick look on Mouser and 16k bytes is pushing it. A few k seems to exist in the $10 range.

[bson]

If you want to experiment with a "soft" Z80 on an FPGA, get a quality devkit if you don't have one already, like a Digilent Arty A7-100T.  While obviously more expensive than a Z80 it's also not a one-trick pony and can be used for all kinds of future FPGA experimentation and design work.  A kit like this will have tons of memory, JTAG, serial port, USB-UART, Shield connector, LEDs, buttons, etc etc.  As much as I dislike Arduino, a shield header makes it very easy to add things like an SD card slot.