The eight-bit Z80 is dead. Long live the 16-bit Z80!

The Z80 has a long series of successor models – some compatible and some not. There are multiple options for hobbyist computer builders. Zilog has announced that it will end production of the original, classic DIP-format, pin-through-hole eight-bit Z80 processor. Although there will be stock in the channel for a while, all the …

[deanb01]

Agon eZ80

As the author of the original Agon Quark firmware (you linked to my Git) and long time Z80 developer, I've grown to like the eZ80. It has its quirks, the most annoying being that there is no straightforward way to access the top 8 bits of 24-bit registers without storing the register in memory first. Yet that not withstanding, being able to access more than 64K without paging is incredible, and the three-stage pipelining means that most code without modification runs much faster. It's also got some neat extra instructions such as MUL. And the C compiler, provided by Zilog, works quite well on it.

It implements kind-of-backwards-compatibility by letting you run a standard Z80 program in a 64K page in memory - the Agon allows the user to run older 16-bit apps in a 64K page or 24-bit apps. I ported the first version of BBC BASIC for Z80 by R.T.Russell originally in 16-bit mode with very few modifications (the full-fat 24-bit version required much more work - source on my Git if anyone is interested).

And it's a complete Z80 SOC with DMA, UARTs, timers, etc.

That all notwithstanding, one can't beat the original in its 40-pin DIP package!

[Anonymous Coward]

Re: Agon eZ80

If memory serves, the eZ80 was also the first microcontroller to support a fully embedded TCP/IP stack.

Unfortunately, after the moderate success of the eZ80, bean counters took over Zilog and over the years they squeeeeeezed it for every bit of personal profit they could pay themselves until the company had shed 80% of its staff and was a shadow of its former self.

[Pete Sdev]

Re: Agon eZ80

Picked up an AgonLight 2 the other month due to reading too much of El Reg's retro computing week, very pleased with it.

BBC BASIC running on a Z80 is the best of both (80s ) world, even if does feel a little perverse.

Thanks for your contributions! -->

[deanb01]

Re: Agon eZ80

I owe an immense debt of gratitude to R.T.Russell who wrote the original Z80 version and who kindly open sourced it a few years ago. I've ported it to a handful of computers since, including the Spectrum Next, Cerberus 2080/2100 (also by the creator of the Agon, Bernardo Kastrup), and my homebrew Z80 computer. Thankfully Richard's source code is very easy to work with. The 16-bit version took a weekend to port over. The 24-bit version is still a work-in-progress - currently refining the inline assembler to handle eZ80 mnemonics.

[Steve Graham]

Before I graduated, I did a lot of 6502 coding. I wrote structured extensions (WHILE, IF-THEN-ELSE etc.) to the Basic interpreter on my Ohio Superboard. But when I started work it was on Z80 embedded systems. We only had an 8080 compiler though, and I wrote a peephole optimizer that replaced common 8080 sequences with Z80 instructions.

Of the two, I preferred the 6502. "Page Zero" essentially gave you a bank of extra registers, and most of the instructions were regular, or "orthogonal", as I believe the experts say.

Later, I did VAX 11-780 assembly coding, and that instruction set is very orthogonal. So the best 32-bit microprocessor CPU was definitely the Nat Semi NS32032, which was "inspired" by the VAX.

[usbac]

I still have my Ohio Scientific Challenger 1P (the boxed version of the Superboard). I haven't powered it on in many years.

[Rich 2]

Wow. A Ohio Challenger. That was my very first introduction to computers. My secondary school started doing “computers” for the first time when I was in 4th year. No exam or anything - just for fun!

Later that year (maybe the following year?) they got some of those new fangled BBC things. Didn’t have the same punched tape storage though :-(

[CowHorseFrog]

6502 orthogonal ?

Not sure how you can say that given each of the registers had very different abilities, Accumulator could add bytes, but x and y couldnt, x and y could increment and decrement but A couldnt. The Accumulator couldnt be used as an index in anyway to memory, Y could be used as indirect to an address in zero page, but x couldnt and so on.

There was hardly anything orthogonal at all.

[toejam++]

The original MOS 6502 definitely wasn't orthogonal. The WDC 65C02 was better, but there were still many shortcomings. The Synertek SY6516 was probably as close as you were going to get, but I don't believe it ever made it beyond development.

As a final project for one of my university comp-sci classes, I wrote an emulator for a 6502-like processor that was as orthogonal as possible. But it required the use of an extend byte so I had room to fit all of my indirect, stack relative, and index register ops that couldn't fit within the first 255 op codes. Some folks considered that heresy, though.

[CowHorseFrog]

The WDC6502 was hardly any different from the original 6502, its still the same in the end, just an A, X, Y, SP, PC. If it wasnt then programs would work.

For your emulator, then its not a 6502, if you are going to add a prefix to provide the register arguments then its a completely different cpu. Secondly your new cpu would work in a real system, because if it had extra registers they wouldnt be preserved/restored in interupts Any interrupt that used one of your new registers would be trashing the same register outside the interrupt.

THi sis the entire reason why the MMX registers are mapped to the FPU registers.

[Version 1.0]

The Z80 taught me so much

I was just a young hardware worker soldering when the Z80 appeared, I'd been reading about the 8080 and I started building a Z80 unit running code from an EPROM and sending data to a pair of ADC's, I manually digitized clinical EKG data on paper and wrote the code to create a dual channel EKG simulator in paper too, then manually converting the code to digital values that I manually stored in the EPROM - I was so happy that it worked great and my company paid me to make four more simulators.

A few years later I was doing CP/M programming, MSDOS programming and then move to the DEC RSX11M world - Gosh "assemblers" ... After learning to easily manually code Z80's on a few sheets of paper had taught me so much and tripled my annual corporate income because the company was so happy.

[STOP_FORTH]

Time to replace my Rodnay Zaks?

[Dan 55]

I'll give you my Zaks when you pry it from my cold dead hands.

I don't want to start thinking about how long that'll be.

[bombastic bob]

the Z80 is a simple chip by modern standards and is easy to implement on an FPGA

as stated in the article. So we could either wire it up with different pin assignments or build a sort of daughter board with the programmed FPGA so that the pins line up.

So yeah. Worth mentioning at any rate.

[deanb01]

Re: the Z80 is a simple chip by modern standards and is easy to implement on an FPGA

You'd have to throw in a bunch of level shifters as the Z80 is 5V, FPGAs are typically 3.3V.

[John Brown (no body)]

Re: the Z80 is a simple chip by modern standards and is easy to implement on an FPGA

"You'd have to throw in a bunch of level shifters as the Z80 is 5V, FPGAs are typically 3.3V."

Not a problem. There are already similar "boards" out there in retro-land to act as pin compatible replacements for rare, hard to find and/or unobtanium chips. Well, i say "not a problem", but that doesn't mean I could do it!! But better people than me have done it :-)

[Dan 55]

Z180 is also no more

Also listed as end of life by Mouser. Only the eZ80 is left.

Now that is a little sad, as fixing original hardware means making an FPGA replacement instead of an adaptor board for a Z180.

[the spectacularly refined chap]

Re: Z180 is also no more

Won't be an issue for donkeys years yet. It was earlier this year I picked up up a dozen 8085s for perhaps £15, out of production since the millennium and produced in smaller quantities than either the Z80 or the 8080.

[diodesign]

Re: Z180 is also no more

Ah, OK, thanks - we've tweaked that part. Don't forget to email corrections@ if you spot something that could be improved, please.

cheers,

C.

[theOtherJT]

Just a thank you...

...both to the z80 for all the good times it gave me when I was a lad, and to the author of this article. Love these long form historical retrospectives. I even learned a new thing: the communicator. I had no idea they ever made that!

[David Harper 1]

I learned to program on a Z80-based machine

The Research Machines 380Z, to be precise. It was the first computer that my school bought, back in 1980. I was allowed to use it in my free time, and I taught myself BASIC programming on it. The school later acquired a Commodore PET, which was easier to use, but I'll always have fond memories of the RM 380Z. It was the starting point for a 40+ year career in scientific computing.

[Rich 11]

Re: I learned to program on a Z80-based machine

I remember the RM 380Z from my college years in the early 80s. A chunky black box sitting in a rack, with sharp-edged aluminium handles. I'd learned a bit of Assembler out of a ratty book and some hobby magazines on the ZX-80, 81 and Spectrum, so it was an easy move. I can't for the life of me remember anything about any software I wrote for it now. I might even be mis-remembering the black box!

Do we have an icon for 'fuck, I'm old!'?

[John Brown (no body)]

Re: I learned to program on a Z80-based machine

Seems to be the best available --------------->

[NXM]

Re: I learned to program on a Z80-based machine

Me too- I was one of the first to do computer studies at my school on one of these.

The first thing me and my mates did was write a bogus command line, which sometimes passed the user's input to the system, and other times would come back with 'are you sure you want to do that' messages in increasingly frustrating ways. We left it running for victims. If they'd found out who did that we would've been expelled.

Then we did a space invaders program. In basic. On a really slow machine. But it worked!

Happy days.

[Red Ted]

An end of an era, but 48 years is still very impressive

A few years ago my daughter and I built an RC2014 and she the programmed it with help of some of my 1980s reference books…

Whilst it is a pity for the firms that depend on them, it is still an amazing testament to the success of the design that it lasted so long.

[J.G.Harston]

I was doing work experience in 1985 building a super-fast ADC interface and writing the software to read from it, and somebody gave me a copy of the new HD64180 datasheet. Though he also insisted that there was "no point" in having any interest or competency in the programming that I'd been doing for several years at that point, as "it will all be done by compilers and stuff for you". I kept trying to argue, well *somebody* has to write the low-level code, the code that starts from reset, the code that the bigger code calls to access the stuff in the machine itself (what I later would learn to call drivers), *that's* what I'm good at and enjoy. "No, that will be done by compilers". Well, *somebody* still has to write what the compilers spit out. "No, that will be done by compilers". I seem to have spent 40 years with people pestering and nagging me to *STOP* doing stuff I have aptitude, skill, and enjoyment in.

[Bebu]

"it will all be done by compilers and stuff for you"

"Turtles all the way down."

Yes. But you have to breed those turtles first."

The world is plagued with the value of everything, worth of nothing crowd.

Knowing a bit of assembly and instruction set architecture helps when writing resource (time/memory) constrained code in a high level language. Doubly so when debugging application bugs which are compiler problems. I recall the Sun C compiler on x86 SunOS ('94 around pentium FDIV bug) produced incorrect code when compiling htonl() which could only be determined by inspecting the assembly (sparc and x86 have different endianess.) You could see it by attaching a debugger to the running application but the developer had spent days scouring his own code.

I learnt machine code/assembly as a sthdent first on pdp10, then pdp11 and 6809 on my first pc, the z80 on an amstrad and x86 on a PC. A smattering of mips, sparc, parisc, rs6k powerpc over the years. All pretty much move, modify and test with the true and proprietary complexity and cleverness of decoding, dispatching, scheduling etc inside the CPU itself.

[martinusher]

I still have fond memories of PL/Z

Zilog sold a box that included a very fine for the year programming language, PL/Z. It was block structured in the Algol/C/Pascal tradition but had the interesting property that you could return more than one result from a procedure. Its other great feature is that you could choose to either compile to a machine code object or to a semi-interpreted byte code, the interpreter being linked in as another object module. This allowed you to trade off speed and space. "Not bad for 1980" (and people have been reinventing and also, for all I know, patenting it on and off ever since!).

As for my original CP/M machine, it was an Osborne 1. I had all sorts of things for it, my two favorites being Forth and Prolog. I left it in the UK when I moved to the US as I then had PCs; somewhat better hardware but it literally took years for the available software to catch up.

[CowHorseFrog]

Not really sure how you could say the Z80 was an improved 8080, it was still horrible ISA, they hardly tried to make something better.

[IvyKing]

Don't forget Technical Design Labs

TDL were the folks that wrote much of the software (e.g. Zapple monitor) to make a Z-80 run like a Z-80 as opposed to be an 8080 clone that just needed 5V. TDL went the way that most of the early micro pioneers went, going under ca 1979. Many of the people from TDL later went on to write a successful IBM PC BIOS clone.

I wonder what the micro-computer industry would be like if Zilog had announced the Z-800 in early 1979 with sampling later that year?

[Dr Fidget]

Nascom 2

God, I remember with such fondness the Nascom 2 which came as a circuit board and a bag full of chips and a bare Hall-effect keyboard. It had a 2K (yes that's 2K) OS on a ROM complete with the assembler source code.

I was a freelance COBOL programmer when my roomie brought one home and I sat down to scoff - an hour later I was gob-smacked. At the time at RX I was waiting 24 hours for a test compile of a COBOL program just to find I'd missed a full stop at the end of a line somewhere and here was a machine I could probably write a COBOL syntax checker on which would run immediately.

[Doctor Syntax]

Re: Nascom 2

The move from batch on cards to direct access on anything more than a crystal set was a game changer for anyone working in those days, even if the batch turn round was just(!) a few hours.

[tychosoft]

Speaking of Ragnarök, the whole premise of CatastropheOS was that there would always at least be plenty of old and salavagble Z80's waiting around to reboot civilization the day after. I think they supported 6502's as well, but they bet big on at least Z80's surviving. Personally, I favor the land of the lost 6809, and what might-have-been in an OS/9 derived world.

[Dan 55]

If you mean Collapse OS (and Dusk OS), I just can't see it myself. As well choosing 8-bit CPUs they chose Forth as the language. I would say there aren't many people who still know what that is and Fuzix has more chance of being useful.

[Liam Proven]

[Author here]

Point 100% taken about Fuzix and C, but I still regard C as tantamount to a crime against humanity.

I wish it were Oberon.

And Oberon can be bootstrapped and loaded on top of DOS.

Now we have FOSS DOS and FOSS Oberon, maybe that will be the way. Almost all the other DOS multitasking environments and programming languages are closed-source and thus black boxes: TurboPascal, indeed TurboEverything, JPS Topspeed everything, Lattice-everything,. Zortech everything...

[Dan 55]

As Churchill might have said in another life and career, many languages have been tried, and will be tried in this IT world of sin and woe. No one pretends that C is perfect or memory safe. C is the worst language, apart from all the others that have been tried from time to time.

[Doctor Syntax]

Great article, the diversion into Z8000 history is much appreciated. One system builder was Onyx with the imaginatively name Onix OS (Unix 7th ed, of course ported by the company who also did the original port of AIX)

Zilog built whole systems. I suppose this must have been the result of Exxon's ambitions. I went on to work in a shop that had two different models. One of them had a disk drive just under the top cover that gave it a rotating external magnetic field so that text on the monitor of the terminal placed on top of it would wobble all the time.

[sjmudd]

like others I started with the z80 on a Research machines 380Z with 5 inch floppies. Then played with the zx80 and moved to a BBC micro model B which used the 6502. great learning experience. shame to hear the chip is going but made it’s time.

[Dan 55]

The king is dead, long live the king

Seems there's going to be FOSS clone of the Z80.

Perhaps they've jumped the gun because the first batch is going to be manufactured in June, probably at the same time as the last orders for the real Z80.

[SCP]

Aermacchi AMX International

I believe the Z8000 was used in the Flight Control System of the Aermacchi AMX aircraft. It was an interesting device - if memory serves correct its implementation was effectively a 2 level NAND gate which gave the instruction set a very deterministic performance.

(The Z80/Nascom 2 heritage was also quite nostalgic)

[Neil Barnes]

Later, in Nick Stacey's computer lab

Captain Zilog suddenly realised that with a right hand on the end of his right arm, instead of a left, things might have been so different...