The Legendary Zilog Z80 CPU Is Being Discontinued After Nearly 50 Years

Long-time Slashdot reader AmiMoJo shares an article from TechSpot: Zilog is retiring the Z80 after 48 years on the market. Originally developed as a project stemming from the Intel 8080, it eventually rose to become one of the most popular and widely used 8-bit CPUs in both gaming and general computing devices.

The iconic IC device, developed by Federico Faggin, will soon be phased out, and interested parties only have a few months left to place their orders before Zilog's manufacturing partner ends support for the technology... Federico Faggin, an Intel engineer, founded Zilog in 1974 after his work on the Intel 4004, the first 4-bit CPU. The Zilog Z80 was then released in July 1976, conceived as a software-compatible 'extension' and enhancement of the Intel 8080 processor.
Back in 1999 Slashdot was calling Zilog's updated eZ80 "one of the fastest 8-bit CPUs available today, executing code 4 times faster than a standard Z80 operating at the same clock speed."

Another headline, from 2001: Zilog To File For Chapter 11...

Question

[quonset]

In the modern age, what purpose would this chip serve? 8 bits is insignificant unless you're sending SOS.

Re:

[backslashdot]

Hence why they're discontinuing it. My guess is it was being used in appliances or devices that were designed in the 1980s that were "good enough" and didn't need a redesign.

Re:

[Vrallis]

For example a Triad Series 12 point of sale / inventory management system. Used to be tens of thousands of them in auto parts stores everywhere. There are probably still a few hanging on out there. I spent 1999 trying to fix all our communications with them to get Y2K workarounds in place, and a number of years migrating data off of them to modern systems.

FPGA V

[Roger W Moore]

I suspect that for anything that still needs it you can use an FPGA implementation of the Z80.

Re:Question

[arglebargle_xiv]

This is the original 1970s-vintage Z80 and more recent derivatives like the Z180, it looks like the eZ80 is still active, so it's not the entire Z80 family that's going away, just the more-awkward-to-work-with older members.

These chips are not going anywhere

[Excelcia]

Zilog hasn't been the only manufacturer of Z80s and clones for a long time. There are a lot of smaller outfits that still produce it. Simply for the reason that there are so many embedded applications that have one. Traffic lights, oil well-head pumps, older CNC (and other industrial automation cases), embedded pipeline valve controllers... the list is a long long one. They are rock solid reliable, deterministic, and even un-hardened ones are still naturally immune to levels of radiation (ionizing and no

Re:

[NotAMarshallow]

True. These chips are extremely cheap with all of the manufacturing advancements.

Re:Question

[Opportunist]

It's probably dirt cheap and for a lot of applications "good enough".

Re:Question

[Casandro]

Actually no, the chip alone is somewhat more expensive than a much faster microcontroller with >64k of RAM... and you still need support chips. So it's simply to expensive to be used in an embedded system, sadly.

Re:

[Anonymous Coward]

It's probably dirt cheap and for a lot of applications "good enough".

This is a common misconception with mature processors, they aren't all that cheap. You can get a 32-bit microcontroller that is way faster, uses way less power, has more peripherals and features, has integrated ram and flash, etc. for perhaps 1/4 the cost of a Z80.
Same as when the pic16f628 came out, it was a pin-compatible dropin replacement for the 16f84, and it was better in every way, and it cost less.
32-bit microcontrollers are so cheap now that they handily beat 8-bit controllers in value-per-dollar.

Graph of lifetime prices may be 'U' shaped

[drnb]

This is a common misconception with mature processors, they aren't all that cheap.

A lifetime graph of a device's price is not necessarily a descending asymptotic line. Its often 'U' shaped, after some optimal point the price starts going up. Declining use leads to a loss of efficiency of scale..

Re:Question

[edwdig]

Yeah, the really old processors aren't the cheapest option because of the unit cost. They're cost effective because changing your production line is expensive. You need engineers to update the product, testing, updating the production lines, new supply contracts, etc. You need to sell a lot of units to make back those costs.

It's especially a big deal with heavily regulated productions. Things like cars and medical devices will often rely on very old components because changing them would require a lot of expensive testing to ensure they meet the legal standards.

Re:Question

[Cafe Alpha]

We've gone from a 6502 with 3000 transistors in it to an NVidia half tensor unit with 100 billion transistors (double for the whole tensor unit).
I like to imagine that one say someone will make a computer with a different computing model where every memory cell will have it's own processor.
So maybe something like that?

Re:Question

[ffkom]

The idea of building processors where memory and arithmetic/logic units are not separate from each other is decades old and has been implemented now and then. In the form of FPGAs you can essentially configure such processors on the fly. But it seems the flexibility of general purpose CPUs remains an important feature, even when more and more specialized units accompany generic CPUs cores on the same chips.

Re:

[kriston]

It's important to note that those nVidia processors are simple vector units and not full-featured microprocessors like the Z80, x86, MC68K, etc.

Re: Question

[OrangeTide]

Z80 is turing complete. And 64k bytes is enough to count all the atoms in the visible universe.

Re:

[swillden]

64k bytes is enough to count all the atoms in the visible universe.

With 65,502 bytes left over.

Re:

[OrangeTide]

that leaves for a lot of multiverses.

Re:Question

[williamyf]

In the modern age, what purpose would this chip serve? 8 bits is insignificant unless you're sending SOS.

A good example is controlling trafic lights. A Z80 is good enough, even for traffic lights controlled from a central office (via those small yaggi antenae, or some other method). And, in serious countries, they have to be subjected to technical review and approved by the transit authority. The cost of the approval is not insignificant. The installed base is HUUUUUGE, so, you keep pumping Z80 based units, instead of making a new controller every time a new micro(processor/controller) comes out.

At some point, either your company or a competitor will come with a unit using another Microcontroller (in this day and age, a Z80 is hardly a microprocessor), but then you have to retrain the service personel, go through all the certification again. So this is seldomly done, and even when done, there is the installed base thing.

Most likely, Zilog saw diminishing orders for these, meaning users were either moving to more modern solutions, or finding the New-Old-Stock, recycled, clones and/or counterfeit ones good enough for their needs.

Re:

[thegarbz]

A Z80 is good enough

It's not good enough. It's an obsolete part that was single sourced single component without a long term viable supply. That makes it unsuitable for use in infrastructure today. I may have been good enough 20 years ago, but the part would have been on the verge of being considered NRND back then (Not Recommended for New Design). In the world of traffic lights speed isn't the issue, security of supply and programmability is.

Re:

[narcc]

I may have been good enough 20 years ago, but the part would have been on the verge of being considered NRND back then

In 2004? The z80 was still going strong. The ez80 was just a few years old, after all. That product line is not being discontinued, only the z84C00 line [pdf] [mouser.com].

So, no, it would absolutely not have been considered NRND back then. Even today, as a platform, the z80 is as stable as ever.

It's an obsolete part [...] security of supply and programmability

Old does not mean "obsolete". It's been around for ages and a lot of people are familiar with it. Finding software and developers isn't an issue. There's a lot of value to be had from the kind of stability the z80 offered.

Re:

[stabiesoft]

I'm thinking many here don't really understand "part life status". Manufacturers disclose it and suppliers like Digi-key even have selection criteria for it. I just picked up a hundred 5534A's as they are going end of life. Digi is nice enough to send an email if you've ordered a part before and it is going EOL. They even tell you when the "last time buy" is. There is also a company called Rochester that buys up parts at EOL and resells them. They must have one whopper of a warehouse.

Re:

[BeaverCleaver]

There is also a company called Rochester that buys up parts at EOL and resells them. They must have one whopper of a warehouse.

Thanks for the tip. I'd seen Rochester listed as a supplier on Digikey's website, now it makes sense if they are an aggregator of EOL hardware.

Re:

[stabiesoft]

For enough money, I think they will do a clone.

Re:

[kackle]

There are lots of applications for a low-resource processor or MCU if you look closely. A basic microwave oven doesn't need more, for example. Nor do most appliances, nor furnaces (which might not even have one at all). We make pump controllers/accessories, and most of our stuff uses cheaper/smaller 8-bit MCUs. We only added 32-bit versions within the last decade or so when "cloudy" was (ill advised?) added to some of our products.

Re:

[guruevi]

They're still used in appliances, although if they ever redesigned the board, they are using the eZ80 which is binary-compatible but not pin-compatible with the Z80.

There are plenty of places you find ancient processors like the 8088, PIC, the Z80 and Motorola 68000. They all can have mini-webservers and other relatively modern plugins, they all have sufficient power for everything from thermostats to washing machines and even vending machines. If it ain't broke, don't fix it applies tremendously to the app

Re:

[ffkom]

8 bits is insignificant unless you're sending SOS.

You do realize that 4 bit wide numbers are the latest hype in computer technology, so much so that the popular press is citing performance metrics based on 4 bit number computations per second [forbes.com]?

Re: Question

[wed128]

4 bit computation is only really useful for ML, where a guess is usually good enough. For any traditional computing task where some kind of repeatability or determinism is required, generally 4 bits isn't quite enough

Re:

[ceoyoyo]

You can do lots of computing with one bit. Logic, for example. It's arithmetic where YMMV.

All of which is more or less irrelevant anyway. The Z80 has mostly 16-bit registers, a 4-bit ALU, and can do whatever precision math you want, so long as it fits in 64 KB. Unless you want to get creative.

Re:

[UnknownSoldier]

Funny enough a STB even last decade were using the 6502 in the ACP chip!

This is a great talk: How Do I Crack Satellite and Cable Pay TV? (33c3) [youtu.be]

Re:

[Z00L00K]

There are still clones around as well as more modern evolved variants, so I wouldn't say it's dead.

Some of those evolved variants are embedded solutions, even variants with built-in ROMs.

Re:Question

[misnohmer]

Your comment is a perfect illustration of the state of software today. Z80 based microcomputers, such as the Sinclair ZX Spectrum, had built in BASIC interpreter, supported compiled and interpreted languages, had a slew of very popular computer games available for it them (with color graphics too), as well as productivity tools such as word processors, spreadsheets, etc. All those ran on computers with less than 64KB of ROM and RAM. NASA managed to land on the moon with its navigation computer having 74KB ROM and 4KB RAM. The towers of abstraction that today's programmers require to write a code require MB's or RAM to just print out "Hello World!" on a console.

Re: Question

[Fons_de_spons]

There are lots of (new) 8 bit CPUs out there. Any electronics device that has to do a bit of math has enough with 8bits at a few MHz. You can still 64 bit arithmetic on those if desired. It will just take a few more clock cycles. For a lot of electronics, math at a few MHz is more than enough.
Most electronics I design for hobby runs on a cheap 8bit AVR attiny mcu. (1.3 euro a piece last time I checked) Usually it runs at a few KHz, uses only a few bytes of ram, less than 2kb of program memory. and can run

Re:

[stabiesoft]

Are they really 1.3EU? That seems expensive. A pi-pico (chip only) is USD 0.80, so less for a dual core arm uC, albeit no builtin flash with the RP2040. Now the other advantage for the older parts is they come thru hole, which is great for hobby, but more expensive than their smaller surface mount equivalents. I can do SOIC, but I've not tried BGA. I may try the toaster oven trick on something cheap to see if I can give it a go. Custom PCB prices are really not a barrier anymore. And some houses will even d

Re: Question

[Fons_de_spons]

Yes, at least where I buy them. This is for Dip, soic are cheaper. I grew up with Atmel microcontrollers a decade or two ago. I love the simplicity of them. Suits my needs so I have little incentive to learn something new. But not sure I'd use them if I had to start over,

Javascript and a hierarchy of 40 libs ain't cheap

[drnb]

God, you're an idiot. You think your coffeemaker or microwave need more than 8-bits?

Sure, that Javascript and a hierarchy of 40 libraries ain't cheap. And that's before internet connectivity is added.

Was great

[backslashdot]

But now there's better alternatives for both cost and lower power consumption. STM32 for example, the cheapest of which goes for less than 25 cents?

Re:

[Opportunist]

Certainly. But it's not a drop-in replacement. You can't just replace a Z80 with a STM32 and be done.

There's nontrivial costs associated with redesigning hardware, along with writing a new firmware. I highly doubt that anyone who designs anything new will reach for a Z80. Or has for the last 20 years.

But believe it or not, there are implementations that have gone unchanged for 30 and more years simply because what the hardware needs to do didn't change. So why design a new hardware?

Re:

[serviscope_minor]

There's a virtue in free they're are a crapton of 8051 based things because it's fast enough, had all patents expired and the tool chains are already out there too.

It's not that people think oh I need an MCU, I'll use an 8051, it's that someone designing a chip uses the core for cheap and that serves as the MCU for whatever is built around it.

Re:

[cstacy]

it's that someone designing a chip uses the core for cheap and that serves as the MCU for whatever is built around it.

MCU?
Good for powering Iron Man suits.
You may want to add Flash memory, too.
Benchmark smash!!
And a Maximoff chip for ML (4-bit WANDA logic).

Give a Z80 to Coulson and he'll use it as a weapon...

Re:

[tepples]

You may want to add Flash memory, too.

Flash would have to be DC powered.

Re:

[AmiMoJo]

But also the STM32 line are microcontrollers and SoCs, not microprocessors. Some do have an external memory bus, but they aren't really suited to building computers with, or dropping in as a replacement for a Z80.

Oh, and STM support is total crap. Their HAL library is dire and their peripherals are awkward at best. Contrast them with Atmel^W Microchip SAM, or AVR.

Re:

[Opportunist]

STMicro support isn't the best. True. Still heaps above Espressif who'll just suddenly and for no valid reason simply pull the rug out under you and disable whole portions of their firmware because it would allow you to actually do with the hardware what you want to do instead of what they deem appropriate.

Re:

[Randseed]

But now there's better alternatives for both cost and lower power consumption. STM32 for example, the cheapest of which goes for less than 25 cents?

According to Mouser [mouser.com] a Z80 costs around $9.56 each. I'm sure there are cheaper equivalent processors out there. Actually, that's a Z84 but I think it's compatible.

Re:

[cstacy]

STM32 for example, the cheapest of which goes for less than 25 cents?

According to Mouser [mouser.com] a Z80 costs around $9.56 each.

Companies that design the kind of embedded systems that use chips like a Z80 consider amounts like two cents (00.02$US) to be a lot of money.

15 years ago things like credit card auth terminals were based on Z80, with banked memory, running tightly coded assembler on bespoke operating systems. Since then, those kinds of little terminals are written in Java running on Linux.

Z80s are no longer economical -- it's more like the situation wit Air Traffic Control systems that were still using vacuum tube computer

Re:

[AmiMoJo]

There were several manufacturers producing Z80 and Z80-based variants back in the day. NEC made some, as I think Hitachi did back in the 80s. As the market dwindled they didn't keep developing them.

It was a very different time, when often the licenced products were as good or better than the originals. NEC were well known for that, making better 8086 chips than Intel, in the form of the V30. Intel actually sued them for breach of licence terms, but lost and set a precedent that microcode can't be copyrighte

This is the saddest news I ever read on /.

[Koen Lefever]

My first computer was a Sharp MZ-80K [wikipedia.org] in 1980.

I was doubting if I should get a Cerberus [github.com] or an Agon [github.com], I will now quickly get both. And an RC2014 [rc2014.co.uk] also while I'm at it.

Re:

[morcego]

My first computer was a Sinclair ZX81 clone, running on a Zilog Z80A processor, with 16KiB of RAM (and 8KiB of ROM).

Re:

[fullgandoo]

My ZX81 had 1KB. Out of which just over 600 bytes was actually usable. I remember typing in a chess program in machine code from a magazine. Still amazes me that someone could fit all the chess rules and a rudimentary engine in 600 bytes.

Re: This is the saddest news I ever read on /.

[morcego]

Zx81 has 1KB. The clones got more memory. TK82C has 2K. The one I had was 16KB. There was even some with 32KB.
Interesting days

Re:

[AmiMoJo]

There will still be new Z80s available, just not in the DIP package. I'm sure adapter boards with SMD Z80s to fit a DIP socket will appear. Not quite the same but functionally equivalent. That said, it does make me wonder if even those have a limited lifespan now.

TI

[Dwedit]

Oh no! How will Texas Instruments continue to manufacture their Graphing Calculators?

Re:TI

[Samantha Wright]

The answer is that this is somewhat misleading; there are plenty of Z80 compatible chips still being manufactured. The TI-84 uses the eZ80, which is four times faster and will still be around [zilog.com]. There are also many (many) clones that are closer to cycle-exact.

Re:

[Casandro]

They are apparently only retiring their Z84 series CMOS systems, I doublt Texas Instruments still uses those, instead of the advanced Z18x ones from Zilog which are binary compatible.

Obligatory

[93 Escort Wagon]

Okay, what the hell, TI? [xkcd.com]

Re: Obligatory

[beelsebob]

Iâ(TM)m pretty sure at this point being extremely old tech is the point. If they stuck sure modern microprocessor in there that could do actual real computation in a meaningful way itâ(TM)d be hard to get exam boards to accept students having them during exams.

That said, the price is ludicrous and clearly only based on schools requiring exactly that model.

Re:

[Dwedit]

You can run a TI83 or TI89 emulator on a phone, provided you have a way to obtain a ROM. TI has given out the firmware update files for free, don't know if they still do that or not.

Oldest?

[kackle]

Is it the oldest in production then?

Re:

[Koen Lefever]

The COSMAC 1802 [renesas.com] is in production since 1974.

Re:

[FrankSchwab]

Oh my lord -
My first engineering job involved hardware using the 1802, in around 1985, and that POS was completely and totally obsolete even then. This was, of course, a defense program, so the prototypes had been built in the mid-1970's and nothing had changed since then.

Re:

[hawk]

it was slow, but could be extremely low power compared to the others, and was silly-rich with registers. 16 general purpose 16 bit registers, iirc. (or pairs of eight bit). And ISTR that you could use all but one or two for program counter and reference (a pair of four bit registers [P & X ?] that pointed to which 16 bit to use]

Also, significantly more radiation resistant than the others of the time (or was that another version? Even so, its design should have been more resistant).

My first computer

Re:

[dryeo]

Wouldn't be surprised if 6502's are still manufactured. Seems that the 65C02 is still manufactured and similar chips with the original NMOS 6502 also available, though it is hard to find if it is still manufactured. The 6502 is a year older then the Z80 and generally the 65C02 is close to a drop in replacement and as long as you weren't depending on unofficial opcodes or bugs, works fine.
Interesting article, https://hackaday.com/2022/12/0... [hackaday.com]

Re:

[hawk]

cool link! (at least if you disregard things I remember being cast as ancient history!)

the 8080 had at least one or two undocumented instructions that worked their way into code. IIRC, the Heathkit chess program needed a byte changed from that to a documented instruction on the Z80 to run [a one byte patch!].

And there were a pair (?) of quirks where 8085 instructions took a cycle ore or less than the the same instruction on the 8080.

The Z80 executed some instructions in less cycles than the 8080 (but wasn

Hot Rod Z80

[Chelloveck]

Ah, my first job out of college. Circa 1989, I got hired at Motorola to code Z80 on their EMX series of cellular telephone switches. But this was no ordinary Z80! This was a Z80 with blackjack! And hookers!

The board had an external MMU and a bank register gave it an effective 24-bit (16 MB) address space. There was an active processor and a standby processor, and 4 MB of the address space was shared between the active and standby. The MMU even had an NMI mask, so we could mask out non-maskable interrupts. I was super proud of myself that I taught our HP logic analyzer to understand the bank register and actually decode the full 24-bit address bus.

So many fond memories. Like counting clock cycles to make sure a task would run in the time allotted. Or the time a co-worker thought the I register was just another general purpose register. (It was actually the interrupt vector. Hilarity ensued.) Or the fact that we only had half a dozen machines in the lab, so lab time was scheduled 24 hours a day and you grabbed a slot whenever you could. Or the two weeks I spent all night, every night, sitting on the floor of a customer site in a refrigerated Bangkok switch room with a microfiche reader and stack of fiche about the size of a brick, typing opcodes into the debugging terminal on the running production system.

Wait, did I say fond memories? I meant nightmares.

Re:

[Z00L00K]

I'd call that an experience building job. The amount of things you'd learn from such a job is something that is worth a lot in future projects, especially things that shouldn't be done.

Re:

[AmiMoJo]

Interesting that Motorola would use a Z80, given their had their own line of 8, 16, and 32 bit CPUs. I suppose by 1989 the 6800 line was mostly dead as far as Motorola was concerned, with only some licenced versions and single customer variants in production.

Re:

[msk]

6809 enters the chat.

6309 makes popcorn.

Re:

[Chelloveck]

I figured someone would notice this. The system was designed around 1980, well before I started there. I'm not sure why the Z80 was chosen over the 6809. The explanation I was given was that the Z80 was more capable than the 6809. The Z80 was also less expensive, though given how much we sold the boards for I don't think a $20-ish difference in CPU price would have made a difference. It could have just been that the people designing it were more familiar with the Z80 and could design faster for it.

Re:

[hawk]

My guess would be that 24 bit address space for the MMU, and that this worked better with the Z80.

There were ways to extend the 6809 space by a couple of bits, but not by eight.

The greater abundance of registers on the Z80--including an entire second set of the 8080 registers, which could be toggled between--sounds like a likely reason. IIRC, the 6809 didn't have any extra data registers as compared to the 6800.

hawk

Makes me wonder

[dicobalt]

What kind of clocks could a Z80 get if it was on a 5nm process.

Re:

[Dwedit]

The speed limiting factor becomes your system bus. To go faster, you need a cache. And to have a cache, you need a way to describe the memory map, so it knows which addresses are plain RAM, plain ROM, bankswitchable memory, memory-mapped IO, memory shared with other hardware, etc...

What happens to the Z800, Z8000, and Z80000?

[kriston]

What happens to the Z800, Z8000, and Z80000 successors?

Or had they already been discontinued?

Re:

[Z00L00K]

They were different families than the Z80 and all of them are basically dead end solutions. There are Wikipedia articles on them. As well as the Z8.

You may still find special embedded solutions using at least some of those architectures, but I'd expect them to be replaced with ARM and Atmel AVR.

So the 6502 wins!

[tlhIngan]

The 6502 is still available today brand new along with several of its companion chips as well as the 16-bit version.

Though, honestly, I don't see why those chips haven't been long discontinued - the Z80 and 6602 basically should be soft cores these days and implementable in basically a reasonably cheap FPGA able to run at full speed. I know Western Design Center (the current provider of 6502 chips) already sells such a design so you can implement a 6502 core in your product and achieve reasonable integration density.

I don't see many people implementing the 6502 or Z80 as a standalone chip other than retro enthusiasts. If you need a Z80 core, you probably will just have one for your FPGA already or build it into an ASIC.Though, oddly, the 8051 has shown incredible robustness for an embedded controller - they're seemingly everywhere especially in USB applications

Re:

[AmiMoJo]

A surprising amount of stuff still uses them, often running ancient software that is well tested and which would be expensive to replace.

Same issue applies to FPGAs, dropping one in would require re-qualifying many products at considerable expense. Cheaper to just buy a real Z80/6502.

Interesting for me.

[NotAMarshallow]

While the Z80 and it's variants have been a memory for me, I was more surprised that EDI (Electronic Data Interchange) of which I was an early adopter of is still being used today. Let me say this, if you read Wikipedia about EDI, you'll be totally lost and mislead about EDI in the modern age. In the early 90's EDI had moved to the WWW. ANSI standards had been approved and every large and mid-sized company was being forced to send information electronically,. Any paper document could be sent and translate

Re:

[NotAMarshallow]

Correction to the above. It has been nearly three decades since doing this and I made an error in my example. The "+" and ":" were actually field delimiter and sub-field delimiter.

Oddly Gaining Popularity with Retro-Computing

[BrendaEM]

Oddly, it seems like the Z80 variants started enjoying some popularity with projects such as the Agon. It would be interesting if someone made an 8-bit processor, made at a modern fab, in other words: an 4gHz 8-bit CPU.