TR Forums

Back to the original question.

The motorola 68008 was essentially a 68000 with an 8-bit data bus. So you could call it a 8-bit computer although you would probably be better to call it a 8/16/32 hybrid and it came with either a 20 or 22 bit address bus. It did see one major usage in a small computer, the not very successful sinclair QL: https://en.wikipedia.org/wiki/Sinclair_QL

thx1138r wrote:

Back to the original question.

The motorola 68008 was essentially a 68000 with an 8-bit data bus. So you could call it a 8-bit computer although you would probably be better to call it a 8/16/32 hybrid and it came with either a 20 or 22 bit address bus. It did see one major usage in a small computer, the not very successful sinclair QL: https://en.wikipedia.org/wiki/Sinclair_QL

I wonder if it had any better success elsewhere such as the embedded market it was intend for?

whm1974 wrote:

I'm still wondering what could be done with a 8-bit CPU w/ wide address space and large number of registers designed around a cleaned up and modernized ISA. And it would be a multi-core design as well.

It doesn't make sense.

To use a large register set effectively, you need to be able to encode the numbers of the registers to be operated on in the instructions. With (say) 32 registers, you've already used up 5 bits out of your 8-bit instruction, just to encode the source register. And that leaves you only 3 bits to encode the rest of the instruction opcode, so this hypothetical CPU would have only 8 unique instructions!

Since you don't have enough bits left to encode a destination register, the target of all operations is an implied accumulator register (as with most real 8-but designs). But this reduces the utility of having the large register set, since everything must funnel through the accumulator anyway.

Sure, you could deal with this by adding prefix/suffix bytes to certain instructions (i.e. variable length instructions like x86 uses), but at the end of the day this is a lot of additional complexity for questionable gain. You'd basically be creating something like a crippled x86 with 8-bit internal data paths. So much for the "cleaned up and modernized" ISA!

And you're still ignoring the fact that with a large flat address space, it is going to be horribly inefficient to manipulate memory addresses if your internal logic is all 8-bit. Address calculations for simple array indexing are potentially going to take many (possibly dozens) of clock cycles, unless you impose arbitrary restrictions like requiring all array elements to be a power of 2 in length.

Sure, you could make a CPU that can address lots of memory in spite of being 8-bit internally. But it would be inefficient and slow compared to contemporary 32- and 64-bit processors, and more complex/expensive than the 8-bit designs still in use.

Other than as an academic exercise, what's the point?

OK I'll admit that I'm defeated. Sounds like for very good reasons everyone went to 32-bit and 64-bit designs. Because they had to. Oh well at least I learned what not to do.

Aranarth wrote:

The problem with this sort of question is there are several ways to look at it:

Performance: Processing power, memory access speed, bus speed etc.
Library: how much software was written for that processor
Market share: how many machines actually sold

Considering how many machines were created and how much software was available and then future impact I'd say the most powerful was the 8086 and it's weird brother the 8088.

In raw performance I think there is a z80 compatible chip that runs at 20mhz that was an absolute beast in comparison to other 8bit machines but it was produced in the mid 90's.

The 8086 could also be overclocked to 20mhz making it a really fast machine but that required some modification of the system board by replacing crystals and putting heat sinks on various components and this was also a 16bit chip not an 8bit one.

I believe the commodore 128 could also be overclocked to 4mhz? which might have made it the fastest real 80's 8bit computer.

My 8086 based Compaq DeskPro (1983) ran at 7 or 8 MHz. It was like having your own IBM360 once you added the 8087 math coprocessor and the MS FORTRAN compiler. I upgraded the CPU to the Japanese clone, an NEC V30 that ran at 10MHz. I had 640K ram and 2MB of extended memory in each of two Tall Tree Systems JRam cards. So, I had 4MB of extended memory and 704K of DOS memory. Both DoubleDos and DesqView could use this memory for task switching. At the time, that was flying!

Edit: oops the V30 ran at 10MHz.

whm1974 wrote:

I wonder if it had any better success elsewhere such as the embedded market it was intend for?

It had some at least, I remember when the sinclair QL came out, a few of it's detractors pointed out that the 68008 was used in some of the more upmarket washing machines of the day.

thx1138r wrote:

It had some at least, I remember when the sinclair QL came out, a few of it's detractors pointed out that the 68008 was used in some of the more upmarket washing machines of the day.

Thanks, I just went looking for info on modern washing machines, having been reading for 5 years or more about smart ones so suspecting they probably use 3+ core CPUs, a good chunk of RAM and a Linux distro nowadays, and found all sorts of silliness and rise of the machines fun, but no actual hard details (yet). Anyone got any good info they can link?

The video at the end of that second link is great (IMO), how amazing that the touchscreen computer built into the fridge door can even show you what's inside!!!

Unlikely to be Linux, more likely to be a more embedded OS in your smart appliances. It's probably a 32 bit ARM or an x86 processor, maybe only one or 2 cores and quite low clock speed in comparison to what you are used to.

Topinio wrote:

thx1138r wrote:

It had some at least, I remember when the sinclair QL came out, a few of it's detractors pointed out that the 68008 was used in some of the more upmarket washing machines of the day.

Thanks, I just went looking for info on modern washing machines, having been reading for 5 years or more about smart ones so suspecting they probably use 3+ core CPUs, a good chunk of RAM and a Linux distro nowadays, and found all sorts of silliness and rise of the machines fun, but no actual hard details (yet). Anyone got any good info they can link?

The video at the end of that second link is great (IMO), how amazing that the touchscreen computer built into the fridge door can even show you what's inside!!!

What is so hard about simply opening the fridge door to check what you have? I think putting computers in everything, never mind connecting them to internet is getting to be way overblown.

just brew it! wrote:

Zilog also produced a 16-bit follow-on to the Z80, the Z8000. It never caught on.

I vaguely remember being told that the Z8000 never caught on because it was very late to the party. It was very late because the decode logic was apparently actual hand designed logic and took a huge amount of time to debug. If they'd used a ROM type system maybe things might have gone differently for them.

Anyone know if I've remembered that correctly?

I'm pretty sure the Z80 or MOS 6502 were the undisputed champs of 8-bit CPUs. The 6502 was clocked slower, but it had around 4x the IPC of the Z80, so they ended up pretty close in performance.

Takeshi7 wrote:

I'm pretty sure the Z80 or MOS 6502 were the undisputed champs of 8-bit CPUs. The 6502 was clocked slower, but it had around 4x the IPC of the Z80, so they ended up pretty close in performance.

Too bad the WDC 65816 wasn't more successful then it was. I still think that both Atari and Commodore should have released systems around the CPU to provide an upgrade path for users of their 8-bit lines. Yes I know the Apple IIgs was one of Apples best selling computers for awhile, but still.

whm1974 wrote:

Takeshi7 wrote:

I'm pretty sure the Z80 or MOS 6502 were the undisputed champs of 8-bit CPUs. The 6502 was clocked slower, but it had around 4x the IPC of the Z80, so they ended up pretty close in performance.

Too bad the WDC 65816 wasn't more successful then it was. I still think that both Atari and Commodore should have released systems around the CPU to provide an upgrade path for users of their 8-bit lines. Yes I know the Apple IIgs was one of Apples best selling computers for awhile, but still.

One other 6502 descendant was much more popular. The Ricoh 5A22 was the heart of the Super Nintendo, for example, and that sold WAY more units than any 8-bit computer.

Non-x86 16-bit computers, for whatever reason, just gravitated towards the Motorola 68000. The Mac and the Amiga are the two most popular examples but Atari's ST used a 68000, too. The 68000 was also at the heart of arcade boards from Sega, Capcom, Konami, SNK (the Neo Geo) and plenty of others. The most famous arcade games of the late 80s and 1990s used some combination of a 68000, a Z80, and some special video and sound chips.

derFunkenstein wrote:

whm1974 wrote:

Takeshi7 wrote:

I'm pretty sure the Z80 or MOS 6502 were the undisputed champs of 8-bit CPUs. The 6502 was clocked slower, but it had around 4x the IPC of the Z80, so they ended up pretty close in performance.

Too bad the WDC 65816 wasn't more successful then it was. I still think that both Atari and Commodore should have released systems around the CPU to provide an upgrade path for users of their 8-bit lines. Yes I know the Apple IIgs was one of Apples best selling computers for awhile, but still.

One other 6502 descendant was much more popular. The Ricoh 5A22 was the heart of the Super Nintendo, for example, and that sold WAY more units than any 8-bit computer.

Non-x86 16-bit computers, for whatever reason, just gravitated towards the Motorola 68000. The Mac and the Amiga are the two most popular examples but Atari's ST used a 68000, too. The 68000 was also at the heart of arcade boards from Sega, Capcom, Konami, SNK (the Neo Geo) and plenty of others. The most famous arcade games of the late 80s and 1990s used some combination of a 68000, a Z80, and some special video and sound chips.

Yeah but the Motorola 68000 was actually was a 32-bit CPU with a 16-bit data path and 24-bit address bus.

whm1974 wrote:

Yeah but the Motorola 68000 was actually was a 32-bit CPU with a 16-bit data path and 24-bit address bus.

The Motorola 68000 was an otherwise 16-bit CPU (16-bit ALU and databus), just with 32-bit registers for forwards compatibility.

srg86 wrote:

whm1974 wrote:

Yeah but the Motorola 68000 was actually was a 32-bit CPU with a 16-bit data path and 24-bit address bus.

The Motorola 68000 was an otherwise 16-bit CPU (16-bit ALU and databus), just with 32-bit registers for forwards compatibility.

Didn't know it had 16 bit ALU.

whm1974 wrote:

srg86 wrote:

whm1974 wrote:

Yeah but the Motorola 68000 was actually was a 32-bit CPU with a 16-bit data path and 24-bit address bus.

The Motorola 68000 was an otherwise 16-bit CPU (16-bit ALU and databus), just with 32-bit registers for forwards compatibility.

Didn't know it had 16 bit ALU.

Yeah, that's why I like to think of the 68000, still as a hybrid, but more as a 16-bit CPU than a 32-bit one. The MC68020 and later are 32-bit through and through.

There could actually be quite a bit of variation:

8086: 16-bit regs, 16-bit ALU, 16-bit Data Bus
8088: 16-bit regs, 16-bit ALU, 8-bit Data Bus
68000: 32-bit regs, 16-bit ALU, 16-bit Data Bus
68008: 32-bit regs, 16-bit ALU, 8-bit Data Bus
80386SX: 32-bit regs, 32-bit ALU, 16-bit Data Bus.
80386DX: 32-bit regs, 32-bit ALU, 32-bit Data Bus
68020: 32-bit regs, 32-bit ALU, 32-bit Data Bus

Please correct me if I'm wrong I've always heard that the Intel 8080 CPU ISA was thought up over the a weekend and the actual CPU was design in a week or two, and it was intend for anything but a microcomputer. Any truth to this?

Takeshi7 wrote:

I'm pretty sure the Z80 or MOS 6502 were the undisputed champs of 8-bit CPUs. The 6502 was clocked slower, but it had around 4x the IPC of the Z80, so they ended up pretty close in performance.

In terms of popularity, yes. But in terms of computing power, no. I think that would go to the 6809. Back in the day, as a young home computer geek you were either in the 6502 or Z80 camps - a bit like GM vs Ford. I was in the 6502 camp with my family owning a C64. Although I later bought a C128 which had both processors. The 6502 could at most execute one instruction every 2 cycles although due to it's "pipelined" architecture this could be fetching 2 bytes from memory such as "LDA #42" or "load accumulator with immediate value of 42". From memory (too lazy to look up my copy of "Programming the Z80" by Rodnay Zaks) the shortest Z80 instruction was 3 (or 4?) cycles with one memory fetch. The 16-bit stack pointer of the Z80 made it a vastly more flexible and capable processor and hence had decent implementations of Pascal, FORTRAN and other high-level languages.

The question is then whether the 6502 qualifies as an 80's CPU? Yes, it was used quite a lot in 80's computers, but the CPU itself was launched in 1975.

demolition wrote:

The question is then whether the 6502 qualifies as an 80's CPU? Yes, it was used quite a lot in 80's computers, but the CPU itself was launched in 1975.

While that's true, and while some computers that used it were built in the late 70s, those times weren't like today where there are a bunch of companies working together to build the latest technology right away. The successor wasn't waiting in the wings until much later, too.

defaultluser wrote:

The most powerful 8-bit CPU DESIGN in the BAD OLD DAYS was the 6809.

https://en.wikipedia.org/wiki/Motorola_6809

It quickly gained a following in demanding arcade machines.

It was pretty beefy, but got outshone by the 68000 released only a year later. The two weren't source-compatible, so the 6809 died off in no time.

I'm sure there are better 8-bit micro-controllers released since then, but this really was the Swan's Song for powerful 8-bit computing.

The 6502 had many more design wins because it was WAY cheaper. We're talking $25 for a processor, in a time when the Motorola 6800 and Intel 8080 cost several hundred dollars. It also had pretty good performance (despite the low price), but came with a lot of restrictions.

256-byte fixed stack (instead of 16-bit), ONE accumulator, and 8-bit index registers. Really primitive compared to most other architectures. But when you saw that $25 price, it suddenly made it all worth it

The fact that the 6809 had a multiply instruction made it my instant favorite. Graduating from a 6502 to a 6809 was like graduating from a bike to a sports car. I didn’t experience this type of elation from hardware until I used the 8087 years later.

Does the 6809 count as an 8-bit CPU if it is capable of full 16-bit arithmetic? That's what the wikipedia article says.

I think we have a new candidate!

https://en.wikipedia.org/wiki/Hitachi_6309

The HD6309, a souped up CMOS 6809 with a 32-bit accumulator, all static design. Plus it as divide and bit-field instructions (not not as fast as what can be done with 6809 instructions).

Anyone interested in a 80186 at 27 MHz on a Intel FPGA? Code is available on Github if you want to try it out.

[Jamie] has implemented the entire set of 80186 instructions in Verilog, and included some of the undocumented instructions too.

Not content to simply simulate a CPU, all the necessary peripherals for a complete working system have been worked into the design as well. There’s RAM, a UART, as well as CGA graphics and a PS/2 controller that is necessary if you’d like to actually use any sort of human input device.

https://hackaday.com/2017/11/03/386-too ... n-an-fpga/

Hudson Soft HuC6280 8-bit microprocessor operating at 1.79 MHz and 7.16 MHz. It features integrated bank-switching hardware (driving a 21-bit external address bus from a 6502-compatible 16-bit address bus), an integrated general-purpose I/O port, a timer, block transfer instructions, and dedicated move instructions for communicating with the HuC6270A VDC. P.us, it had an embedded 6-channel wavetable sound synthesizer that alternatively could output PCM samples at 5-bit resolution, or greater resolution by ganging sound channels. About 6 million produced for home market.

blindsorrow wrote:

Hudson Soft HuC6280 8-bit microprocessor operating at 1.79 MHz and 7.16 MHz. It features integrated bank-switching hardware (driving a 21-bit external address bus from a 6502-compatible 16-bit address bus), an integrated general-purpose I/O port, a timer, block transfer instructions, and dedicated move instructions for communicating with the HuC6270A VDC. P.us, it had an embedded 6-channel wavetable sound synthesizer that alternatively could output PCM samples at 5-bit resolution, or greater resolution by ganging sound channels. About 6 million produced for home market.

So can this 8-bit CPU directly access more then 1MG of RAM or is it still stuck with a mere 64K without resorting to Bank Switching?

whm1974 wrote:

blindsorrow wrote:

Hudson Soft HuC6280 8-bit microprocessor operating at 1.79 MHz and 7.16 MHz. It features integrated bank-switching hardware (driving a 21-bit external address bus from a 6502-compatible 16-bit address bus), an integrated general-purpose I/O port, a timer, block transfer instructions, and dedicated move instructions for communicating with the HuC6270A VDC. P.us, it had an embedded 6-channel wavetable sound synthesizer that alternatively could output PCM samples at 5-bit resolution, or greater resolution by ganging sound channels. About 6 million produced for home market.

So can this 8-bit CPU directly access more then 1MG of RAM or is it still stuck with a mere 64K without resorting to Bank Switching?

Well, he mentions bank switching right in the description...

Sort of related, Alan Cox of Linux kernel fame is developing a Unix-like operating system aimed at these old computers, called Fuzix:

https://github.com/EtchedPixels/FUZIX
http://www.fuzix.org/

One target is the RC2014 modular computer:
https://rc2014.co.uk/

bthylafh wrote:

Sort of related, Alan Cox of Linux kernel fame is developing a Unix-like operating system aimed at these old computers, called Fuzix:

https://github.com/EtchedPixels/FUZIX
http://www.fuzix.org/

One target is the RC2014 modular computer:
https://rc2014.co.uk/

So the rc2014 based on the Z80 8-Bit CPU, what is the limit on Bank Switching?