Sorry for the OT. But I've been reading some stuff about the 68K and came across the DTACK Grounded newsletter. Now, if you're a fan of the 68K then you will know that the newsletter (and author) is quite respected in that world.

But I couldn't help but get a chuckle when I read the following from the first issue (circa July, 1981):



OK, in 1981, a high-end 8086 system could probably beat an Apple II or VIC-20. But I don't quite think the 6502 went the way of the 4004/4040. At least not by 1982! HAHAHA

Source:

http://www.easy68k.com/paulrsm/dg/dg01.htm

_________________
Cat; the other white meat.

Ah yes, that newsletter, interesting and amusing as it could be, was always written with a certain tone of voice. I suppose it's one way to try to persuade people: and that first issue seems to be about persuading people that 68k is the future, by means of telling them that everything else is old hat or a dead end.

It did turn out that 68k was a good bet, and had a pretty good run!

Not ever, actually. This seems closer to "completely wrong" than "arguably wrong."

The 8086 is basically dead. It's preserved within the '386 and every subsequent CPU, but that seems to be mainly some sort of weird inertia related to the endless backward compatibility of everything that came after it. The 386 flat model is nearly as different from the 8086 segmented model as the 68000 is from the 6809; nobody (except for retrocomputing hobbyists) been doing new designs or writing new programs for the old 8086 architecture any more.

On the other hand, the 6502 seems even now still to be more successful than its intended successor, the 65816. It's still in production (even in new forms, such as FPGA cores), and seems to be used even for new designs.

Oh, and there was a TRS-80 Model 4, and it used a Z80; it was basically just a slightly improved Model III.

_________________
Curt J. Sampson - github.com/0cjs

Yep. At least the first issue was honest up front. It wasn't a secret that the mission was to sell his hardware.

But I did find it amusing that everything else in the world was junk when the 68K came along.

_________________
Cat; the other white meat.

Well, that part (as opposed to "nobody will be using the 6502 in a few years") I do find to be a plausible argument. Aside from cost and compatibility, I don't know that there's any advantage an 8-bit processor has over a 68008.

(Come to think of it, I have a 68008 in my parts bin. I guess I should wire it up one day to see if it works. I doubt I'll do much with it, though; I'm not really into all this newfangled 16-bit stuff.)

_________________
Curt J. Sampson - github.com/0cjs

Interrupt performance immediately comes to mind. The 6502 is way, way, way ahead of the 68008 or 68000 in interrupt performance.

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

Really? 68008 taking twice as long makes sense becuse it has to move 16-bit words across an 8-bit data bus, but I thought the 68000 interrupt handling cycle was essentially the same as the 6502: push PCH, PCL, PSW, read vector high and low address parts and then jump to that vector.

_________________
Curt J. Sampson - github.com/0cjs

I'm not familiar with the 68000; but when I was collecting information for comparison, the numbers I got, which might have been from WDC or maybe from 68K data sheets I got rid of a few years ago, were that the 68K took 46 clocks or 5.75μs minimum @ 8MHz for interrupt acknowledge plus interrupt sequence, not including the time it took for the currently executing instruction to finish, which in the extreme case was something like 170 (which was probably for a DIVide instruction). That was for a 68000, not 68008. I don't have information on a return from interrupt.

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

68000 had looked like quite the thing: it had been the standard for the rather expensive 19" related industry control stuff.
Motorola MC68000 was second sourced by Hitachi (HD68000), Mostek (MK68000), Rockwell (R68000), Signetics (SCN68000) SGS-Thomson (EF68000) and Toshiba (TMP68000).
Also, there were microcontrollers from Motorola (MC683xxx, core was a 68020 derivate) and Philips\NXP (68070, core was a 68000).

And at some point, poof: everything 68k related suddenly and completely had disappeared out of the catalogs of the distributors. //Like in "dinosaurs versus big asteroid".
The Motorola processor division had turned into Freescale, which went bought by Philips\NXP.
Nowaday, everything at the NXP homepage seems to be ARM related, while 68000 and 683xxx are 'not recommended for new designs', some parts already are out of production.
There still is the Innovasic (now a subdivision of AnalogDevices) Fido 1100 in ball grid array package, but IIRC the Manual doesn't explicitly mention 68k compatibility, so Fido doesn't count.

To put it short: new 6502 parts still are available in DIP, while IMHO 68000 is about to go the way of the 4004/4040.

Yeah. There's a lot of things to like about the 68k series, but tight, highly-deterministic instruction/interrupt timing is not one of them.

That's why Fido 1100 has 4 sets of registers: for highly-deterministic interrupt response time (IIRC within a single clock cycle at 66MHz).

Initially, Innovasic had the plan to build something like a 200MHz ARM, but customers had insisted in having a 68k compatible instruction set.
That's why Fido 1100 has the Motorola CPU32 instruction set (like the 68332 has),
it's binary compatible to the 68020 minus a few of the more sophisticated addressing modes and such.

But back on topic: the advantage (and beauty) of 68k was the linear and flat memory addressing with 32 Bit registers.
65816 bank switching still feels a bit awkward, and the 80x86 real\protected mode plus segmentation stuff really could hurt your brain.

6800 had been designed at Motorola, but the design team later had jumped ship and did the 6502.
Later, Motorola had put a lot of time, money and effort into inventing a "future proof" ISA, the 68k.
68k later went (more or less) obliterated by the 6502 insipred ARM, IMHO there is some irony to this.

It's interesting to see, how management\marketing decisions had shaped the landscape since the invention of the 4004.
In a parallel universe, we all would have 68k powered PCs, running nice and reliable without Microsoft.
In another parallel universe, we all would be tinkering with (godforbid) 80x86 instead of 6502 (and 68k).


Edit:
Between 1979 and 1994, it had been exciting to watch the arms race between Motorola 68k and Intel 80x86.
For me, it had been hard to predict which of them two ISAs would be going to make it into the far future...
...until Motorola had axed 68k development after the invention of the 68060 while introducing the ColdFire.

ARM had the great advantage of running on 32 bit wide memory (and making very good use of it) - that's partly a matter of (historical) timing, the 68k would have been much less desirable if that was the story. Even 16 bits wide felt a bit high-end, needing pairs of ROMs and sets of 16 DRAMs. Which is why the 68008 had a certain amount of appeal, early on.

I think the other problem is that a CISC machine is difficult to iterate on to chase ever-higher performance: eventually, at huge effort, Intel managed to do that, but it was an exceptional case. They tried several times to ditch x86 and start again with something simpler and more scalable. So, I think, the 68k line couldn't keep up with performance expectations. For embedded applications, MIPS was very popular for a while, and eventually ARM took over as the obvious choice.

It's an interesting question, perhaps, if WDC had never got started and not offered the '816, would it make any difference to us here? There were several other manufacturers of 6502 besides Commodore/MOS, and they had CMOS versions. And we know from much later work by many other individual efforts (21 HDL cores linked here) that reimplementing a 6502 in an HDL is a practical proposition, which means any number of companies could have written their own and used it or licensed it for custom or semicustom chips.

That is, the great popularity of 6502-based micros in the past, and the simple nature of the CPU, means we might still have a keen community of enthusiasts even without WDC.

If WDC would have introduced the 65816 shortly before Motorola had intruduced the 68000, I think this would have made a lot of a difference.
Unfortunately, the 65816 was introduced in 1983, four years after the 68000.
So the 68000 already was established on the market, not much space there for the poor 65816.

In 1979, RAM still had cost an arm and a leg, so a CISC instruction set with a variable instruction length and an 8 Bit or 16 Bit data bus for 68k had made a lot of sense.
But at some point, to me it looked like the complexity this had caused had made it difficult for Motorola to build faster 68k compatible CPUs.
IIRC the superscalar 68060 internally had translated the 68k instructions into 48 Bit RISC instructions.

After the 68060, the ColdFire instruction set had looked somewhat similar to 68k at assembly level, but at binary level it was completely different from 68k,
I still think this was for reducing the cost for designing and manufacturing CPUs.
There was a story, that Apple had put some pressure against Motorola for lowering the price of the 68000...


The 6502 CPU chip internally is, in no way, simple.
But the ISA and the system integration are simple enough to be handled by just one "lone wolf" hobbyist, for most of the other and more modern CPUs it might be a little bit different.
There is plenty of software and design examples for the 6502, and all of the bugs and oddities of the 6502 related chips are supposed to be well documented and known by now.
If you start tinkering with a new (and maybe ARM based) microcontroller, the bug list tends to be longer than your arm, most of it is related to the peripherals.

Looks like the design of the chips in the 6502 family hasn't changed much since 1983, so it isn't considered to be a "state of the art" or at least a "modern" design.
But to put it this way: as a human, you are supposed to be the pinnacle of evolution, cutting/bleeding edge and such, development still ongoing.
The shark is a design that hasn't changed much for the past 16 million years: It didn't have to.
Now try to compete with it for food in its natural enviroment.
See: in its natural environment, IMHO nothing can beat the 6502.

If you try designing a CPU which uses as few chip space as possible while making efficient use of a small memory size, no matter what approach you take:
the end result somehow always shows a tendency to resemble the 6502 ISA.

No need to compete with the shark for food - we are the food!

In calling the 6502 simple, I suppose I mean mostly to say that it's a very manageable task to cook up an implementation - not that it's trivial. Also, the LUT count of an efficient 6502 implementation is impressively low, and of course the original transistor count is low too. So, for microcontrollers, it can be a good architecture, because size is cost. Hence the huge numbers of toys and other mass market products which use a 6502 or 6502-lite.