Addendum:

>> Dave Haynie has clarified that the Hombre's PA-RISC core was not supposed
>> to also act as the CPU in new Amiga desktop class computers (but only in
>> new Amiga gaming consoles): [...]

> New old info has surfaced: [...]

All the details have emerged:

http://aminet.net/package/docs/hard/Hombre
http://aminet.net/package/docs/hard/Hombre_shrinked

In Hombre_part_1.pdf, Dr. Hepler identified several classes of potential Hombre-based hardware (see also Hombre_Presentation_part_1.pdf, pages 3 and 13-20):

1. Cable box system
2. CD-game playing system
3. CD-game based MPEG player (based on #2)
4. CD-game based home computer (based on #2)
5. A1200 class machine (based on #4)
6. Multi-media graphics accelerator on a PCI card (based on #2)
7. Desktop / Tower systems using Hitachi PA/50L as system CPU

In all hardware classes except #6 (obviously) and #7, the Hombre's PA-7150 was supposed to also act as the system CPU. (See also Hombre_part_2.pdf, pages 1 and 16, and Hombre_part_3.pdf, page 3.)
In hardware class #7, an additional PA-RISC chip was supposed to act as the system CPU. This is in contrast to Dave Haynie's statement who said the system CPU would have likely been PowerPC or Alpha.

Supposed date of completion as of September 1993:
"The time frame for this chip set is CY 2H 1994."
(Hombre_part_3.pdf, page 3)

"Current status" in October 1993:
"Block diagrams of all chips have been generated. Interfaces to memory and between chips have been defined. Schematics of the CPU chip and Video chip have been started and various blocks have been simulated."
(Hombre_part_1.pdf, page 19)

Supposed date of completion as of April 1994:
"Late 1994 "First Silicon" Availability for 1995 Production"
(Hombre_Presentation_part_1.pdf, page 3)

"Status" in April 1994:
"- Schematics for many major blocks have been captured
- M language, synthesizable, behavioral models have been written for much of the functionality of the chip set.
- Some simulation has been started, much remains to be done."
(Hombre_Presentation_part_2.pdf, page 30)

A lot of people have memories of the Amiga that don't match the reality. The first is it's CPU: many people still believe the 68k to be a super CPU, but in reality it was a 1979 design actually not much faster than the Zilog Z80 inside a ZX Spectrum, and by that by the mid-80s had been clearly surpassed by the 286. Apple even had an 8-bit CPU that was faster. Updates from Motorola were anaemic and more meant for Apple's needs. Pushing a 680EC20/14 out in a 1991 A1200 was almost a joke, although AGA hid it for a long time.

The second was the Amiga's graphics, which by 1985 were clearly ground-breaking, but by 1990 compared poorly to VGA in terms of colour and data transfer rate. AGA, as Haynie admitted, was too little too late. A good in-place upgrade in compatibility terms, but only a stop-gap.

Hobre was an attempt at solving both problems at once, even fixing Commodore's decade-long indecision on whether their computers should be for productivity or games. Hombre was intended to be both, and to not only restore the Amiga's technological edge, but push it even further over the rest of the market than the A1000 had.

Unfortunately by this point, senior CBM management was no longer listening to the engineers. They'd ridden the C64 all through the 80s, failing with everything else but the Amiga, gotten rid of Tramiel, treated their engineers like garbage, and no doubt believed they should ride the Amiga all through the 90s. The company was haemorrhaging cash left right and centre, and the execs were only in it for themselves - stories abound of corruption and cocaine and hookers. So it never happened. And Commodore finally went bankrupt, and were still making both Commodore 64s and ECS, 68000/7 computers when they did.

TL;DR: It was a non-compatible Amiga replacement that Commodore engineers wanted to develop but didn't because management were self-absorbed idiots.

Quote:

I remember the 68000, especially in regard to its status compared to other cpus, as I sold both 68K and Intel based systems.
You are WAY off when you make that comparison between the 68000 and the Intel '286.
Because of its address mode limitations, the '286 was fairly useless to my company as we couldn't port our 68K code (which relied on position independent addressing) to it.
We actually looked forward to the introduction of the '386 (particularly the i386SX as it could be had relatively cheaply), and attended several meetings with other companies anticipating its availability.

Your opinion of the 68020 is equally off base, as the '20 is, in many operations, several times faster than the 68000. It was a nice leap forward, and was the cpu we used in our higher end offerings.

As to Motorola's later developments, they weren't produced primarily for Apple.
In fact the '040 was only used in a limited number of systems and the '060 wasn't used at all. The '60 being introduced too late to even be considered by Apple, which was already shifting to the PPC.

Although, before the Pentium, 68K cpus were holding their own in comparison to Intel.

As the ISA grew more dated, the focus did shift.

The reason Coldfire has some functionality reduced, is that Motorola was shifting 68K focus toward its most successful market, embedded products. This is where 68K and related cpus continued for some time to come.

Next, the Z80, an eight bit cpu with a 64K address range, is hardly comparable to a 68000. Its a knock off design promoted as a cheap alternative to Intel, plain and simple. Doesn't really have that much to offer outside of being able to run CP/M.

You are right that Amiga graphics were dated ,but you are preaching to the choir here on that, as MorphOS only supports re-targetable graphics.

Its interesting to note that PA-RISC doesn't exist anymore (or Itanium), and that PowerPC is winding down, but 68K seems to just keep trudging along, with a recent re-introduction of the 68000 and 68020 cpus by a licensee.

To close, I'm of the opinion that if the Amiga had survived, either at Commodore or at one of the future IP holders, that it definitely would have transitioned to a PPC cpu.
The alternative cpus, outside of Intel, held little appeal.
And at the time, CISC seems to be a dead end approach that even Intel thought was limited

[ Edited by Jim 17.07.2017 - 11:41 ]

> Commodore finally went bankrupt, and were still making [...] ECS, 68000/7 computers
> when they did.

According to Wikipedia, the production of the last ECS-based model (A600) ended mid-1993, so almost a year before the declaration of bankruptcy.

Quote:

Which would leave the Motorola 68EC020 powered A1200 as the base model, when CBM went bankrupt.
A computer that CAN run Linux (which i'286 powered computers can NOT do).

That, btw, being one of my benchmarks for the value of a cpu.
Can it run UNIX derived or UNIX-like operating systems?

And while you can (sort of) torture that out of an i'286, it will do it REALLY poorly.
While ALL members of the 68K family WILL run that type of software (it just had MUCH better addressing modes).

SO, please don't point me toward Intel, as they were not a serious contender until the i386, and really it wasn't even that (or the i486) that got them on the map, it was the Pentium.

And THAT cpu I would acknowledge was the death knell for the 68K family (at least on the desktop).

> You are WAY off when you make that comparison between the 68000 and the
> Intel '286. [...] Your opinion of the 68020 is equally off base [...].

I guess KennyR doesn't refer to when the CPUs were released but when official Amiga models with these CPUs were available, which was sometimes much later:

- 68000 released in 1979, A1000 released in 1985
- 68020 released in 1984, A2500/20 released in 1988
- 68030 released in 1987, A2500/30 released in 1989
- 68040 released in 1990, A3000T/40 released in 1991/1992
- 68060 released in 1994, A4000T/60 released in 1997

As can be seen, official Amiga models based on 68000 and 68020 were first available as much as 6 and 4 years, respectively, after the release of the respective CPU, which can be considered a long time. It became better with the 68030 and 68040, and worse again with the 68060 (because of the Commodore bankruptcy, obviously).
Note that the above list only mentions big box Amiga models, while the much cheaper and much better selling wedge-type models took even longer to adopt the respective CPUs (A500/A1200: 8 years, A600: mind-boggling 13 years).

> In fact the '040 was only used in a limited number of systems

According to Wikipedia, there were:
- 12 systems with 68(HC)000
- 2 systems with 68020
- 45 systems with 68030
- 46 systems with 68(LC)040

To me this looks like the 68(LC)040 was used in the biggest number of Apple systems.

> I'm of the opinion that if the Amiga had survived [...] at one of the future
> IP holders, that it definitely would have transitioned to a PPC cpu.

Some may argue that this is what actually happened ;-)

> The alternative cpus, outside of Intel, held little appeal.

Dave Haynie mentioned DEC Alpha as a serious contender to be heir to the m68k in the early to mid-1990s (see comment #3).

> the Motorola 68EC020 powered A1200 [...] CAN run Linux

Linux requires an MMU present, so the 68EC020 won't cut it. Minimum would be 68020+68851, so the stock A1200 is out.

> please don't point me toward Intel

I certainly didn't :-)

Quote:

You know, I hadn't considered the need for an MMU.
And my primary experience with UNIX (or UNIX-like) OS' (outside of Microware's OS9) was Xenix.
That doesn't require an MMU (although it could benefit from one), and its damned slow on a 6-8 MHz 68000.
Which, come to think of it, was one of the reasons we were working with OS9.
That and the fact that the OS required reentrant code, making it pretty efficient with memory.
After all, I don't think anyone remembers how damned high memory prices used to be.
Every 512K of memory cost our users as much as the average price for a 8 bit PC.
It was not cool.
These days, code reentrancy is not a particularly useful feature, as the utility of multiple processes using a single module (outside of memory savings) isn't that important.
But, two other features that OS highlighted, position independent code, a microkernel core...still pretty useful stuff.

BTW - You CAN create a Linux distro without an MMU, but some commands get left out (as they just won't work).
But, the kernel itself can be ported, its been done.

Which IS more than you can say for early X86, which can't support anything much more complicated than CP/M or DOS.
Even Windows didn't run well until you bumped your specs up to an i'386 or better.

So again, no, the i286 in no way obsoleted the 68K series.

> You CAN create a Linux distro without an MMU [...].
> [...] the kernel itself can be ported, its been done.

Indeed, see μClinux.

> early X86 [...] can't support anything much more complicated than CP/M or DOS.

Xenix does run on 8086 and 80286. After the 8086 port, it took one more year to port it to the 68000 (and 80286).
Further "UNIX (or UNIX-like) OS'" which do/did run on 8086 and 80286 are:
- QNX
- MINIX
- Venix
- PC/IX
- Idris
- Coherent
- ELKS
Besides, also OS/2 was originally created for the 80286.

>Xenix does run on 8086 and 80286. After the 8086 port, it took one more year to port it to the 68000 (and 80286).

Interesting, I didn't try the earlier port.
Its damned slow on a 68000, I can't imagine it on something less competent.

And it would not surprise me if OS2 could be run on a '286, as you can run Windows on that (or even an 8086, I had it running on a V20 at one point), although it probably runs rather poorly.

We simply could implement some of the things we were doing before the '386, and the '486 and above were really much better at producing the kind of results we'd been getting with 68K family cpus.

Glad to know you're aware of μClinux, although its a port I'd avoid, it exists.
After your mention of the requirements for more mainstream ports I checked out Debian and a few others.

Its really annoying that a few of the available '020 based accelerator use EC processors (like the Furia), but hey I'd probably want an '030 or higher anyway.
And of course here's another issue with the Vampire (unless they've enabled an MMU).

Minix...hmm, almost stepped in that one in the '80s and early '90s, but since Minix 3 is still pretty crude, its one I'm avoiding. Which is a pity since its likely to be more efficient than BSD. Microkernels...Tanenbaum definitely has my respect for his position on that one as opposed to Torvalds' advocacy of the messier alternative. Pity Minix itself has never been as well fleshed out as some of the process control OS' based on the concept.

>To me this looks like the 68(LC)040 was used in the biggest number of Apple systems

Interesting count, I'll keep it in mind.
By the time we got to the '040, I rather liked what Next had to offer, but must admit I never paid that much attention to Apple's offering (even if they are pretty much derived from that technology).

I still think an AROS 68K port to Next makes sense.

And its a pity Apple didn't use the '060, as it was easily competitive with their early PPC based systems.

AND TO ALL, my apologies for the off topic content.
On Hombre, we keep hearing about this project in connection with Amiga.
But isn't it fair to say that it was a replacement for the Amiga?
Or at least that the CD64 was intended as a follow up to the CD32, but not necessarily as an 'Amiga' (except as a possible trademark only).
The Soc Hepler is quoted as mentioning that would provide backward (Amiga) compatibility didn't have to be included (and probably wasn't even planned for inclusion in the CD64 - it would have added to the cost).

I guess what I'm saying is Commodore didn't seem as hung up on the Amiga trademark as many are today.

[ Edited by Jim 18.07.2017 - 09:59 ]

Not sure whats your beef with EC20 on some new cards as these don't feature enough RAM to need more than 16MB addresspace.

An 020 with intregrated MMU is called 68030 btw.....

>An 020 with intregrated MMU is called 68030 btw.....



That IS pretty much all the '030 is, other than the fact that its available at higher speeds.
Actually, the '020 holds a special place in my heart (and for my former uses an EC is fine and an MMU is unnecessary).
I loved how much more powerful, clock for clock this was than the 68000.
We didn't get another comparable jump until the '040 was introduced (but the '30 did scale pretty high).

I think I mentioned before that the last system we offered was '020 based (there was an '030 in the works, but the company founder passed away unexpectedly).

BTW - No speculation here on whether or not Hombre based systems would have been branded Amigas?
I'm not 100% convinced they would have been. Particularly not if they wound up running Windows NT.

Quote:

Yes indeed. Thanks for the excellent summation; exactly my point.

This state of affairs was of course nothing special. The C64, released in 1982, was provided with a 1976 MOS6502, more or less the same processor the Apple II had used in '77. The Z80 was still being used in consoles until the early 90s, and in arcade machines for who knows how much longer (albeit with a much higher clock than any ZX Spectrum).

To be fair, the popularity of the Amiga was ignited by its capabilities, but sustained more by reputation, especially as time went on. Fact is, both of Commodore's landmark machines (C64 & A500) were technologically behind contemporary machines, but boosted by excellent custom chipsets that boosted them way above. And of course, beyond all, a massively talented userbase was the key to the Amiga's success, as it was also the secret behind the ZX Spectrum's, a computer so basic that it couldn't even colour a moving sprite without attribute clash.

Hombre may have, or may have not put Commodore back in front, but personally I don't think so. The PC may have been awful for games until '94 or later, but by the time they did come, it had a set of established APIs and an OS and software base that could take advantage of modular hardware. In contrast, the Amiga was still an oddity that many treated like it was still an 80s computer, with hardware hacks, pokes and under-the-hood assumptions, and was ill-prepared for an age where custom chipsets just didn't cut it any more.

> By the time we got to the '040, I rather liked what Next had to offer,
> but must admit I never paid that much attention to Apple's offering
> (even if they are pretty much derived from that technology).

Yes, the fact that Jobs took m68k Mac hardware engineers with him to NeXT probably meant that NeXT hardware was in part derived from Apple hardware. But then probably only to a small degree as these engineers only worked on 68000-based hardware at Apple until 1985 while the first NeXT computer in 1988 was 68030-based.

> isn't it fair to say that it was a replacement for the Amiga?

Hombre was supposed to be new Amiga technology replacing the old m68k+OCS/ECS/AGA technology.

> the CD64 was intended as a follow up to the CD32, but not necessarily as
> an 'Amiga' (except as a possible trademark only).

Oh, not that futile "what's an Amiga" discussion, please ;-)

> The Soc Hepler is quoted as mentioning that would provide backward (Amiga) compatibility

Can you point me to this mention?

> Commodore didn't seem as hung up on the Amiga trademark as many are today.

I see no indication that Commodore wouldn't have used the 'Amiga' name for Hombre-based products.

> No speculation here on whether or not Hombre based systems
> would have been branded Amigas?

I guess they would have, at least for some of the hardware classes Dr. Hepler envisioned.

> if they wound up running Windows NT

See comment #7 for information about which OS Commodore planned to specifically create for Hombre. Windows NT could have been ported, of course, but was never meant to be the main OS.

@Kenny,

Yes, I'd agree that the custom chips definitely make the C64.
Since I'm a fan of 68000 based systems besides the Amiga, there the use of a custom chipset isn't as important to me.

In fact, the systems I worked with in the late '80s were rather generic using a off the shelf ET4000 based video card.
Then again, gaming was NOT part of our market.

@Andreas

>> The Soc Hepler is quoted as mentioning that would provide backward (Amiga) compatibility

>Can you point me to this mention?

Its not a direct quote, so it could be hearsay.

And would Commodore have bothered to add this? I'm unsure.

In any case, examining the plans for Hombre, I'd have to say that it doesn't look that promising.
The approach reminds me of early TI accelerator boards.
From the documentation it looks like the PA-RISC cpu was incorporated primarily to be a video controller, and that in higher end applications additional processor would be used.

If that second cpu wasn't another PA-RISC processor, the entire platform begins to look kind of messy.

>>> The Soc Hepler is quoted as mentioning that would provide backward (Amiga)
>>> compatibility

>> Can you point me to this mention?

> Its not a direct quote, so it could be hearsay.

Interesting nevertheless, as I don't remember I ever heard of this before.

> From the documentation it looks like the PA-RISC cpu was incorporated primarily
> to be a video controller

The video controller is the other chip of the Hombre chip set. The functions of the two individual chips are as follows (according to Hombre_Presentation_part_1.pdf, pages 11/12):

Nathaniel:
- RISC Int. Core (MMU/Cache)
- Blitter/Renderer
- 8-bit DMA Interface
- Audio Processor
- CD-ROM Interface
- PCI Bus
- Memory Copier
- Copper

Natalie:
- Line Buffers
- Cursor Ram
- 512 x 25 CLUT
- Framegrabber
- Genlock
- YUV -> RGB (MPEG)
- Mouse/Game Cntrlr Intf

> and that in higher end applications additional processor would be used.

Yes, this has been established in comments #3 and #8.

> If that second cpu wasn't another PA-RISC processor, the entire platform begins
> to look kind of messy.

Yes, that's what the last sentence of comment #3 refering to Haynie's statement was essentially about. Fortunately, this newly emerged documentation of Dr. Hepler's clears that mess up.

Quote:

How many cycles does it take for a z80 to perform a 16*16 bit multiplication to a 32 bit product?

Z80@4Mhz is *much* slower that 68000@7Mhz. Maybe 10 times slower (or even worse).

8x16 bit Z80 multiplication...

Mul16_8:
ld b,8
ld hl,0
Loop:
rra
jr nc,Jump
add hl,de
Jump:
sla e
rl d
djnz Loop

[ Edited by Jeckel 20.07.2017 - 10:24 ]

Quote:

Yes, it is, and I'm a big fan of the 68000, but really its yet another apples and oranges comparison.
I can get better multiplication performance out of a Hitachi 6309 at half that speed (and I'm running those at 3.58 MHz in one project). Although...that can only do 8 x 8 multiplication or 16 x 16 multiplication (which obviously could be used for 8 x 16).
In any case, my point is...comparison of an 8 bit cpu to another 8 bit cpu would be more valid.

And that's what we have to remember, the Z80 is an earlier 8 bit design intended to compete with Intel's 8 bit cpus.
Its not fair to compare it to the 68000.

Edit - BTW, a 68000 processor is what you ADD to a Tandy Model 2 (which already has a Z80) in order to run something more powerful than CP/M, like Xenix. They're known as Model 16s or 6000s.



[ Edited by Jim 20.07.2017 - 10:42 ]

Quote:

Yes indeed, they're just 8 bit CPUs. Even at the simplest a word would require at least two cycles. In practice, most operations used multiple cycles.

But by 1985, Z80s were often running at 8 MHz, and by the time the CMOS versions came along (I don't know what year), at 20 MHz and up. The original 4 Mhz did 0.58 MIPS raw; the 68000 did just 1.4 MIPs at 8 MHz.

As a modern comparison, the Raspberry Pi 2's SoC does 4744 MIPS.

My point was, that the Amiga "super CPU" wasn't quite so super as we believed it was, even in its day.

[ Edited by KennyR 22.07.2017 - 21:51 ]

> The original 4 Mhz [...]

The Z80 was introduced at 2.5 MHz.

MIPS ratings don't really give you an accurate measure of a processor's relative power when comparing dissimilar processors.
Especially when considering early microprocessors.

[ Edited by Jim 22.07.2017 - 18:43 ]

> since Minix 3 is still pretty crude, its one I'm avoiding. [...] Pity Minix itself has never
> been as well fleshed out as some of the process control OS' based on the concept.

Interestingly, in 2015 Intel‘s Management Engine was switched from running ThreadX on ARC to running MINIX 3 on x86.

"Starting with ME 11, it is based on the Intel Quark x86-based 32-bit CPU and runs the MINIX 3."
https://en.wikipedia.org/wiki/Intel_Active_Management_Technology#Hardware

"MINIX 3 is believed to be used in the Intel ME found in Intel PCH beginning with the introduction of ME 11 which is used with Skylake/Kaby Lake processors."
https://en.wikipedia.org/wiki/MINIX_3

Tanenbaum's open letter to Intel:
http://www.cs.vu.nl/~ast/intel/

[ Edited by Andreas_Wolf 08.11.2017 - 15:19 ]

Addendum:

> All the details have emerged:
> http://aminet.net/package/docs/hard/Hombre
> http://aminet.net/package/docs/hard/Hombre_shrinked

Some (further) non-public marketing material on the Hombre-based plans:

Hardware technology history/plan
Amiga systems plans
Game system enhancement plans