Commodore 900 vs Commodore Amiga 1000

[WolfToTheMoon]

In 1984, Commodore bought Amiga and released what is now known as the Amiga 1000. The 1000 was, in honesty, a rather poor seller, and not untill the Amiga 500 and 2000 did Amiga gain significant market traction.
In 1984, Commodore was also getting ready to release the Commodore 900 UNIX machine, that was developed inhouse by Commodore Germany(the same team would later design A500 and A2000). Looking back, was it really smart to choose Amiga over the 900?

The 900 was marketed(before it was cancelled, about 500 units were made, and later, sold for 4000$ in Germany) at around 2700-3000$ for a base machine, with 512kB RAM(expendable to 2 MB on motherboard), 20 MB HDD, 1.2 MB 5,25" floppy and 10 MHz Z8001, running Coherent OS(An UNIX clone) + Commodore's own windowing software by Rico Tudor. There were 2 versions, one a server and other a workstation(1200x800 monochrome display with a dedicated 128 kB video chip, 14" and 20" monitors offered as option).
On the other hand, Amiga 1000 had a 256k of RAM, 880 kB 3.5" floppy, 7 MHz 68000 and no disk drive, yours for 1295$ in 85'.

When Amiga was released, it was released into a pretty competitive market, with Mac, Atari ST and, up to a point, IBM PC compatibles. However, C900 had little competition for 2700$ on the UNIX market.

A post from Dr. Peter Kittel from usenet groups illustrates the impact of the C900 had, at the time...

Yes, at that time it meant just below DM 10,000. You should have seen
the hords of worried HP and DEC people coming over to our booth(ceBIT 85') and
look at that beast and recognize it did practically everything their
much more expensive, established machines did.

Comparing the 2 system, IMHO, the Z-machine had a bigger potential(except maybe in regards to Z8001 vs 68000)... in 2-3 years, cca 88-89' timeframe, Zilog would have probably finished the (initially buggy) Z80000 and thus give it a full-on 32 bit upgrade path(Z80000 was a fully pipelined(6 stage) design with 256 byte of on-chip cache, comparable in some regards to 68040 and 486, but few years earlier). Commodore even considered buying Zilog in 84-85', obviously for intentions of full vertical integration of the C900 line, just like with the 8 bits.

In the end, C='s poor financial situation in 84-85 and the acquisition of Amiga spelled the end of the Z-machine. I wonder if, knowing what we know now, C= would have been better off going with the C900 vs the Amiga. It would give C= a presence in the UNIX market, a modern, easily portable UNIX based OS and possibly even Zilog and their pretty advanced 32 bit chip.

[WolfToTheMoon]

Commodore may or may not have done well to introduce the 900 (The Z80000 is certainly a pretty nifty chip for 1986,) but it wouldn't have been a good alternative to the Amiga anymore than the Amiga would've been a good alternative to the 900. Unix was even less of a home-user OS in the mid-'80s than it is today (hell, XWindows had just been introduced a couple years prior, and had a long, long way to go to catch up to where even the Amiga started usability-wise, to say nothing of Mac OS's polished GUI.)

C900 used Commodore's own windowing system, described by Dave Haynie as being very fast and polished.
I'm not arguing that C900 would replace Amiga... it wouldn't... it was 2 times the price of the A1000 and aimed at higher educational and business market. I'm arguing it would have been better for C= as a whole.

And as good as the hardware may have been, there was and is a lot more to success in the big-iron market than pure specs - software options, reliability, quality of support, etc. Could Commodore have delivered on those? I dunno...

Agreed... That was one area where C= badly lacked.

[WolfToTheMoon]

Interesting.  I never heard of the C900.

That said, I think Commodore made the right choice.  Back in the 80s $2700-$3000 was an awful lot of money.  

Indeed... it is, but relatively speaking... Macintosh 128k was 2595$ in US(the quoted price for the C900 is the European price, as C900 was not initially planned to go on sale in the US).
Compare a Macintosh(128k RAM, 7 MHz 6800, no disk, lousy OS, no software) with C900(512k RAM, 10 MHz Z8001, 10 MB disk, UNIX based, thus has available software - C900 was planned to ship with an C compiler, BASIC compiler and 50ish apps built in) and I think it's actually a steal at that price.

[WolfToTheMoon]

Put that way, it does seem a bargain.  

However, from what I remember of the average family living on an average wage, it was in the "business" price-range, rather than directed at the home user.  The same can be said for the A1000 and it was really the arrival of the cut-down and much cheaper A500/Atari ST that brought a new generation of computing into the home.

I imagine that C900 would excell in higher education, on universities, and in small to midsize companies. Also, it's 1 megapixel screen in 85' might establish C= as a market leader in DTP, and not Apple.
For home users, C128 and it's follow ups( let's call it C256), on which Haynie and his crew worked before A2000, would replace A500.

Somewhere in, say 88-89, C= would release a Z80000 based C900 successor, while producing something cheaper for the lower end. Possibly a cheaper, simplified C900, which would be competitive with lower end Macs and 286 PCs untill 90-91.

[WolfToTheMoon]

Commodore were a really lame company and did not do Amiga justice most of the time. Despite that, they did the wise thing of staying out of the 80s business trend of trying to captitalise on the aftermath of the AT&T split and revive a terrible (for the desktop) OS.

It's a business and scientific targeted machine. And still has multiuser and protected memory in 85, which AmigaOS still doesn't have.

In the end of the day their only chance would be to sell el-cheapo Z8000-based machines to the enterprise (Z8000 was introduced in 1979)

which is the same year MC68000 debuted

rather than risking with a new product (Z80000) of a dying company (Zilog was in rebuilding/restructuring/survival mode under Exxon).

C= was in talks to buy Zilog at that time.

[WolfToTheMoon]

Yes. MC68000 was a more capable processor and 68020 was released 2 years before the Z80000 - the latter was good on paper but got delayed and did not sell.

68000 was no more capable than Z8000... The benchmarks from  BYTE show that Z8000 is somewhat faster per clock than a 286. That should make it comparable to a 68000.

In 1987, if Commodore goes down the C900 route(and buys Zilog), you would probably have a fully pipelined Z80000 with MMU on board vs a vanilla 68020, which didn't have an MMU and had 3 pipeline stages. In theory, Amiga, as happend historically, would not have an comparable CPU until late 92' and 040 A4000.

Zilog rated a 10 MHz Z320(CMOS version of the Z80000) at 10 MIPS peak. Motorola rated a 50 MHz 030 at 18 MIPS and 44 MIPS for a 40 MHz 040. While MIPS rating is not the best benchmark for different CPUs, it shows the potential the Z80000 architecture had.

[WolfToTheMoon]

I don't understand what this thread is about.

Commodore made Amigas.
Commodore made Unix machines.

Nobody wanted the Unix machines so they did not sell well.

When is the last time you saw a for sale ad on any Amiga website for a Commodore Unix machine?  Never?  Unix sux.  Its 10x slower than AmigaOS.  Complete junk.

Commodore put Unix on the A2000 with a 68020 accelerator card and called it the A2500/UX.

Commodore put Unix on the A3000/030 and called it the A3000/UX.

We used to "sell" them. Ha!

I think they later had an A4000/UX too.


Commodore offered multiple Unix machines at cheap prices to the free market and the free market rejected cheap unix machines.

SUN wanted to sell A3000UX, but Commodore botched the deal. How's that for not selling?

When Commodore planned to release C900 in 85' for around 3000$, SUN was selling 68000/68010 UNIX systems for +25 000$ !!!(OK, it had a bit more memory and disk space, but still) ... You think C900 wouldn't sell? I think it would sell far better than Amiga 1000 ever did.

What Commodore got with the Amiga was a very good gaming system, but which highly depended on the Motorola CPU and it's custom chips and was not easily portable.

 With the C900, they could have made a C= variant of Mac OS X, only much earlier. OK, by the mid 90s they will probably be f*cked by Intel and x86, but that's just it, UNIX is highly portable and it wouldn't take much to change ISAs. There would be no OCS, EGS and AGA dependencies...

Oh, and as to gaming.... Haynie claims that C900 had a more sophisticated blitter chip than the Amiga.

[WolfToTheMoon]

Coherent never had a TCP stack until its publisher folded in the mid nineties. That doesn't exactly scream "next big thing" at me.

Yes, I know... but adding a TCP/IP stack to an UNIX clone wasn't rocket science, was it?

[WolfToTheMoon]

Zilog's Z8000 had some interesting features but also bugs, compatibility issues and it was not ready from the start (not the MMU). It didn't translate good for the desktop either. Olivetti was selling $5000+ HD-less and unixless Z8000-based machines then; it was the HD, the RAM, the support that was expensive. The Z80000 was only on paper and the company under Exxon was a mess and many people had left in the mid 80s. This was not the case of 70s with CBM buying up and coming MOS, it was a case of entertaining opportunities to enter the enterprise after the party was over. It would take some time to catch up and find clients, the trend in the CPUs had already changed, it was not a sure thing they would manage to license Z8000 to other (i.e. Olivetti) and it could take years before Z80000 was ready. You never know but I guess they would fail miserably.

My theory is that C= wanted Zilog because they were after vertical integration of their 16/32 bit line just like they did with the 8 bits. C900 was started under Tramiel, that was his style. And Zilog, beeing in bad shape, was probably cheap enough to buy.

When Tramiel left, obviously, the opinion to buy Amiga prevailed, thus tying the fate of the 16/32 bit line to Motorola.

There was a 32 bit 68000-like CPU project by MOS but Tramiel killed that shortly after buying MOS. Ah, the missed opportunities...

[WolfToTheMoon]

denis ritchie and AT&T examined Coherent code... it passed their scrutiny.

Z8001 had 16 x 16 bit registers, true, but instructions could "see" them as anything from 8 bit to 64 bit.

[WolfToTheMoon]

I managed to find a Sieve benchmark of Z8001 vs some other CPUs of the time...

http://www.atarimagazines.com/v4n6/STperformancetest.html

According to this, a 5,5 MHz Z8001 running Unix was 2 times faster than a ST with a 8 MHz 68000 in this particular benchmark.
So a 10 MHz Z8001 would give a very solid performance in a C900

[WolfToTheMoon]

That's not a good benchmark. It compares different thnigs at the same time. And certainly not indicative of CPU performance, unless you are ok seeing the same CPU (68k) performing 3 times faster/slower using different OSes and compilers.

That's to be expected, with different compilers. It just shows that Z-8001 is certainly very competitive to a 68000 and 80286.

BTW, I think that the prime reason for poor Macintosh performance is that the C compiler on 128k models barely ran due to little available memory. ST is much better in that regard(512k)

[WolfToTheMoon]

What if Apple had bought the Amiga rather than CBM?

Unlikely, since they had Lisa and the Mac.
But, theoretically, had they bought Amiga, they would have positioned it as a high-end workstation and charged much more than C= ever did (Macintosh II was around 10 000$ with a video color card, and it still disn't multitask)

[WolfToTheMoon]

By 1989, C= could be using Z80000,  which blows the NeXT's 68030 away.

[WolfToTheMoon]

The C900 looks like an interesting machine, the Z8001 being an early RISC CPU, albeit one with a lot of flaws.

The biggest flaw was that it was late to the market... performance was very good, once bugs were ironed out. The Z80000 had a huge potential.
And one more plus is that it was cheap... Z8000 had only about 51000 tranistors... That's some 17 000 less than the 68000 that came out the same year. Z80000 had 91 000 transistors, vs 273 000 for the 68030.

The video chip is actually the same as the C128's video chip (http://en.wikipedia.org/wiki/MOS_Technology_8563) but presumably running at a faster speed to generate the 72Hz 1024x800 display. Despite having its own blitter, it was probably not a patch on the Amiga's graphics chip.

video chip had it's own 128k of memory. Blitter was an add-on card, Haynie claims it was superior to Amiga's blitter.