Some thoughts on why the Amiga didn't thrive

[Bruce Abbott]

Originally Posted by duga

Well, everyone wants more speed. By adding fast mem you more than double the speed of the system.

Only if the 'fast' RAM is separated from the ChipRAM bus. This was why the trapdoor memory in an OCS A500 is called 'slow' memory - it's connected to the ChipRAM bus so it is constrained to the same speed.

With 2 MB on the board and some kind of "fast mem controller", you could configure the system (via physical jumper, or via Early Startup Menu) to 2 + 0 MB ("graphics mode") or 1 + 1 MB ("fast mode"). Later, when you add a memory/accelerator board with fast mem you of course set the jumper to 2 + 0 config.

1 jumper would not be sufficient. The RAM would have to be connected to the bus via 2 sets of buffers, one for use as ChipRAM and another for FastRAM. This would make the circuit more complex and take up extra room on the motherboard - for little benefit because when you have the AGA chipset you want as much ChipRAM as you can get!

[Bruce Abbott]

Originally Posted by Foebane

Which looks JUST as UNRELIABLE to me. You said PLACE, not even LOCK or SECURE, meaning that it could pop off again during use and fry both CPUs! MADNESS!!

Yep, if you just 'place' it over the CPU it can be unreliable (had that problem with the Viper CPU board in my A600). The Vampire came with screws that clamp the board down to the motherboard. The socket was so tight that I had to remove the motherboard from the case and press the two boards hard together with packing under the CPU. Once this was done the Vampire stayed well connected.

Of course with any socket there is always the possibility of a bad connection developing due to contact corrosion, and PLCC sockets are particularly bad for it because they only have one contact point.

Commodore rightly decided to solder the CPU onto the motherboard. They could have provided separate pin headers to add a CPU board, but this would have raised the cost and gotten in the way of the hard drive cradle. Looking at the motherboard we see that there was barely enough room for the CPU as it was, and placing extra connectors around it would have been difficult if not impossible (you have to have designed a few PCBs yourself to really appreciate how hard it can be to get such a compact layout, especially with the larger clearances that were required on boards of that era).

BTW around this time PC motherboards also tended to have soldered in CPUs with no socket for adding a more advanced type. Users were expected to buy a new motherboard or even a whole new computer when they wanted a faster machine.

[grond]

I still don't see any convincing reason the A600 had to be so small that you could only fit ridiculously expensive laptop hdds (into a desktop computer!) and couldn't fit a full featured expansion slot that would have allowed proper accelerators and fastmem. It is understandable that Commodore continued to sell the A500 after the A3000 was introduced and started putting ECS chips into them. It is also understandable that they eventually turned the fact that it was cheaper to stop producing the old chips into a something that could be advertised as an update of the A500 (the A500+). But the A600 was just wrong on so many levels that it just didn't make any sense at all. It's like they added an ashtray and a cup holder to the Volkswagen Beetle after thirty years of production and expected people to be excited about it and buy it at twice the price.

[Foebane]

@Grond:

Why would you add an accelerator to an ECS machine? Best leave that stuff for AGA.

[Solo Kazuki]

Originally Posted by Lord Aga

Best graph ever. Totally readable.
Must have been made on a superior PC machine.

This is (more or less) picture of real sales, but overall. Some people tend to forget that for industry and business almost no one get other computer solutions than XT/AT clones. So it looks like this, different may be sales for private/home usage.

[grond]

Originally Posted by Foebane

@Grond:

Why would you add an accelerator to an ECS machine? Best leave that stuff for AGA.

AGA didn't exist when the A600 was designed. And ECS added productivity modes which should have been accompanied by a wider chipmem bus and a faster CPU just as in the A3000.

[Solo Kazuki]

Originally Posted by grond

AGA didn't exist when the A600 was designed. And ECS added productivity modes which should have been accompanied by a wider chipmem bus and a faster CPU just as in the A3000.

AGA chipset was almost complete when A3000 was released, there was A3000 prototype with AGA chipset (and additional DSP). That was CEO (Ali M.) decision to not release AGA hardware yet until 1992 year.

[Bruce Abbott]

Originally Posted by grond

I still don't see any convincing reason the A600 had to be so small that you could only fit ridiculously expensive laptop hdds (into a desktop computer!)

2.5" hard drives were not 'ridiculously' expensive. They cost a bit than typical 3.5" and 5.25" drives, but then so did SCSI drives (which Amiga users were used to having to pay extra for).

One popular IDE hard drive introduced in 1996 was the Quantum Bigfoot. Why was it popular? Because it was cheap. Should Commodore have made the A600 big enough to take this and other 5.25" drives? If so then we are looking at something significantly larger than an A500, certainly not an 'all in the keyboard case' design.

The smaller the motherboard and case the cheaper it should be to manufacture. Of course as we know the A600 wasn't cheaper to produce than the A500 (that's what happens when you manufacture stuff in the UK rather the far East) but had it been bigger it would have been even more expensive.

Another factor to consider is that 2.5" drives generally use much less power than larger drives, so the power supply doesn't have to be as large. The A600's PSU was significantly smaller and lower wattage than the A500's 'brick' power supply (which should have made that cheaper too).

...and couldn't fit a full featured expansion slot that would have allowed proper accelerators and fastmem.

The idea was the A600 would be a 'condensed' version of the A500 designed to match the specs of a typical A500 system but at a lower price. IOW it was supposed to be an entry level system for people who didn't expect to expand it (just like most A500 users didn't expand their machines other than adding trapdoor RAM). Originally it was going be called the A300, which gives you an idea of where they expected it to be in the new lineup. Perhaps if they had stuck with that name Amiga fans might have had fewer complaints!

It is understandable that Commodore continued to sell the A500 after the A3000 was introduced...

Especially since the A3000 was ridiculously overpriced (and I say that as a 'first adopter'). Not only that but the case design was poor too - it had no 5.25" drive bay so you couldn't install an internal CROM drive, a very noisy power supply fan, and barely enough room to fit an accelerator card (I had to cut hole in the drive tray to get a fan onto my Blizzard 060 board). Considering the A3000's flaws it's surprising that it that it doesn't get criticized more.

So of course they continued to make the A500 and A2000. If they hadn't sales would have ground to a halt very quickly.

It is also understandable that they eventually turned the fact that it was cheaper to stop producing the old chips into a something that could be advertised as an update of the A500 (the A500+).

I remember the day I received the first shipment of A500+'s in my shop. Despite getting regular sales information from Commodore NZ, I had no clue that this new model was coming until I saw the logo. We had ordered regular A500's and these were what we got!

But the A600 was just wrong on so many levels that it just didn't make any sense at all. It's like they added an ashtray and a cup holder to the Volkswagen Beetle after thirty years of production and expected people to be excited about it and buy it at twice the price.

The A600 was not twice the price of an A500+ (at least not in NZ anyway), and the IDE hard drive and PCMCIA slot were a lot more than 'an ashtray and a cup holder'.

But to take your analogy further, the new VW beetle was pretty much as you describe - still a 2 door, still the same size, still not enough room in the engine bay for a V8 engine - but it does have cup holders! So why would anyone buy one?

[Bruce Abbott]

Originally Posted by Solo Kazuki

AGA chipset was almost complete when A3000 was released, there was A3000 prototype with AGA chipset (and additional DSP). That was CEO (Ali M.) decision to not release AGA hardware yet until 1992 year.

If an AGA A3000 had been released in 1991, existing A3000 owners like me would have been pissed!

But it wasn't going to happen. Schematics for the prototype are dated 15th November 1991, which means there was no way it would be released until at least mid 1992. If it came out at a similar price to the original A3000 it would have bombed. The DSP chip would have been an orphan, and the case design was not appropriate for a machine of this class.

So what did Commodore do instead? They released the A4000 - which is what the 'AGA A3000' should have been.

[grond]

Originally Posted by Bruce Abbott

2.5" hard drives were not 'ridiculously' expensive. They cost a bit than typical 3.5" and 5.25" drives, but then so did SCSI drives (which Amiga users were used to having to pay extra for).

We had that already. 2.5" hdds cost about 2.5x as much as they same capacity in a 3.5" form factor.

One popular IDE hard drive introduced in 1996 was the Quantum Bigfoot. Why was it popular? Because it was cheap. Should Commodore have made the A600 big enough to take this and other 5.25" drives?

By 1996 Commodore should have sold 040-based Amigas as minimum configuration. How should a supposed low-cost A600 be relevant in 1996?



And also you can't invalidate my point about the needlessly small "low-cost" A600 that required needlessly expensive 2.5" harddisks by taking my argument to another extreme. There was meaningful middle ground and that was 3.5" harddisks.

Of course as we know the A600 wasn't cheaper to produce than the A500 (that's what happens when you manufacture stuff in the UK rather the far East) but had it been bigger it would have been even more expensive.

Why would it be more expensive? Just because it was more PCB space? It could have had less layers, would have been less densely packed and thus probably less expensive.

Another factor to consider is that 2.5" drives generally use much less power than larger drives, so the power supply doesn't have to be as large. The A600's PSU was significantly smaller and lower wattage than the A500's 'brick' power supply (which should have made that cheaper too).

I already addressed that before: Commodore saved a dollar on the PSU and expected customers to spend hundreds more on a ridiculously expensive 2.5" harddisk. That was typical of them and hurt them in the long run. Much in the same way they soldered a 68020 onto the A1200 PCB because it would have cost them a dollar to put it on a daughterboard and offer every possible combination from 020 without fastmem up to 50 MHz 030s. Every 030+ user of the A1200 thus paid for a useless 020 CPU but at least Commodore saved a dollar.

The idea was the A600 would be a 'condensed' version of the A500 designed to match the specs of a typical A500 system but at a lower price.

That may have made sense but then they should have dropped the PCMCIA and IDE stuff altogether.

Perhaps if they had stuck with that name Amiga fans might have had fewer complaints!

Commodore clearly didn't care about Amiga fans, otherwise they wouldn't have tried to sell the same stuff all over again. There was very little reason for anyone who already owned an Amiga to buy an A600.

Considering the A3000's flaws it's surprising that it that it doesn't get criticized more.

Probably because it was a dream machine that few people ever owned.

The A600 was not twice the price of an A500+ (at least not in NZ anyway),

AFAIR the A600 started at 1100DM which was precisely what an A500 cost in 1987, five years earlier when it was still new and powerful technology. And the only thing the A600 offered over the A500 was the ability to add very expensive harddisks and very expensive and almost nonexistent PCMCIA expansions.

[activist]

Originally Posted by grond

How should a supposed low-cost A600 be relevant in 1996?

the design brief for the A600 was something to replace the C64. The 64 was still in production in 1994 as you know.

Originally Posted by grond

And also you can't invalidate my point about the needlessly small "low-cost" A600

it was designed small so as to appear 'console like' and cater for the C64 market. They messed up with the low cost part when the spec changed from the original A300 brief. I think the A600 design and look has held up well actually.

Originally Posted by grond

I already addressed that before: Commodore saved a dollar on the PSU and expected customers to spend hundreds more on a ridiculously expensive 2.5" harddisk.

just scope creep unfortunately. There was never supposed to be a hard drive in it. Good to have today of course. And PCMCIA. But that's just pot luck.

Originally Posted by grond

There was very little reason for anyone who already owned an Amiga to buy an A600.

anyone except those who owned a C64. Then you might be interested in the the type of machine the A600 was. But not at the launch retail price.

Originally Posted by grond

AFAIR the A600 started at 1100DM which was precisely what an A500 cost in 1987, five years earlier when it was still new and powerful technology. And the only thing the A600 offered over the A500 was the ability to add very expensive harddisks and very expensive and almost nonexistent PCMCIA expansions.

the design was fine for the intended C64 market. But again scope creep pushed the cost out of reason.

[AmigaHope]

Originally Posted by Bruce Abbott

If an AGA A3000 had been released in 1991, existing A3000 owners like me would have been pissed!

But it wasn't going to happen. Schematics for the prototype are dated 15th November 1991, which means there was no way it would be released until at least mid 1992. If it came out at a similar price to the original A3000 it would have bombed. The DSP chip would have been an orphan, and the case design was not appropriate for a machine of this class.

So what did Commodore do instead? They released the A4000 - which is what the 'AGA A3000' should have been.

They should have never made ECS in the first place, and started work on a better chipset (something like AGA but with a few more improvements) from day one. Unfortunately Commodore was slow to understand that the computer market was changing and you couldn't just sell the same spec for years like they had been doing with the C64.

It was ironically the same mistake Atari made with game consoles years earlier. They thought they could just sell the Atari 2600 forever, and when it stopped selling they thought that people were just tired of video games.

[TEG]

On the A600, here is the inside view of David Pleasance, this what activist described:

- David Pleasance: "It isn't I particularly dislike the A600 for what it is, rather I hate what it did to Commodore as it was wrong in every possible way.

My initial idea was to provide a purpose-built product to fill the gap that the C64 left and to attempt to reach a price point low enough for the socio-economic population who could not afford an Amiga 500 - so I asked for a 'cut down' (cost reduced) Amiga that we could sell for a maximum price of £249 to get the basic machine.

But this product could be expanded as and when users could afford it. I specifically requested we call it the Amiga 300 - so there was no misunderstanding as to where this model fitted in the line-up.

I have no idea who ultimately changed the model number to A600. The first several thousand A600's that were built had A300 on the motherboard.

The last straw was of course Bill Sydnes who was head of engineering at the time. He added some specs and it ended up costing more to make than the A500, so we shot ourselves in the foot.

A team of professional saboteurs could not have done a better job."

Source: [ Show youtube player ]

About Bill Sydnes:
(1981)

‘Skirt the Bureaucracy’

The only way to do this, according to William Lowe, was to skirt the IBM bureaucracy. He had two proposals. The first was to buy a microcomputer company outright (he mentioned Atari by name); the other was to create a brand new microcomputer.

The plan was submitted to IBM’s Corporate Management Committee (CMC), and it was enthusiastically received. Microcomputers were gaining a foothold in the business world, so CMC asked that a prototype be built for review in a month’s time. A small team was established in the ELS IBU to create the prototype.

Bill Sydnes, the IBM 5120 manager, was selected to lead the hardware engineering team working on the prototype and was given a staff of 12 engineers. At the time, IBM had no microprocessor ready for release (the predecessor to POWER, the IBM 801, would not be ready for almost two years), so Sydnes decided to use the Intel 8088 processor. Though it was a 16-bit processor (meaning that it could handle larger numbers than other microcomputers), it used an older bus design that slowed it down, unlike the more expensive 8086.

Sydnes noted in the CMC review that other open and processing standards would be adopted to keep down costs.

IBM System 23Jack Sams, the engineer in charge of software development for the prototype, was eager to avoid the problems he encountered during the development of the IBM System/23. The System/23 was a very small minicomputer used mainly for word processing and basic business management.

Source: Origin of the IBM PC

From amigahistory.plus.com:

Bill Sydnes- A former manager at IBM who was responsible for the stripped down PCjr. He was hired by Commodore in 1991 to repeat that success with the A600. However, at the time the Amiga was already at the low-end of the market and a smaller version of the A500 was not needed.

Source

So an IBM guy was implied in the final A600 design.

[Foebane]

This Bill Sydnes guy sounds like a REAL asshole, down there with Mehdi Ali and Irving Gould!

[grond]

Originally Posted by activist

the design brief for the A600 was something to replace the C64. The 64 was still in production in 1994 as you know.

If one thinks it wise to move the Amiga into the C64-market, then an A300 without IDE and PCMCIA would have made sense. However, that essentially meant giving up on the idea of staying competitive tech-wise (Apple never gave up on this idea even though throughout most of their 68k and PPC era they mostly had bold claims about it). To me that was the cardinal sin. Amigans were willing to spend money to keep their Amigas at a competitive hardware level in the 1st half of the 90s but could only spend that money at 3rd party companies and got software in reward that still only targeted the old 68000/OCS configuration.

There was never supposed to be a hard drive in it. Good to have today of course. And PCMCIA. But that's just pot luck.

Indeed. Today the A600 is a very attractive package IMO only superseded by the A1200.

[Minuous]

I'd disagree about a "smaller version of the A500 was not needed". The A500 was grossly oversized.

[AmigaHope]

Originally Posted by Minuous

I'd disagree about a "smaller version of the A500 was not needed". The A500 was grossly oversized.

The biggest problem was that Commodore's legendary vertical integration is part of what allowed them to produce computers with custom silicon so cheaply. They owned their own fab.

Except they had neglected keeping their fab updated with new process nodes, so they couldn't miniaturize their hardware themselves. If they had designed an Amiga-on-a-chip to make it cheaper to build, they could not have made that chip themselves -- and buying silicon from independent fabs was nowhere near as cheap as it is today.

That's part of why AGA was almost identical to OCS. They put almost all the improvements into Lisa, which meant only one chip required a process node they could not produce themselves, which would be outsourced to HP. The other chips were only tweaked to support the new features, hence Alice being almost identical to Agnus -- with the same old blitter almost unchanged from 1985, same copper, etc. And of course Paula was famously completely unchanged.

Most of the miniaturization in the A600 came from replacing the off-the-shelf parts of the A500 with smaller equivalents, and putting their existing silicon into surface mount packaging. This did not decrease costs because the surface mount packaging was actually more expensive, and the silicon wafers themselves were the exact same ones used in the A500. They saved a bit of money with smaller PCBs, but PCBs were only a tiny fraction of the cost of production.

[Foebane]

Originally Posted by AmigaHope

That's part of why AGA was almost identical to OCS. They put almost all the improvements into Lisa, which meant only one chip required a process node they could not produce themselves, which would be outsourced to HP. The other chips were only tweaked to support the new features, hence Alice being almost identical to Agnus -- with the same old blitter almost unchanged from 1985, same copper, etc. And of course Paula was famously completely unchanged.

And here, all this time, I was wondering if the reason they curtailed AGA so much was because of running out of time and/or money during development, not because of something like "process nodes".

What IS a process node, anyway?

I thought it was easy to manufacture silicon chips back then, Commodore bought a chip fab company to do just that!

[Bruce Abbott]

Originally Posted by Foebane

This Bill Sydnes guy sounds like a REAL asshole, down there with Mehdi Ali and Irving Gould!

Yep, that REAL arsehole who invented the PC, the bane of every other computer platform since! If it wasn't for this arsehole the Amiga would have thrived!

What IS a process node, anyway?

I thought it was easy to manufacture silicon chips back then, Commodore bought a chip fab company to do just that!

Manufacturing silicon chips has never been easy, which is why Commodore had to buy a chip fab company to do it.

Process Node

The technology node (also process node, process technology or simply node) refers to a specific semiconductor manufacturing process and its design rules. Different nodes often imply different circuit generations and architectures. Generally, the smaller the technology node means the smaller the feature size, producing smaller transistors which are both faster and more power-efficient. Historically, the process node name referred to a number of different features of a transistor including the gate length as well as M1 half-pitch.

Producing chips with good yield becomes increasingly difficult as the transistor size reduces. Each process generally requires a dedicated production line, which costs millions of dollars to set up. The original Amiga chipset was fabricated in a 5 µm manufacturing process. According to Wikipedia CSG was able to do 1 µm in 1994. In 1993 Intel was using 0.5 µm for their Pentium CPU.

So in 1991 (when the AGA chipset was designed) they probably could have done better than 5 µm. But that would have required a complete redesign of the custom chips to use the new process, which would have cost more and further delayed its introduction. It might also have introduced significant incompatibilities with earlier chipsets, making the A1200 even less attractive.

[Foebane]

Thanks for the explanation, Bruce!