AGA successor: Document about the Advanced Amiga Architecture published

[alexh]

I am an ASIC designer and I work as part of a team that design SoC chips today. When silicon comes back from the fab the first time, it takes approx 2 years before it is qualified to go on sale. There are 3 firmware teams, one developing the HAL, another the low-level test firmware sitting on the HAL to help diagnose all the ASIC bugs and devise workarounds for the HAL and another team developing the mission firmware that sits on top of the HAL. Mission Firmware would be the equivalent of AmigaOS and it doesn't run for a long time.

Even if the chips are functional enough to sell they have to go through what is called corner lot testing to ensure that they will still work at extreme ranges of temperature and voltages. That itself can take half a year. I know I sat putting samples into an oven and a freezer, wrote software to monitor and modify voltages, to find an ASIC bug.

Yes our SoCs are more complicated than AAA, but we have 10x more people.

[grond]

Originally Posted by alexh

Yes our SoCs are more complicated than AAA, but we have 10x more people.

And you have at least ten times more advanced software tools to catch problems early on. I would guess that the timeline would have been similar for Commodore to what you describe.

[Bruce Abbott]

Originally Posted by Thomas Richter

First of all, you say yourself that the software is the critical part, not the hardware, so which difference does it make?

It's pretty obvious that the most 'critical' part of a chipset is the hardware. But of course it also needs to work with the software. In our case that is Amiga OS and any games or apps we want to run. In the case of people wanting to run PC software, they need a PC - not just Windows, but the hardware that Windows runs on. Not necessarily because that hardware is more powerful, but simply that it is compatible with what Windows needs.

In 1997 I had an Amiga that was as powerful as a typical PC of the day, with an 060 CPU, 32MB RAM, RTG card and Ethernet. What it didn't have was the software required to do what I needed. Therefore I reluctantly had to use a PC instead. For most people that happened earlier. In fact most people never even considered the Amiga as an option - they were PC from day one. The name of the game back then was IBM compatibility, and that hasn't changed much today.

And why is that different from today? But even then, we *do* have such chipsets. They are just your average PC graphics cards - which can do a lot better than AAA ever could.

So which 'average PC graphics card' can I connect to my Amiga to get 'a lot better than AAA ever could'?

[Thomas Richter]

Originally Posted by Bruce Abbott

It's pretty obvious that the most 'critical' part of a chipset is the hardware.

A chipset *is* hardware, so that's not what you mean, obviously.

Originally Posted by Bruce Abbott

But of course it also needs to work with the software. In our case that is Amiga OS and any games or apps we want to run. In the case of people wanting to run PC software, they need a PC - not just Windows, but the hardware that Windows runs on.

The games of today need primarily the infrastructure windows provides. The hardware implements it. So the important part is that interface, That is DirectX and not exactly the hardware underneath. The success of a machine is the software it runs, as this is why customers buy the machine. To solve particular problems - or entertain them. Which hardware that is - why does it matter?

Originally Posted by Bruce Abbott

So which 'average PC graphics card' can I connect to my Amiga to get 'a lot better than AAA ever could'?

About every single one? Even stone-old chips like the Cirrus5434 provide better features than AAA would have offered.

[ImmortalA1000]

Originally Posted by Toffee

Oh wow this is new documentation on Triple A! Can't wait to digest, thanks for posting.

No secret we all know but they admit AA/AGA was outdated before it was released:-

The Amiga 1000 was technically demoted to 2nd place in cutting edge performance/£ by the Acorn A305/A310 Archimedes 256/4096 colour/8 true stereo channel sound and 68020 CPU speed ARM series in 1987 before the A500 even went to stores and they never really got to #1 ever again if you include console chipsets too.

AAA would never ever have made it to an A500/1200 price bracket machine, that being the bulk of Commodore's income. Ditto for Hombre. Ditto for absurd CPU successor PA RISC/MIPS etc ideas when clearly the answer was to move over to dirt cheap in terms of price performance x86 (via AMD). Apple made the right choice, pick a powerful spec x86 based off the shelf solution (CPU and graphics chips) to replace your esoteric slowly advanced hardware choice. That's why Apple are still here, they put in the effort of migrating off Motorola and onto cheaper x86 PC world options and they understood the advantage of not owning a PC wasn't the alternate CPU architecture etc it was the superior OS exclusive to their own computers.

[Turrican_3]

Originally Posted by ImmortalA1000

AAA would never ever have made it to an A500/1200 price bracket machine, that being the bulk of Commodore's income. Ditto for Hombre.

This very new document seems to claim otherwise, as there's mention of a low cost AAA system that was anticipated to be slightly more expensive than the A500 to produce (hence ending up having a comparable MSRP)

Same for Hombre, which according to Commodore's own documents had a design goal of a low cost (sub-40$), high performance chipset.

Originally Posted by Bruce Abbott

According to his AAA overview document, it consisted of "four completely new VLSI chips, implemented in high speed CMOS". AFAIK Commodore didn't have the ability to make such chips in-house, so where did they get this 'first silicon' from?

http://obligement.free.fr/articles/aaa.php

Pictures have been publicly available for a good while. It seems manufacturing was courtesy by HP, not sure if this is true for each chip though.

[pandy71]

Hilarious, not sure if this is Dave Haynie upload but it looks like reviving dead body by applying some voltage across muscle and making it contract... AAA would not change anything for Amiga, perhaps only extend a bit misery of falling Commodore...

[Matt_H]

For context, if you haven't watched it already, you need to watch Dave Haynie's video, The Deathbed Vigil. It provides a ton of insight into what Commodore engineering was trying to do versus what Commodore management would allow them to do.

AGA/AA/Pandora was never supposed to exist, but came into development as a "quick" stopgap because AAA development was taking too long (probably because it was starved of R&D resources). But even then, management dragged their feet, insisting on the development of useless new ECS machines like the 2200/2400; AGA was supposed to be available somewhere in the window of late '89 and early '91. Instead, it came out late '92, and too late to be available for the holiday shopping season in high quantities. By the time it was widely available in 1993, VGA had become immovably entrenched.

Imagine if we'd gotten the A3000+ and a 1200-like machine in 1990. VGA existed by that point, of course, but it was expensive and wasn't yet ubiquitous. Most PC games were still EGA at that point. If the original engineering timeline had held, with AGA in ~1990 and AAA in ~1993, the Amiga definitely could have remained competitive as the "obvious" (and drastically less expensive) choice well into the late 90s, or certainly long enough for a post-AAA architecture to be available. A 1280x1024 screen on AAA? That would have been a game-changer for CAD and other workstation/productivity applications. That sort of resolution didn't become common on PCs/Macs until well into the late 2000s.

In other words, if you look at AA/AAA in the contexts for which they were originally intended by engineering, e.g., the markets of ~1990 and ~1993, respectively, they're damn impressive. And if that timeline had held I think computer history would be much different and much more interesting.

[Bruce Abbott]

Originally Posted by Matt_H

Imagine if we'd gotten the A3000+ and a 1200-like machine in 1990.

Realistically this was not going to happen in less than 3 years. How do we know this? Even the biggest computer manufacturer in the world, IBM, with vast experience in chip design and far more resources than Commodore, took 3 years to develop VGA after EGA came out in 1984 - and VGA was a much simpler design than AGA.

On the software side, Commodore faced a much harder task than IBM did with VGA (which just needed a BIOS ROM to use the new screen modes). Workbench 2.0 was a major rewrite of the OS. It was released in late 1990 with the A3000. Had AGA support been necessary it probably would have taken longer.

I think the earliest AGA could have been released was late 1991. Dave Haynie says the development of AA/AGA was paused sometime in 1991 by 'human bus error' Bill Sydnes, then restarted in 1992 when his plan to introduce mid-range A3000 style Amigas failed. This is the only time that realistically could have been saved.

But let's say they did manage it in 1990. At that time the A500 was still selling well with numbers increasing, and OCS software was still getting better. It would be foolish to withdraw the A500 when demand was still high, and it would take a while for AGA to take over. So how much difference would it make? I suspect not much. OCS would still be the 'lowest common denominator' that games were made for. OTOH 1990 was the year that VGA games like Monkey Island and King's Quest V were released. Perhaps if an Amiga which was capable of running such games in 256 colors was available then it might have revitalized the Amiga games market.

VGA existed by that point, of course, but it was expensive and wasn't yet ubiquitous. Most PC games were still EGA at that point.

By 1990 EGA was getting old and VGA games were the 'latest thing'. Around this time developers made the choice to cater only to 'high-end' users. They could afford to do that due to the huge number of PCs and exponentially growing user base, with most new machines having VGA and at least a 386 CPU.

A 1280x1024 screen on AAA? That would have been a game-changer for CAD and other workstation/productivity applications. That sort of resolution didn't become common on PCs/Macs until well into the late 2000s.

I disagree. Nobody was going to buy an Amiga for workstation/productivity applications anyway, no matter how impressive the graphics were. Could you get AutoCAD on the Amiga? No. How about QuarkXPress? On the Amiga you only had unprofessionally designed and poorly supported wannabes. But even if the apps were awesome it wouldn't have mattered. AutoCAD and QuarkXPress were famous, so that's what everyone wanted. And they didn't run on the Amiga.

In other words, if you look at AA/AAA in the contexts for which they were originally intended by engineering, e.g., the markets of ~1990 and ~1993, respectively, they're damn impressive. And if that timeline had held I think computer history would be much different and much more interesting.

In Amigaland perhaps, but for computing history in general I don't think it would have made much difference. Most people had already dismissed the Amiga as just a cheap (and incompatible) gaming computer before well 1990. A new chipset wouldn't change that, whether it was released in 1990 or 1992.

Still, that doesn't stop us from imagining how awesome an AAA Amiga might have been to own in 1993. Your PC friends would be so jealous!

[Bruce Abbott]

Originally Posted by ImmortalA1000

Ditto for absurd CPU successor PA RISC/MIPS etc ideas when clearly the answer was to move over to dirt cheap in terms of price performance x86 (via AMD). Apple made the right choice, pick a powerful spec x86 based off the shelf solution (CPU and graphics chips) to replace your esoteric slowly advanced hardware choice.

Yep, that's what I was telling people when the Pentium came out.

That would have been the death of the Amiga as we know it.

But hey, what did you want a computer for? If it was playing games or running apps the PC was the obvious choice. Just make everything 'off the shelf', and keep your old Amiga to run legacy Amiga stuff if you still want to. Which is what we are doing today. Hell, today we can have an 'Amiga' more powerful than any hardware that could be produced for it using just a stock PC! Of course to do this a few chips need to be emulated...

[stefcep2]

Originally Posted by Thomas Richter

A chipset *is* hardware, so that's not what you mean, obviously. The games of today need primarily the infrastructure windows provides. The hardware implements it. So the important part is that interface, That is DirectX and not exactly the hardware underneath. The success of a machine is the software it runs, as this is why customers buy the machine. To solve particular problems - or entertain them. Which hardware that is - why does it matter?

You need both software and the hardware to run the software *well*.

[stefcep2]

Originally Posted by ImmortalA1000

The Amiga 1000 was technically demoted to 2nd place in cutting edge performance/£ by the Acorn A305/A310 Archimedes 256/4096 colour/8 true stereo channel sound and 68020 CPU speed ARM series in 1987 before the A500 even went to stores and they never really got to #1 ever again if you include console chipsets too.

AAA would never ever have made it to an A500/1200 price bracket machine, that being the bulk of Commodore's income. Ditto for Hombre. Ditto for absurd CPU successor PA RISC/MIPS etc ideas when clearly the answer was to move over to dirt cheap in terms of price performance x86 (via AMD). Apple made the right choice, pick a powerful spec x86 based off the shelf solution (CPU and graphics chips) to replace your esoteric slowly advanced hardware choice. That's why Apple are still here, they put in the effort of migrating off Motorola and onto cheaper x86 PC world options and they understood the advantage of not owning a PC wasn't the alternate CPU architecture etc it was the superior OS exclusive to their own computers.

For a long time after 680x0, Apple used PPC hardware as a selling point of difference, and this was in addition to MacOS eg the "altivec", "super computer" ads they ran.

When it became clear that PPC could not match the price/performance of x86 they moved to intel, and then as the hardware point of difference no longer existed, they focused on the OS as their point of difference.

Today with the M chips, its back to hardware and software as their point of difference.

Seems this thread has to many software guys. Hardware matters.

[Thomas Richter]

I totally disagree with that. If Apple's history tells anything then that hardware did not matter too much - as long as they could preserve their software basis, they changed the underlying hardware and added a translation layer. It is just that x86 development is at standstill, and that is because the hardware is about to collapse about its legacy.

[desiv]

Originally Posted by Thomas Richter

I totally disagree with that. If Apple's history tells anything then that hardware did not matter too much - as long as they could preserve their software basis

This!
And this is why I think Commodore's biggest mistake wasn't hardware or R&D cuts by management (which was dumb, but)...
It was (IMHO) not going after the education market and dislodging Apple.

I think the Amiga, with it's TV video compatible subsystem, was a perfect match for the education market.

Commodore needed to do what Apple did, but more. Discounts / support for the education market.
It is the market that kept Apple relevant during their down times (well, until they got into the phone/music business).

Of course, for Commodore to recognize that, they would have had to have good marketing and business awareness.
And they just wanted the next C64... Until they wanted the next PC, which was too late and not going to happen.

I'm not as worried about the application software not being there. If you get into the schools, the software will be there. The publisher's would be smart enough to follow that market...

[pandy71]

Originally Posted by stefcep2

When it became clear that PPC could not match the price/performance of x86 they moved to intel, and then as the hardware point of difference no longer existed, they focused on the OS as their point of difference.

Today with the M chips, its back to hardware and software as their point of difference.

Apple own HW allow Apple to increase profits (Intel was too greedy) - that's why Apple started all that M ARM development - it will be not surprising if Apple next step will be to ditch ARM (Apple need to pay royalties to ARM) and they start with something like RISC-V... After so many ISA changes they simply don't care... and btw they didn't care also in past - Rosetta was not 100% guarantee to run legacy software owned by people but hey, Apple don't care.

Forgot to add to be not completely OT, Miner in interviews claimed that before his departure from Commodore, Ranger chipset was ready (silicone proven) yet Commodore decided to go for ECS instead Ranger, later AAA seem to be more or less ready yet Commodore decided to go for AGA (AA) chipset - this also can be considered as one of main reasons for Commodore failure - many projects started but commercialized are those very "conservative".

And 'Advanced AMIGA Architecture' nicely shows cost of Commodore product - they had wide margin.

[ImmortalA1000]

Originally Posted by Bruce Abbott

Yep, that's what I was telling people when the Pentium came out.

That would have been the death of the Amiga as we know it.

But hey, what did you want a computer for? If it was playing games or running apps the PC was the obvious choice. Just make everything 'off the shelf', and keep your old Amiga to run legacy Amiga stuff if you still want to. Which is what we are doing today. Hell, today we can have an 'Amiga' more powerful than any hardware that could be produced for it using just a stock PC! Of course to do this a few chips need to be emulated...

Amazingly we are sort of in agreement, I must be taking Dr Gregory House's medication lol

It happened to all of them, Acorn had a nice CPU but there was no real innovation on the graphics end of their low end machines. Atari did put a DSP in the Falcon and it did magical things for musicians but it wasn't going to suddenly make Ridge Racer possible on your £600 Falcon etc.

Nothing to be ashamed of, happened to Sony too as the PS3 was highly esoteric but PS4 is not much more than a highly optimised x86-64 type system with a highly focuses UNIX core. Xbox was always a PC in disguise except for the Xbox 360 too.

Amithlon is the sort of thing Commodore should have been working towards, the 486SX CPU was a game changer for home computers, they really should have noticed that x86 had already overtaken Motorola 68xxx CPUs in bang/buck terms. It was dumb to assume there would be no blitter equipped SVGA cards, and indeed there were such things whilst Commodore was still pimping the £2000 A4000 right to the bitter end, which in reality was no faster than a £400 33mhz 486SX PC in 1994. To make things worse the Mac LC475 was about the price of an A4000/030 with the same CPU and 16 million colours of A4000/040.

Commodore were stuck, there is no evidence anything they were doing at CSG would have saved them because Commodore didn't have the financial reserves to fight, like Nolan at Atari had he not sold to Warner there never would have been the financial power to make the VCS a thing. You would need to manufacture a million Hombre chipset based motherboards in a single production run to compete with Sega/Sony in the mid 1990s. If you don't have the cash you can't take a seat at this poker game.

Maybe Commodore should have spent more effort making a success of their PC business, it worked for Amstrad in the early 1990s and Amstrad branding was more of a stigma than a bonus lol and yet you still saw Amstrad PCs in offices. They gave up on one sector to concentrate on another that was rapidly disappearing. Maybe if the 3DO chipset was made for an A1200 successor and shoved in a £500 internal CD equipped option it would be OK, there is no way AAA could ever have done anything like that and 3DO by Xmas 1994 is the minimum quality consumers would be interested in.

There was some cool stuff in AAA, something to do with real-time video etc but this is no use for a mass market machine to keep Commodore afloat, they needed RJ Mical & Dave Needle levels of engineering talent not to go bust in 1994, simple as that. Dave Haynie and co at CSG were no Dave Needle/RJ combo of genius that's for sure

[TEG]

Originally Posted by ImmortalA1000

Commodore were stuck, there is no evidence anything they were doing at CSG would have saved them because Commodore didn't have the financial reserves to fight, like Nolan at Atari had he not sold to Warner there never would have been the financial power to make the VCS a thing. You would need to manufacture a million Hombre chipset based motherboards in a single production run to compete with Sega/Sony in the mid 1990s. If you don't have the cash you can't take a seat at this poker game.

As someone pointed out in the other thread, Gould did not wanted to introduce Commodore in the stock market. So Commodore missed the opportunity to leverage money at the time they were N°1 computers seller with the C64. Is it because it would expose too much the books? We don't know, he wanted debt.

It would have allow the modernization of the MOS plant, and the 6502. 2020 Apple of the time. Note that the improvement of the 6502 was finally done by others and was widely used: 6502 Computers and Games, 6502 Variations and derivatives.

[stefcep2]

Originally Posted by Thomas Richter

I totally disagree with that. If Apple's history tells anything then that hardware did not matter too much - as long as they could preserve their software basis, they changed the underlying hardware and added a translation layer. It is just that x86 development is at standstill, and that is because the hardware is about to collapse about its legacy.

Well you can totally disagree, but the reality is that's what Apple actually *did*.

Hell, their entire marketing philosophy was literally that: Think Different! PPC was part of that.

And it worked- many Mac users argued and genuinely believed that PPC was one component that gave them better performance and experience.

[stefcep2]

Originally Posted by pandy71

Apple own HW allow Apple to increase profits (Intel was too greedy) - that's why Apple started all that M ARM development - it will be not surprising if Apple next step will be to ditch ARM (Apple need to pay royalties to ARM) and they start with something like RISC-V... After so many ISA changes they simply don't care... and btw they didn't care also in past - Rosetta was not 100% guarantee to run legacy software owned by people but hey, Apple don't care.

.

Ultimately every business aims or should for "vertical integration". ie make it yourself and sell it yourself. Cut out the middle man. Nothing unique from Apple.

[Bruce Abbott]

Originally Posted by desiv

This!
And this is why I think Commodore's biggest mistake wasn't hardware or R&D cuts by management (which was dumb, but)...
It was (IMHO) not going after the education market and dislodging Apple.

It's harder to dislodge than to get in first. In fact Commodore was there in 1980 with the PET, but the Apple II had color and was far more popular. By the time the Amiga came out Apple was entrenched in the educational market.

I think the Amiga, with it's TV video compatible subsystem, was a perfect match for the education market.

Why?

Commodore needed to do what Apple did, but more. Discounts / support for the education market.

A market only big enough for one player will be bad for 2 players. To dislodge Apple, Commodore would have to sink a large amount of money in it - for little benefit.

It is the market that kept Apple relevant during their down times (well, until they got into the phone/music business).

The education market was going PC because PCs were ubiquitous and more relevant. The Mac was only 'relevant' in higher education. It wasn't enough to carry them. Apple got into the phone business in 1997. Jobs said at that time they were 2 weeks from insolvency. In 1996 their net income was -$816 million. In 1997 it was -$1.045 billion.

Of course, for Commodore to recognize that, they would have had to have good marketing and business awareness.
And they just wanted the next C64... Until they wanted the next PC, which was too late and not going to happen.

Commodore didn't seem to recognize that PCs had the 'personal' computer market sewn up, and this would never change. The C64 and A500 were successful because they filled a niche PCs couldn't. They could have cultivated that niche more than they did, but ultimately they would have to accept a shrinking market.

I'm not as worried about the application software not being there. If you get into the schools, the software will be there. The publisher's would be smart enough to follow that market...

The quality of educational software for schools was generally very poor, and there wasn't much money to be made from it. The real market was home users. But most developers were more interested in producing games or business apps.

Between 1991 and 1994 I was a developer in a small educational software company. Like many it was run by a teacher who developed his own software for classroom use, then attempted to market it. I wish I had started working there earlier when I could have had more impact.