Commodore Amiga in the 1990s: AAA, AGA and the race against the PC

Commodore entered the 1990s with a problem many computer companies would have loved to have. The Amiga was technically distinctive, it had a recognisable place in the market and its custom graphics and sound hardware still gave it capabilities that ordinary PCs had spent years trying to catch. But the gap was closing. Fast. PC graphics were improving. Sound cards were spreading. CD-ROM was becoming a serious part of the computer market. Processor performance was climbing, and new graphics hardware was appearing at a pace that made long development cycles increasingly dangerous. Commodore had several answers. That was part of the problem. There was the Amiga 3000, the ambitious AAA chipset, the CDTV multimedia system, the more practical AGA graphics architecture and, later, the completely different Hombre project. Each represented a possible direction for the Amiga. Some overlapped. Others replaced earlier ideas. Several never made it into production.

The PC was catching up

The original Amiga architecture had one major advantage: custom chips. Graphics, audio and other jobs didn’t have to rely entirely on the main processor. That gave Amiga systems capabilities that stood out strongly in the second half of the 1980s, especially in games, graphics and video work. The same approach created a problem, though.

Once you’ve built a software platform around specialised hardware, replacing that hardware becomes difficult. Programs may rely on specific chip behaviour. Developers may use shortcuts that work perfectly on one generation but cause trouble on the next. Improving the machine isn’t simply a matter of installing a faster processor.

Commodore knew it needed another big jump. That project became AAA, the Advanced Amiga Architecture. The goal wasn’t to catch VGA. Commodore wanted to beat it.

AAA was supposed to provide enough headroom for several years of future Amigas, covering everything from relatively affordable machines to high-end systems. Instead of designing one new computer, Commodore was effectively trying to create the technical foundation for an entire generation. On paper, it was a serious upgrade.

AAA was much more than better Amiga graphics

AAA often gets mentioned alongside AGA, but the two shouldn’t be confused. AGA eventually shipped in machines such as the Amiga 1200 and Amiga 4000. AAA was a far larger redesign. Commodore planned a completely new group of custom chips with names including Andrea, Linda, Monica and Mary. Together they would handle graphics, memory access, audio and peripheral functions. The specifications were ambitious.

AAA was designed for 24-bit colour, much higher display resolutions and far more memory bandwidth than ECS. It could use standard DRAM or VRAM. The graphics architecture could run with wider buses, while some configurations could double up parts of the chipset to produce a more powerful 64-bit graphics path.

Audio was getting a major upgrade too, moving from the familiar four 8-bit Amiga channels to eight 16-bit channels with stereo positioning. There were also plans for higher-end video capabilities and substantially more flexible graphics modes. This wasn’t Commodore polishing the old chipset. It was trying to replace the foundation underneath it. That raised an awkward question: how much compatibility should the company keep?

Compatibility was becoming a burden

Backwards compatibility sounds simple until engineers have to build it. The Amiga had accumulated years of software designed around OCS and ECS hardware. Some developers had also relied on undocumented operating-system behaviour or talked directly to the hardware when it suited them. That works well until the hardware changes. With AAA, Commodore started considering a cleaner break. Instead of forcing the new architecture to behave exactly like older Amigas, compatibility could potentially be handled through additional hardware.

It was a different philosophy. The idea wasn’t to pretend that nothing had changed. Commodore wanted room to build something substantially faster without carrying every old technical restriction into the new design. There were several development projects around AAA. Nyx acted as a prototype platform. Another modular design allowed engineers to experiment with different parts of the system, including the processor.

Early plans still involved Motorola 68000-family CPUs such as the 68040. Engineers were also looking at RISC processors.That tells you how broad the project had become. And broad projects need people.

Commodore didn’t have enough engineers on AAA

AAA needed roughly a dozen engineers if Commodore wanted to keep the work moving properly. It didn’t get them. The project operated with about half that number at points, and some of those engineers weren’t working on AAA full-time. They had other jobs inside Commodore too. That’s a bad setup for any major semiconductor project.

Chip design already moves slowly. Split the engineering team between several products and it gets slower still. Meanwhile, the technology you’re competing against doesn’t stop. Commodore also had another issue. Its own semiconductor operation was struggling to keep pace with newer chip-manufacturing processes. So the company wasn’t just asking whether it could design AAA. It also had to ask where and how it would manufacture the chips.

All of this happened while Commodore was cutting research and development spending. The result was predictable. AAA slipped. Then it slipped again. And every delay made the original specification slightly less impressive compared with whatever was appearing in the PC market.

The target kept moving

This was one of Commodore’s biggest technical headaches. When AAA started, beating VGA looked like a worthwhile long-term goal. But PC graphics weren’t standing still. Resolutions increased. Graphics cards improved. Processor performance rose. A design that looked comfortably ahead in 1988 or 1989 could look much less dramatic a few years later.

That forced Commodore into an unpleasant choice. It could keep changing AAA to match the new market, making an already late project even later. Or it could freeze the specification and risk releasing expensive new silicon that no longer delivered the leap originally promised. Neither option was attractive.

At the same time, Commodore couldn’t ignore the machines already on sale. It still needed new Amigas customers could actually buy. That brought the Amiga 3000 into the picture.

The Amiga 3000 pushed the platform upwards

The Amiga 3000 was a much more conventional product than AAA. It was a real computer, ready for a real market. The machine used a Motorola 68030 and offered more memory capacity, built-in SCSI, ECS graphics, Zorro III expansion and AmigaOS 2.0. Compared with the Amiga 2000, it was a clear move towards a faster and more professional desktop platform.

But even here, Commodore ran into compatibility trouble. Some existing Amiga programs had been written around undocumented software interfaces. AmigaOS 2.0 changed enough that those shortcuts could break. It wasn’t the operating system’s only problem, and it wasn’t unique to Commodore. This kind of thing happens whenever developers build software around behaviour that was never guaranteed.

Still, it added friction at exactly the point when Commodore needed the transition to feel straightforward. The A3000 also arrived while the computer market was starting to talk constantly about multimedia. That word would shape Commodore’s next big experiment.

CDTV tried to take the Amiga out of the computer room

Commodore could see where the industry was going. CD-ROM offered far more storage than floppy disks. That meant bigger games, more audio, reference software, educational titles and video-heavy applications. One option was obvious: sell CD-ROM upgrades for existing Amigas. Commodore went further. It created CDTV.

The idea was unusual. Instead of selling another Amiga computer, Commodore put Amiga technology into something that looked more like a hi-fi component. It was designed for the living room, not the traditional computer desk. You didn’t need to think of it as an Amiga at all. CDTV could play audio CDs and run software from CD-ROM, including games, encyclopaedias and educational programs. Commodore wanted it to sit somewhere between a computer and a home-entertainment system. The concept made sense. The execution was harder.

CDTV was expensive for what it offered

CDTV reached the market at $999. Philips was selling CD-i for $799. That put Commodore in a difficult position from the start. And if you wanted to turn CDTV into something closer to a normal computer, adding a monitor, keyboard, mouse and floppy drive pushed the total price much higher.

The software library wasn’t ready either. Commodore had wanted around 100 titles at launch. It had roughly half that. There wasn’t one obvious piece of software that made people feel they had to own the machine.

So CDTV landed in an awkward middle ground. It was expensive compared with consumer electronics, yet someone who wanted a real computer could simply buy a computer. The technology wasn’t the only issue. The product itself was difficult to explain. Was it a games machine? A CD player? An educational box? A computer that didn’t want to call itself a computer? Commodore never found a convincing enough answer.

The Amiga roadmap started changing shape

While CDTV was struggling, AAA was getting later. Commodore needed something that could ship sooner. By 1991, management reviewed the roadmap and suspended AAA development while it worked out what to do next. A new plan emerged. Existing technology would receive short-term upgrades. The next practical generation would use the AA chipset, later marketed as AGA. A more radical architecture could follow afterwards.

In other words, Commodore stopped waiting for AAA to become the immediate successor to ECS. AGA became the bridge. That decision shaped the Amiga machines most people actually saw in the early 1990s.

AGA wasn’t AAA

This distinction matters. AGA improved the existing Amiga graphics architecture. It increased the available colour depth and offered better display modes, but it remained closely related to the hardware that had come before. AAA was supposed to go much further.

So the Amiga 1200 and Amiga 4000 shouldn’t be viewed as cut-down AAA systems. They came from a different branch of Commodore’s plans. AGA was the technology Commodore could realistically get into shops. And Commodore badly needed something new in shops. The problem was that the product line between ECS and AGA became messy. The Amiga 600 is the clearest example.

The Amiga 600 arrived at the wrong moment

The original idea behind Commodore’s new low-end machine sounded reasonable. Take the Amiga 500 concept, reduce the cost, integrate a few useful features and sell it for less. That’s not quite what happened. The finished Amiga 600 was smaller, used ECS graphics and added PCMCIA expansion, but it ended up costing more than the A500 it was meant to replace.

That created an obvious sales problem. Why buy the newer machine if it isn’t clearly faster and costs more? Worse, people already knew better graphics hardware was coming. AGA had become the interesting upgrade.

So the A600 landed in the uncomfortable gap between two generations. It was newer than the A500, but technologically it didn’t look like the machine buyers were waiting for. Commodore then discontinued the older A500 line, removing the cheaper alternative. The result was a product range that didn’t line up neatly with what customers wanted.

Other planned Amigas disappeared as well

The A600 wasn’t the only sign of a changing roadmap. Commodore had considered an Amiga 3000 Plus with AGA graphics, SCSI, CD-ROM support and a digital signal processor. There was also an Amiga 1000 Plus concept, a lower-profile machine with fewer expansion slots and a price target around $1,000. Neither became a mainstream production model.

The A1000 Plus idea changed into another proposed system called the A2200, and its specification moved backwards from AGA to ECS. That wasn’t popular with Commodore’s European operation.

From their perspective, selling another ECS machine when the market was waiting for AGA made little sense. The product was dropped. Commodore was now improvising. Eventually, two AGA systems would matter far more: the Amiga 1200 and Amiga 4000.

The Amiga 1200 was the machine buyers had been waiting for

The A1200 finally brought AGA into the mainstream Amiga format. It kept the familiar keyboard-computer design of the A500 generation but added a Motorola 68EC020 processor and the new graphics chipset. This was a much easier upgrade to explain. Better graphics. Faster processor. Familiar format. Demand was there. Supply wasn’t.

Commodore had moved manufacturing from Hong Kong to the Philippines as part of its cost-cutting effort. Then Mount Pinatubo erupted, disrupting production. That part wasn’t something the company could reasonably plan around. Other problems were. The logistics infrastructure around Manila wasn’t ideal for the scale Commodore needed. The company later turned to outside manufacturing in Scotland. Then came another problem: AGA chip supply.

Commodore had too few A1200s and too many A600s

This is where the product planning, manufacturing and marketing problems collided. Commodore couldn’t get enough AGA chips for the number of Amiga 1200s it wanted to sell during the crucial 1992 Christmas season. Only around 70,000 A1200s could be produced for that period.

At the same time, Commodore had large numbers of Amiga 600s. But buyers had already been told AGA was the future. So people wanted the machine Commodore couldn’t supply in sufficient numbers, while warehouses held machines the market was less interested in.

That’s a terrible position for a hardware company. The unsold A600 inventory put pressure on cash. Manufacturing problems then made it harder to increase A1200 production. Each issue made the next one worse.

The Amiga 4000 delivered AGA at the top of the range

Commodore also needed a high-end AGA machine. That became the Amiga 4000. It brought the new graphics chipset into a desktop system and provided faster processor options. It also switched from the A3000’s built-in SCSI to IDE.

The A4000 looked similar in broad concept to the earlier A3000 Plus idea, but some planned features had disappeared. The dedicated DSP was gone. SCSI was no longer standard. Still, AGA had finally reached both ends of the Amiga range.

The A1200 covered the mainstream home market. The A4000 handled the higher end. For the short term, Commodore had an answer. For the long term, it still didn’t. AGA could extend the Amiga architecture, but it wasn’t the huge technical reset AAA had once promised. By then, Commodore was thinking about something else entirely.

Hombre was a very different future

Hombre appeared around 1993. It wasn’t just another Amiga chipset. The thinking had changed. AAA had tried to build a dramatically more powerful Amiga while still considering ways to handle compatibility with older systems. Hombre was much less interested in carrying the traditional Amiga architecture forward.

The proposed system used a 64-bit multimedia design with hardware aimed at 3D graphics. Games were a major target. One idea was a CD-ROM-based console, sometimes referred to as CD64, with performance goals aimed at the emerging generation of 3D games machines.

Another idea was a PCI graphics accelerator that could plug into a PC or workstation. That was a big departure for Commodore. Instead of assuming the future had to be a self-contained Amiga computer, Hombre could become technology inside somebody else’s machine.

PCI changed the thinking

PCI was becoming an important industry standard. Commodore recognised what that meant. A graphics accelerator built for PCI didn’t depend on a proprietary Amiga expansion system. It could potentially reach PCs and workstations built by other manufacturers. Hombre therefore pointed towards a much more open hardware model.

The processor plan changed as well. Commodore evaluated several embedded RISC processors and selected a PA-RISC design that allowed custom logic to be combined with the processor core. Windows NT was being considered as one of the target operating systems. Even AmigaOS wasn’t guaranteed.

That tells you just how far Hombre had moved from the original concept of the Amiga. It wasn’t necessarily the next Amiga. It was a possible next Commodore multimedia architecture.

The Motorola 68000 family was no longer enough

Commodore wasn’t making these decisions in isolation. The Motorola 68000 family had powered a huge range of computers through the 1980s. Amigas used it. Apple used it. Workstations used it. But the wider industry was changing.

RISC processors were gaining ground. Silicon Graphics had MIPS. Sun had SPARC. Acorn had ARM. Motorola itself was pushing the 88000. Floating-point performance was becoming more important, especially for graphics and technical applications. The assumption that Commodore could simply keep moving through faster 68000-family processors forever was becoming harder to defend. AAA had already experimented with the idea of alternative CPUs. Hombre went much further. The old Amiga architecture was no longer sacred.

Hombre never reached production

By early 1994, Hombre had moved beyond the level of a vague idea. The design had gone through revisions. Engineers were working on major controller and display chips, and different types of systems had been considered. But semiconductor projects don’t become finished consumer products overnight.

Hombre needed more time. Commodore didn’t have it. The company declared bankruptcy in April 1994. AAA never became the foundation of a new Amiga generation. Hombre never reached the market. The A1200 and A4000 ended up representing the last major Commodore-produced Amiga generation.

Commodore had several futures, but no single roadmap

Looking at the early 1990s Amiga story as a straight product timeline misses what was actually happening. Commodore was running several different ideas at once. AAA was the ambitious next-generation Amiga architecture.

CDTV tried to turn Amiga technology into a consumer multimedia appliance. The Amiga 3000 pushed the existing computer platform upwards. AGA gave Commodore something practical it could ship while AAA stalled. The A600 tried to reshape the low end but arrived without the upgrade buyers were expecting.

The A1200 finally delivered AGA to the mass market, but Commodore couldn’t make enough of them at the right moment. The A4000 brought AGA to the high end. Then Hombre questioned whether the next Commodore machine needed to look like an Amiga at all.

That’s the interesting part of the story. Commodore didn’t run out of technical ideas. Quite the opposite. It had too many projects competing for limited engineering resources, limited manufacturing capacity and increasingly limited time. Meanwhile, the PC market kept moving.

A graphics target that looked advanced one year could look ordinary a few years later. CD-ROM shifted from novelty to expectation. Standard buses such as PCI became more important. RISC processors changed the performance discussion. Dedicated 3D hardware was approaching fast.

Commodore kept responding, but each new response arrived before the previous one had fully settled. By 1994, the company had travelled from AAA’s vision of a far more powerful traditional Amiga to Hombre’s idea of a 64-bit, RISC-based multimedia platform that might run Windows NT or appear as a PCI card. That’s a huge change in only a few years. And it shows just how quickly the computer industry had moved around Commodore.