AGA successor: Document about the Advanced Amiga Architecture published

[cynix]

Saw this linked over on Hacker News and couldn't see it mentioned anywhere here so thought some of you guys might be interested :

https://archive.org/details/advanced...ge/n3/mode/2up

"Document for the Commodore-Amiga Advanced Amiga Architecture. Dated June 18, 1992, this was the never-released whole new Amiga system architecture. This project was started in 1988, and first silicon was available in 1993. Commodore's business problems prevented subsequent progress on this system. "

Apologies if this is old news.

[Bruce Abbott]

Evidence that "first silicon was available in 1993"?

[cynix]

Originally Posted by Bruce Abbott

Evidence that "first silicon was available in 1993"?

It's a quote from the uploader of the document, Dave Haynie, who was a lead hardware engineer at Commodore and worked on the design + made this video about the last days of Commodore: [ Show youtube player ]

So unless someone has evidence to the contrary it seems reasonable to assume that what he says is accurate.

But I'm just an ST guy who knows next to nothing about Commodore and Amiga history, so don't take my word for it.

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No secret we all know but they admit AA/AGA was outdated before it was released:-

[Aulapatchuc]

Originally Posted by Toffee

No secret we all know but they admit AA/AGA was outdated before it was released:-

Dave Haynie said it many years ago in Amiga.org that AA was ready in early 1990, and Commodore slackers only released it at the end of 1992, as we all know.

[Bruce Abbott]

Originally Posted by Aulapatchuc

Dave Haynie said it many years ago in Amiga.org that AA was ready in early 1990, and Commodore slackers only released it at the end of 1992, as we all know.

Cite?

[desiv]

Originally Posted by Bruce Abbott

Cite?

[ Show youtube player ]
His talk about the chipsets is around the 30 minute mark.
He mentions the AA chips would be coming out in 1990 at around 30:50...

It's worth watching all his vids. They are great..

[Bruce Abbott]

Originally Posted by cynix

It's a quote from the uploader of the document, Dave Haynie

Thanks for that, I didn't realize it was Dave Haynie himself who uploaded the document.

The problem here is that his statement was made on April 16, 2023, 30 years after the fact. However in an interview he did with Kees Witteveen of Amiga.org in 2003, he said "...and of course, AAA was abandoned before it was capable of even booting AmigaOS". So if there was a 'first silicon', it apparently wasn't good enough to be used in an actual Amiga.

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? He also states "The interface to any kind of system will be implemented in some kind of “glue” chip, most likely a new gate array, though this function certainly could be implemented in a PAL and TTL based circuit, as on the AAA prototype system.". So did he have a full working chipset, or was the prototype using something else?

It might seem like nitpicking, but when we have people making statements like "AA was ready in early 1990, and Commodore slackers only released it at the end of 1992", I think it's important to stick to known facts and reasonable conclusions. If the AAA chipset was advanced enough to have actual silicon produced in 1993 then IMO Commodore's engineers were not 'slacking'.

But I'm just an ST guy who knows next to nothing about Commodore and Amiga history

Then perhaps you didn't know that this document would be more fuel for the Amiga fan's favorite activity - bashing Commodore.

[oscar_ates]

We talked a lot about AGA was not enough in the other disappointed thread. Apparently it was even obvious to even Amiga engineers that it was "too little, too late".

[Pyromania]

Does this mean AAA chipset based Amigas ship in two weeks fingers crossed?

[Bruce Abbott]

Originally Posted by desiv

He mentions the AA chips would be coming out in 1990 at around 30:50...

But in that 2003 interview with Amiga.org he said:-

The story begins in 1991, when Sydnes took over as VP of Engineering. I was working on the _real_ A3000+, the first prototype of which was the first AA machine ever, back when we called it 'Pandora'. This machine was using mainly A3000 parts (I planned to revise it to the 040 bus once the AA stuff had been proven).

So in 1991 the AA/AGA chipset wasn't 'proven'. This is confirmed by OS developer notes talking about various bugs in the chipset that were still being worked out as late as April 1992.

What this means is that the window of 'slacking' is much smaller than has been suggested.

Regarding AAA, I found this from Dave Haynie.

It wasn't until 1993 that the first silicon from that project was available -- I built the development system. The chips weren't done, they were only partially functional.

In that Quora article Haynie also said,

Applying an extra US$5-6 million to chip development back then, that alone could have been the difference.

Now, this wasn't the only issue in the end. But it was the fundamental one. Professionals were switching to Macs or PCs because they couldn't get the performance they needed out of Amigas.

But most professionals weren't switching to Macs or PCs because the Amiga wasn't powerful enough, they did it because the PC and Mac between them had 95% of the 'professional' market. This just shows how delusional even the 'smart' engineers at Commodore were. AAA in the A4000 or even in the A3000 wouldn't have made much difference.

[desiv]

Originally Posted by Bruce Abbott

But in that 2003 interview with Amiga.org he said:-

So basically, your saying Dave's stories change from time to time, so it's tough to accept anything he says as detailed fact (i.e. what specific year anything happened)?

Well, he wasn't called HazeyDave for no reason... ;-)

[Bruce Abbott]

Originally Posted by oscar_ates

We talked a lot about AGA was not enough in the other disappointed thread. Apparently it was even obvious to even Amiga engineers that it was "too little, too late".

And this is the only comment worth making about this document? Sad.

[Bruce Abbott]

Originally Posted by desiv

So basically, your saying Dave's stories change from time to time, so it's tough to accept anything he says as detailed fact (i.e. what specific year anything happened)?

Well, he wasn't called HazeyDave for no reason... ;-)

People's recollections fade over time, so for accuracy it's often best to get more contemporary accounts.

But perhaps we shouldn't dwell on the past too much. This document is interesting for what we might want to do in the future. Could we use it to help design a 'new' chipset with features similar to AAA?

[Pyromania]

Originally Posted by Bruce Abbott

People's recollections fade over time, so for accuracy it's often best to get more contemporary accounts.

But perhaps we shouldn't dwell on the past too much. This document is interesting for what we might want to do in the future. Could we use it to help design a 'new' chipset with features similar to AAA?

Doesn't the SAGA chipset in the Vampire already do this.

[grelbfarlk]

Yeah you know what if they just said, hey rather than spending $5-$6million, we just put a slot in there, and put the bestest cheapest card in there. Like they did on the A4000. And then proved everyone who made a better card, which was everyone was more competent at it than them. And now we have UAGA and XAGA and žAGA.

[coder76]

To keep the Amiga hardware competitive for a few years more in the 90's, improvements were not needed so much with sprites, blitter, or copper, but rather with chip ram speed. The PC's display hardware before 3D cards came out wasn't often more than a graphics card that offered to open a screen with various resolutions, while the CPU filled the screen with pixels. But Amiga was stuck with an ISA-like slow bus, while PCs got PCI, which was fast enough for fast 486's and Pentiums.

[eXeler0]

Originally Posted by coder76

To keep the Amiga hardware competitive for a few years more in the 90's, improvements were not needed so much with sprites, blitter, or copper, but rather with chip ram speed. The PC's display hardware before 3D cards came out wasn't often more than a graphics card that offered to open a screen with various resolutions, while the CPU filled the screen with pixels. But Amiga was stuck with an ISA-like slow bus, while PCs got PCI, which was fast enough for fast 486's and Pentiums.

yea, they should probably have focused on getting the bandwidrth /speed to do 800x600 @256 colors at usable speeds by the time the A3000 was released, that would have been a good place to start.

[Thomas Richter]

Originally Posted by Bruce Abbott

Could we use it to help design a 'new' chipset with features similar to AAA?

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

Originally Posted by Bruce Abbott

But most professionals weren't switching to Macs or PCs because the Amiga wasn't powerful enough, they did it because the PC and Mac between them had 95% of the 'professional' market. This just shows how delusional even the 'smart' engineers at Commodore were. AAA in the A4000 or even in the A3000 wouldn't have made much difference.

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.

[jbenam]

AAA _could_ have made a difference if released in 1990.

After they missed that target they should’ve just gone with ISA and then PCI - and get access to the massive ecosystem that already existed for the PC.

You could’ve kept the chipset as an integrated entry-level solution while power users could’ve purchased PC graphic chipsets. Which is what happened in the end - they just had to waste time to reimplement it in Zorro.