Continuing the discussion from here: https://amigaworld.net/modules/newbb/viewtopic.php?mode=viewtopic&topic_id=45508&forum=16&start=500&viewmode=flat&order=0#880974

@Hammer

Quote:

Hammer wrote: @cdimauro Quote: Who was the (technical) responsible for that? Was it Porter again? After the C65's 8-bitplane chipset reveal and Rubin's approval, AGA ASIC R&D was done by a CSG LSI team member.

Which proves that there was the possibility to have a new Amiga chipset.
Quote:

AA3000 and AA1000plus are system integration projects.

What's the point here? It's normal that a product is made of different components which are assembled & work together to fit the project scope.

BTW, I'm System Integration Test Lead at BMW, and I'm working on exactly this area (e.g.: SERVING the System Integration by testing the assembled products).
Quote:

AA3000 acted as the primary system integration test platform for AGA.

That's normal: you need a testing platform for the product coming from the system integration (process).

I'm also responsible and manage/handle/steer the entire test farm for BMW's Infotainment system, so I know very well this topic, too.
Quote:

Jeff Porter was a task allocator within the system engineering group within the bounds of management directives e.g. Herni Rubin's monochrome hi-res management directive.

So, it was a project manager which roughly acted also as a "scrum master" (!).
Quote:

Herni Rubin has other directives, not just monochrome hi-res.

Nothing to say, here: we know that Commodore was working at several projects in parallel, and he was responsible for some of them.
Quote:

Quote: The hierarchy was defined, but the technical people were incompetent and not even able to agree on the SPECs of a project... Wrong. Similar-ranked headstrong engineers are not cooperating, which needs proper leadership i.e. the captain is missing in action. A1000's project had Jay Miner as leader for both system integration and custom ASIC design for both ICS and OCS (drop-in 64 color EHB improvement).

I agree on that. There were too many engineers which had completely different ideas on how (IF) to evolve the platform, which stalled its development.

Managers need a single point of contact as a reference regarding technical topics, and this was clearly missing.

I manage several engineers working on different areas / work packets, and I've a responsible for each of them to whom I talk about the activities to be performed, issues, etc.. I don't need to talk with every single engineer regarding topics on a specific area, neither I need the technical knowledge (they are responsible and are paid for that, and they should deliver the expected results).
Quote:

After the A1000 project, the monochrome hi-res mandate was Herni Rubin's.

Like for other projects.
Quote:

Quote: This furtherly proves how engineers were doing whatever they wanted, even against what managers had said them to do. Managers are oblivious to PC graphics.

As you reported from the book, no: it wasn't the case. Managers knew PC graphics (and Mac graphics as well), and they started meetings for discussing the specs of the Next Big Thing exactly with the purpose of contrasting the competition which became very serious.
Quote:

Quote: In fact, Hedley haven't proved why this ‘bitplane to pixel converter’ was so important for the... CDTV. Which was Commodore's multimedia machine. 1. For the CDTV-CR project after CDTV. 2. Improved PC graphics handling without modifying the core Amiga graphics.

Again, that wasn't the purpose for those specific products, as I've already tried to explain.
Quote:

Quote: He has wasted time and resources for the company just to play with his idea... Amiga's multimedia group can only modify surrounding chips (i.e. system integration), and they don't have the mandate to change the core Amiga graphics design e.g. AGA R&D needs Henri Rubin's approval. Amiga's multimedia group only has a system integration mandate. Remember, the system engineering group was responsible for CBM's 8-bit business desktops and C900's system integration. The system engineering group created a VLSI team for AAA and Hombre. VLSI team recycled some engineers from the CSG LSI team. VLSI team is under the system engineering group's administration. The CSG's LSI team is responsible for VIC-20 and C64 for both AISC and system integration R&D, which largely kept the CIC group alive until the A500 takeover. The problem is the weak business plan after A500. Corporate structure limitation is an executive board management problem. It was the executive board that deleted the original Los Gatos Amiga team.

Rubin's reaction to the LSI's C65 has shown that this mandate was certainly possible (e.g.: putting people from different teams together to work on a specific project).
Quote:

The Amiga lacks Ken Kutaragi's leadership capability.

Kutaragi was also a very good engineer, and Commodore clearly lacked them, as it was proven.

@cdimauro

Quote:

cdimauro wrote: Kutaragi was also a very good engineer, and Commodore clearly lacked them, as it was proven.

Commodore had good engineers. If they didn't then we would have all the awesome stuff they produced - 9 wonderful Amiga models that all worked as intended and are still working 30+ years later! Plus many expansion cards and peripherals that were all competently designed.

Tramiel era machines weren't so great though. The C64 was Commodore's most popular model, and yet was full of design flaws and bad engineering. But that was largely due to Tramiel pressuring them to get stuff out the door ready or not.

Commodore's engineers had their faults, but bad engineering wasn't one of them - at least not during the Amiga years. They certainly weren't any worse than the engineers who designed the infamous Apple III and IBM PCjr, or even the Mac and PC, or the Ti-99/4a, Timex Sinclair 2068, Acorn Electron, Amstrad CPC to name just a few others with significant design flaws.

@bhabbott

Quote:

bhabbott wrote: @cdimauro Quote: cdimauro wrote: Kutaragi was also a very good engineer, and Commodore clearly lacked them, as it was proven. Commodore had good engineers. If they didn't then we would have all the awesome stuff they produced - 9 wonderful Amiga models that all worked as intended and are still working 30+ years later! Plus many expansion cards and peripherals that were all competently designed. Tramiel era machines weren't so great though. The C64 was Commodore's most popular model, and yet was full of design flaws and bad engineering. But that was largely due to Tramiel pressuring them to get stuff out the door ready or not.

Which kind of innovations brought the Amiga engineers after the Amiga 1000?

Kutaragi brought the first Playstation and all its innovations. The Amiga engineers?
Quote:

Commodore's engineers had their faults, but bad engineering wasn't one of them - at least not during the Amiga years.

Please, tell me more about the Slow Mem which was worse than Chip and Fast Mem, the ECS (after FIVE years from the OCS) with its ridiculous 4 cycles Chip mem & registers access time for the 68020, the Amber chip (instead of integrating a scan doubler to che chipset), Super Buster and its "awesome" memory bandwidth, the FrankenPatchwork which AGA represents which was still a 16-bit chipset and most of the things left untouched, Akiko (crappy C2P WITHOUT using the integrated DMA controller), and Hombre with the HP PA-RISC with its ridiculous amount of caches.

Just to report the first ones which come immediately to my mind.
Quote:

They certainly weren't any worse than the engineers who designed the infamous Apple III and IBM PCjr, or even the Mac and PC, or the Ti-99/4a, Timex Sinclair 2068, Acorn Electron, Amstrad CPC to name just a few others with significant design flaws.

Compare them to the OVERALL stupidities that Commodore engineers have made in just 9 years, and draw your conclusions...

Hint: April 1994...

@cdimauro

Quote:

cdimauro wrote: Please, tell me more about the Slow Mem which was worse than Chip and Fast Mem,

a) it wasn't worse, it combined the flaws of the 2
b) it was a significant cost cutting measure (no need for a separate controller) that had minimal performance impact on 68000 systems.
c) it was only meant to be in the low cost A500, the A2000A had real 512k FAST in the MMU slot. The A2000B was a rushed fix for the A2000A failure
d) once Agnus went 1MB C= never again shipped a system with SLOW. A500 of that time needed to have just one Jumper cut when upgraded with a trapdoor card.

Plenty worse WFTs in pretty much every piece of HW of that period.

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@cdimauro

in my view management was most to blame. The main problem was, there was no real longterm plan where to go. Amiga was just one product, 16bit home computer. Most money was earned with C64. And they also sold PCs but they were too small for that. They certainly had losses there. But in the company they used PCs for administration, not amigas. For them the amiga was just a toy system they could sell at young people, not a key technology for Commodore to survive. Additionally they did a lot of costly wrong decisions. certainly engineers have a tendency to overengineer and develop products nobody want to buy being too expensive or set wrong priorities. But even for that is management to blame. There must be competent project managers who oversee a project and report to the top management.

Last edited by OlafS25 on 08-Sep-2025 at 09:07 PM.

@cdimauro

Quote:

As you reported from the book, no: it wasn't the case. Managers knew PC graphics (and Mac graphics as well), and they started meetings for discussing the specs of the Next Big Thing exactly with the purpose of contrasting the competition which became very serious.

Managers rejected an immediate solution for PC graphics handling. Akiko's C2P is abstracted by an API anyway. CDTV-CR/CD32's multimedia team doesn't have the mandate to modify the core Amiga graphics design.

The Next Big Thing like the AAA project is tainted with a crazy high-end 64-bit VRAM-equipped workstation when the mainstream SVGA with improved FPM DRAM is the immediate competitive threat. Commodore HR failed to hire enough chip engineers when there was a single engineer bottleneck for C65, AAA, and ECS Agnus B. Brain drain is real.

AAA scales from 32-bit FPM DRAM to 64-bit VRAM with up to 16MB Chip RAM (dual 32-bit AAA chipset mode), while projects like Amiga ICS/OCS and 3DO focus on cost-effective single memory configuration.

Early SVGA chipsets such as ATI VGA Wonder (Q1 1988) and ET3000AX (Q4 1987) have the fastest FPM DRAM for their time without using expensive VRAM.

For CeBit 1988, Commodore PC engineers and Herni Rubin were impressed with ATI's 1988 era SVGA chipsets(e.g. VGA Wonder), so they demoed them instead of the Amiga 500 with a flicker fixer, even though it was Herni Rubin who issued the monochrome hi-res directive in 1986.

System engineering leadership ignored the following arguments:
1. 1987 era 32-bit chipset evolution argument = ignored.
2. 1987 era 8-bitplane chipset evolution argument = ignored.
3. 1987 era 16-bit (65,536 colors) graphics chipset evolution before moving into 24-bit graphics argument = ignored.

The Unix camp wanted the high-end workstation. The Amiga Unix (Make AT&T Great Again) project is another C900 mk2 distraction that departed from Commodore's core market and mass production strengths.

There are many other project distractions e.g. C64GS with C65's cartridge size support, PC genlock, various mice control devices, A2410 TiGA, A2024 monitor with 5000 production scale, x86 sidecart for A1000, system integration Commodore PC clones, and many others.

Mac LC II's best-selling Mac model for 1992 was done with a modest 256-color and 65,536-color display modes.

Commodore management failed to replace Jay Miller's Amiga OCS and CSG's C64 core business.

Last edited by Hammer on 09-Sep-2025 at 01:29 AM.

_________________

@cdimauro

Quote:

Kutaragi brought the first Playstation and all its innovations

Most of PlayStation 1's innovations were done by external party IP i.e. LSI Coreware CPU and Toshiba GPU.

With a DSP ASIC background for SNES, Kutaragi provided good leadership direction for various entities to work together.

Kutaragi stated that most of his Sony meetings are dominated by price vs performance debates.

Quote:

Please, tell me more about the Slow Mem which was worse than Chip and Fast Mem, the ECS (after FIVE years from the OCS) with its ridiculous 4 cycles Chip mem & registers access time for the 68020, the Amber chip (instead of integrating a scan doubler to che chipset),

1. 1MB ECS Agnus A in a 512KB Chip RAM + 512KB Slow RAM configuration can still access the slow RAM as Chip RAM. This feature only works for ECS Agnus A.

1989 era A500 Rev6A is the de facto Amiga configuration standard.

A certain demo scene exploited the A500 rev6A ECS Agnus A's 512KB Chip RAM + 512KB Slow RAM configuration as 1 MB Chip RAM.

Detect C0$ Slow RAM and ECS Agnus A existence and exploit CDTV-style's 1 MB Chip RAM.

2. Amber chip with a high-speed serial frame buffer chips is a workaround for "read my lips, no new chips" with the Amiga graphics context.

3. https://www.retroisle.com/commodore/amiga/Technical/Firmware/customchips.php
From Dave Haynie

But that year (1991), CSG was changing over from channelled gate arrays to sea-of-gates technology, and their first chips were running at 1/4 speed or so. Six months later, I had fast chips, but also the new Engineering Administration, which considered Gemini either an illegal research project, or "something from the last administration". So, no big pounding on the Zorro III bus was possible before the A4000 shipped.

In all fairness, the Zorro III project was a large thing to bite off. When the PC Industry made EISA, there were hundreds of people involved: some to hammer out the specs, some to make chips, test, etc. At Commodore, it was basically just me for most of the life of Zorro III.



Zorro III is mostly a solo effort by Dave Haynie, while the PC world has a large engineering resource pool for the chipset designs.

Blame Commodore management for hiring engineers for Commodore's PC clones instead of having the "Amiga 1st" ideology.

Apple didn't betray its in-house platforms like Commodore!

Commodore management's Acutiator rejection shows management's exit direction from the Amiga.

From Commodore - The Final Years book

On December 20, 1991, Dave Haynie and Greg Berlin finished their
proposed spec for a fourth generation Amiga system architecture.
They called the proposal Acutiator, which was a medieval term for a
sharpener of weapons. They wrote, “One of the main design goals of
the Acutiator architecture is to separate functions into modular
pieces. This gives us the flexibility to design low-cost systems which
make use of some subset of these components, or to use them all to
create a ‘feature enriched’ machine for those customers willing to
pay the additional cost.�

The engineers wanted to allow the then-new Motorola 68040
processor to work with the next generation of Amigas. And of
course, the architecture would work with the upcoming AAA chipset,
as well as the more imminent AA. And because AAA was designed to
work with different processor families, Haynie wanted his Acutiator
motherboard to also handle different processors. Specifically, there
were at least three major RISC processor families at that time and
he wanted Acutiator to accept these RISC chips.

Their architecture required three custom chips: EPIC, AMOS, and
SAIL. In cost comparisons, Haynie calculated that the Acutiator
architecture would add approximately $125 to a system (including
the cost of a 68EC040 chip), resulting in a $300 retail price increase.
This was a bargain, considering the user received a significant
processor upgrade. Haynie proposed that Commodore should assign
Scott Shaeffer, Paul Lassa, and himself to each create the three
required gate array chips. He expected prototypes in 7 to 9 months,
with the first systems shipping in 1992.


Without a cache-coherent chipset, Amiga couldn't exploit $100 68EC040-25 which is about the same price as Am386-40, i386DX-25 and 68030-25 (with MMU)


from https://archive.computerhistory.org/resources/access/text/2013/04/102723262-05-01-acc.pdf
DataQuest 1991, page 119 of 981

For 1992,
68EC020-16, $16.06
68EC020-25, $19.99
68020-25, $35.13
68EC030-25, $35.94 (missing MMU, not Unix capable, used in A4000/EC030)
68030-25, $108.75
68040-25, $418.52
68EC040-25, $112.50 (missing MMU and FPU, Commodore management rejected glue chips for Amiga)

AM386-40, $102.50
R3000-25, $96.31

386DX-25, $103.00
80486SX-20, $157.75 (still has MMU for Xenix 386, Windows NT, and Linux)
80486DX-25, $376.75
80486DX-33, $376.75
80486DX-50, $553.25
80486DX2-50, $502.75

Q4 1992, I would select 68EC040-25/AGA Amiga ahead of the i386DX33/Am386-40 PC with a similar price range i.e. AU$1500 or US$1000.



Quote:

Super Buster and its "awesome" memory bandwidth, the FrankenPatchwork which AGA represents which was still a 16-bit chipset and most of the things left untouched, Akiko (crappy C2P WITHOUT using the integrated DMA controller),

AGA was a rush job to save the system engineering administration's ego against C65's reveal. Corporate politics and "losing/saving face" (ego) are real.

Akiko C2P has to overcome upper management roadblocks. The idea about bitplane/chunky pixel conversion hardware started in the CDTV-CR project.

The "read my lips, no new chips" with the Amiga graphics context was removed with the C65 reveal, hence AA3000's original Q4 1991 release was the fix against press criticism for A3000's lacking 256 color mode.

Quote:

and Hombre with the HP PA-RISC with its ridiculous amount of caches.

Commodore's PA-RISC selection is Hitachi PA/50. Hombre's L1 cache structure resembles PA/50's.

Commodore's RISC selection is with the "cheap" RISC camp since Motorola 88000's asking price is anti-game console price.

Last edited by Hammer on 10-Sep-2025 at 05:44 AM.
Last edited by Hammer on 10-Sep-2025 at 05:35 AM.
Last edited by Hammer on 09-Sep-2025 at 01:57 AM.

_________________

@cdimauro

Quote:

cdimauro wrote: @bhabbott Which kind of innovations brought the Amiga engineers after the Amiga 1000?

You know what kinds. I sense what you are really asking is why didn't they make the kinds of innovations you think they should have made.

To understand the engineers actions after the A1000 we need to look at what came before it. The C128 was launched on June 2, 1985 at the Chicago CES. At the show Commodore CEO Thomas Rattigan announced that 100,000 pre-orders had already been taken, and that 15% of C64 users were expected to upgrade to it within the year (which turned to be an accurate prediction). The C128 was much more successful than the engineers expected, eventually selling 4 million units by the end of 1989 (almost as many as the entire Amiga range up to 1994).

The A1000 was released a month later with big fanfare. However it was expensive to produce so the price was well above most other home computers, and Sanyo had limited production capacity so they couldn't sell many even they had customers for them. With $55 million spent on advertising and only 70,000 sold they lost money on every unit. Reducing manufacturing cost was essential if the Amiga was to survive.

The next model could have the OS in ROM (once it was finished) which removed one the expensive parts of the A1000. But that wouldn't be enough. Atari was doing well with the much lower cost ST, then introduced the 1040ST which was the first home computer to have 1MB for 'under' $1000. The lower cost Amiga would be competing against it and the C128.

The Los Gatos 'innovators' who had designed the original Amiga were tasked with designing a lower cost A1000, but they stuck closely to the original design and were not able to save much. So the Commodore engineers at West Chester decided to work on their own lower cost model. A few of them dismantled a 520ST and compared it the A1000. The ST had far fewer parts. They figured that a form factor similar to the ST or C128 was what would sell to the masses. As well as making the keyboard integral they would need to integrate as much of the discrete logic as possible. Thus Fat Agnus and Gary were conceived. the 84 pin PLCC Agnus (the first PLCC and largest custom chip Commodore had made to date) integrated the clock logic, while Gary incorporated the expensive power-hungry PALs and other logic into a low cost low power gate array.

When the Los Gatos Amiga team found out about the A500 they scoffed, suggesting that Fat Agnus wouldn't work and that removing the WCS was a mistake. This shows that Commodore's engineers were more innovative than Jay Miner's engineers.

You may complain that they didn't innovate on enhanced features, but that was not what was needed. The A1000 got developers familiar with the Amiga's custom chips and OS so they would be able to produce software that made best use of it. By 1987 when the A500 was released a number of titles were showing what the Amiga could do, making ST users jealous.

The A500 was not the machine to change the ends again, but to consolidate the architecture and bring the A1000's capabilities to the masses. It's just unfortunate that IBM released VGA with 256 colors in the same year, though it took several years to become popular and couldn't match the Amiga in action games until the 90's (when x86 CPUs got powerful enough to make up for not having a blitter etc.).

Quote:

Please, tell me more about the Slow Mem which was worse than Chip and Fast Mem

SlowRAM was an innovation. You may scoff, but the Amiga needed a cheap way to get 1MB to counter the 1040ST. FastRAM wasn't cheap because it needed a dedicated DRAM controller. SlowRAM was practically free to implement, and made the expansion board cheap to manufacture. They added a RTC at the same time because the OS needed it for serious applications. This was also very cheap to implement because it only needed a few more pins on the trapdoor connector and Gary did the I/O address decoding.

You say SlowRAM was worse than ChipRAM or FastRAM, but it was no worse than ChipRAM for code and non-custom chip data. Most games would compress data anyway to save memory, and didn't need that much actual ChipRAM. SlowRAM gave customers a 1MB expansion at the lowest possible price with performance equal to the base machine. Games and apps that didn't push the bandwidth too high wouldn't be slowed by a noticeable amount compared to FastRAM.

In practice nobody but people like you got upset about it. Fans were happy to get 1MB cheaply for software that needed it, and it didn't make their machine any slower than it already was. If they wanted more then they could always install a RAM expansion on the Zorro slot like A1000 owners did. In 1989 Commodore provided up to 2MB of true FastRAM in the A590, along with a 20MB hard drive and DMA SCSI interface. That turned the A500 into a very capable machine for 'serious' use. And yes, that was innovation too!

Quote:

the ECS (after FIVE years from the OCS) with its ridiculous 4 cycles Chip mem & registers access time for the 68020, the Amber chip (instead of integrating a scan doubler to che chipset), Super Buster and its "awesome" memory bandwidth, the FrankenPatchwork which AGA represents which was still a 16-bit chipset and most of the things left untouched, Akiko (crappy C2P WITHOUT using the integrated DMA controller), and Hombre with the HP PA-RISC with its ridiculous amount of caches.

No, the 4 cycle ChipRAM access was not 'ridiculous ' just because you had an 020 (which needed 3 cycles) running at a higher clock speed. By that logic anyone with an 060 would expect ChipRAM to be doing single cycle access at 50MHz. Once the chipset was designed to work with that timing it had to keep it for compatibility, just like ISA bus video cards did (which BTW were no faster than AGA - actually slower unless the ISA bus was overclocked, which might stop other cards from working).

Just because the engineers didn't implement what you think they should have doesn't make them bad engineers. It's actually quite arrogant to suggest that you could have done a much better job. You spew out 'Integrated AMBER' and 'DMA Akiko' as if these things would be a doddle, then dump on Dave Haynie for not managing to get real silicon running at the theoretical maximum speed of Zorro III. Let's see your designs to do this - I'm sure the community would welcome them. Should be a doddle with the chips we have today, right?

Quote:

Compare them to the OVERALL stupidities that Commodore engineers have made in just 9 years, and draw your conclusions... Hint: April 1994...

Ti was out of the home computer business with their tail between their legs by 1981. Sinclair sold out to Amstrad in 1986 for a mere £5 million. Amstrad stopped making home computers after the CPC Plus line flopped in 1990. the Acorn Electron was almost a complete disaster with missed launch dates, ULA reliability problems and less than half the performance of the BBC micro (partly due to using 4-bit RAM!).

And that's just a few of the horror stories about home computer engineering. Commodore was by no means the worst.

Last edited by bhabbott on 09-Sep-2025 at 02:55 AM.

@bhabbott

Quote:

The A500 was not the machine to change the ends again, but to consolidate the architecture and bring the A1000's capabilities to the masses. It's just unfortunate that IBM released VGA with 256 colors in the same year, though it took several years to become popular and couldn't match the Amiga in action games until the 90's (when x86 CPUs got powerful enough to make up for not having a blitter etc.).

286-16 with VGA clone is enough for the Body Blows game port from the Amiga 500.

It's programming mastery on VGA's Mode X / Mode Y.

https://www.youtube.com/watch?v=5o9yOBBWPgM
This YouTube video shows a 286 PC clocked at 16 MHz with VGA and Sound Blaster adapters

10:00 Tubular Worlds,
15:54 Body Blows,
31:46 Gods,
44:47 Pinball Fantasies,
58:52 Prehistorik 2 with background parallax.

VGA is more than Atari ST's effort.

Read https://www.phatcode.net/res/224/files/html/ch24/24-01.html
Title: Parallel Processing with the VGA.

This heading refers to the ability of the VGA chip to manipulate up to four bytes of display memory at once. In particular, the VGA provides four ALUs (Arithmetic Logic Units) to assist the CPU during display memory writes, and this hardware is a tremendous resource in the task of manipulating the VGA’s sizable frame buffer.

The latches can be used to perform 4-byte-at-a-time (one byte from each plane) block copying.



https://youtu.be/NudEusAMzr0?t=5671
Am386DX-40 with 256K L2 cache and "SpeedStar 24" ET4000AX running Star Wars Dark Forces (1995). Escom's Cx486DX-40/S3 805-based PC is also compared.

32-bit S3 805 VLB is shipped with 60 ns or 70ns FPM DRAM memory chips. S3 805's 86C801 chipset has a full 32-bit graphics architecture.

Acutiator's 68EC040-25 configuration was designed to compete against the fast 386DX-40 and low-end 486SX.

From Commodore - The Final Years

Their architecture required three custom chips: EPIC, AMOS, and
SAIL. In cost comparisons, Haynie calculated that the Acutiator
architecture would add approximately $125 to a system (including
the cost of a 68EC040 chip),


Notice PC's mainstream SVGA used lower latency FPM DRAM in a 32-bit architecture. FPM DRAM was replaced by EDO DRAM e.g. S3 Trio 64UV+.

At the higher end, VRAM was replaced by SGRAM (SDRAM for graphics, SGRAM is single-ported, which can open two memory pages simultaneously). SGRAM was a precursor to Graphics Double Data Rate (GDDR) memory.

PlayStation 1 replaced expensive VRAM with cheaper SGRAM. My point, VRAM was replaced by DRAM-based SGRAM.

On price, PC's mainstream SVGA with FPM DRAM has undercut IBM's 8514 with VRAM.

In modern times, Apple M-series silicon uses mainstream DDR5 with a 256-bit bus to undercut GDDR6. AMD responded with Strix Halo's mainstream DDR5 with a 256-bit bus.

About finding a sweet spot between performance vs cost. HBM has a high-end graphics memory position.

A500's original success was about good performance vs cost.

Quote:

Once the chipset was designed to work with that timing it had to keep it for compatibility, just like ISA bus video cards did (which BTW were no faster than AGA - actually slower unless the ISA bus was overclocked, which might stop other cards from working).

ISA's 8.3 Mhz x 16 bit is 16.6 MB/s bandwidth. ET4000AX enables zero wait state on the ISA bus.

MCA has 10 Mhz. If the add-on card's chipset is designed for MCA and ISA, the ISA card variant can also do 10 Mhz.

https://www.deltabeta.it/2010-2016/corso_rete/varie%20-%20connettori/connettori/co_ISA_Tech.html

8.3 MHz is specified as the maximum, but many systems allow this clock to be set to 12 MHz and higher.


ET4000AX supports both MCA and ISA bus types, hence it's tolerant of 10MHz.

Last edited by Hammer on 09-Sep-2025 at 05:51 AM.
Last edited by Hammer on 09-Sep-2025 at 05:27 AM.
Last edited by Hammer on 09-Sep-2025 at 05:24 AM.
Last edited by Hammer on 09-Sep-2025 at 04:42 AM.

_________________

@bhabbott

Quote:

The next model could have the OS in ROM (once it was finished) which removed one the expensive parts of the A1000. But that wouldn't be enough. Atari was doing well with the much lower cost ST, then introduced the 1040ST which was the first home computer to have 1MB for 'under' $1000. The lower cost Amiga would be competing against it and the C128. The Los Gatos 'innovators' who had designed the original Amiga were tasked with designing a lower cost A1000, but they stuck closely to the original design and were not able to save much. So the Commodore engineers at West Chester decided to work on their own lower cost model. A few of them dismantled a 520ST and compared it the A1000. The ST had far fewer parts. They figured that a form factor similar to the ST or C128 was what would sell to the masses. As well as making the keyboard integral they would need to integrate as much of the discrete logic as possible. Thus Fat Agnus and Gary were conceived. the 84 pin PLCC Agnus (the first PLCC and largest custom chip Commodore had made to date) integrated the clock logic, while Gary incorporated the expensive power-hungry PALs and other logic into a low cost low power gate array. When the Los Gatos Amiga team found out about the A500 they scoffed, suggesting that Fat Agnus wouldn't work and that removing the WCS was a mistake. This shows that Commodore's engineers were more innovative than Jay Miner's engineers. You may complain that they didn't innovate on enhanced features, but that was not what was needed. The A1000 got developers familiar with the Amiga's custom chips and OS so they would be able to produce software that made best use of it. By 1987 when the A500 was released a number of titles were showing what the Amiga could do, making ST users jealous.

A500's original success was between the Los Gatos Amiga team and the system engineering group.

Amiga ICS and OCS core graphics R&D are credited to the Los Gatos Amiga team.

Amiga 500's cost-reduced system integration work is credited to the system engineering group.

Amiga needs both groups to balance the leading-edge R&D, cost reduction, and system integration.

Ex-Los Gatos Amiga leadership was able to complete 3DO (completed in 1993) and 3DO M2 (completed in 1995, sold to Panasonic for $100 million). 3DO MX was led by an ex-SGI engineer and assimilated into Xbox.

Last edited by Hammer on 09-Sep-2025 at 06:19 AM.

_________________

@OlafS25

I don't think so. After reading the Bagnall's books, as everybody has already done here I believe, it seems the truth is in the middle as usual. Yes, they were brilliant and talented engineers but also youngsters, full of egos and weird, that lead to nonsense engineering decisions - because they felt cool putting questionable stuff (like CP/M in C128...) and ignoring and staying far away from marketing/executives guys' scrutiny as much as possible. Commodore was a kind of creative magnet for all those strange guys that didn't want to wear suits. Indeed, we had the various Bill Herd and his Animals team, G. Robbins, D. Needle, the two guys who worked on C65, and so on... often they all had erratic, drunk and even violent behaviours. On the other hand the toxic/very demanding environment in Commodore didn't help at all, thanks to Tramiel's legacy, too.
In the 80s Commodore executives felt the Amiga was designed to be only a high-end system, Los Gatos team was even more intransigent about this - only D. Morse wanted Amiga as a game machine, but he left in 1985. Irving Gould pushed personally to make the A2000. Clearly Commodore wanted to distance itself from the world of games, after the infamous crash of 1984, at least until late 80s, when A500 started to invade the world.
The fact the "they used PCs for administration because they didn't believe in Amiga" is just a BS, they were already using mainly VAX/Unix machines and still some PETs, as a lot of other companies at that time. It was a nonsense to replace the IT infrastructure with machines that lacked of software. Amiga never had serious productivity software, I mean, to manage a billion dollar company like Commodore.

Last edited by dipsomania on 09-Sep-2025 at 09:45 AM.
Last edited by dipsomania on 09-Sep-2025 at 09:43 AM.
Last edited by dipsomania on 09-Sep-2025 at 09:41 AM.
Last edited by dipsomania on 09-Sep-2025 at 09:39 AM.

@dipsomania

propably both sides involved like always

but you seem to give the main blame to the developers but I have doubts there

take the C65... the project was finished and then stopped by marketing after already first batch produced to be delivered to developers

how can sthi happen? What exactly was the orders of the hardware developers? How was communication in the months before?

that it would become incompatible to the old C64 was certainly known during development phase. How can it happen that this was no topic before?

I think there were serious problems in management and organization of Commodore

Last edited by OlafS25 on 09-Sep-2025 at 02:46 PM.
Last edited by OlafS25 on 09-Sep-2025 at 02:45 PM.

@topic

Why is this even a debate? The role of Commodore's management in the company's failures and ultimate demise is definitive and complete.
The proverbial buck stops with leadership. Commodore's was, to put it mildly, dysfunctional.

Before you cry a river, it has been my experience that the leadership of most global public companies is dysfunctional, though the level of dysfunction varies and is only tested during a crisis. Apple were in the same boat during that time. However, their momentum generated enough inertia in key markets that in an opportune time they were deemed worthy of a bailout by Microsoft.

Engineers are never to blame. If the organisation lacks or has the incorrect mix of engineers to deliver on its goals, it's the responsibility of management to hire the correct mix of engineers, and set them to their tasks.

End of story.

_________________
All the way, with 68k

@dipsomania

Quote:

I don't think so. After reading the Bagnall's books, as everybody has already done here I believe, it seems the truth is in the middle as usual. Yes, they were brilliant and talented engineers but also youngsters, full of egos and weird, that lead to nonsense engineering decisions - because they felt cool putting questionable stuff (like CP/M in C128...)

C128's Zilog Z80 with CP/M was part of Zilog Z8000-based C900 (Unix clone) direction, and Commodore obtained Zilog's second source license. This direction shows Commodore management has an "anti-toy" bias.

C128 has aging C64 gaming hardware with "productivity" Z80 CP/M mode. The mindset mirrored Amiga's ECS " productivity mode" while "toy" graphics mode was neglected.

Meanwhile, IBM VGA upgraded both productivity mode's 640x480p with 16 colors and "toy" mode's VGA 320x200/240 256 colors with hardware scrolling (eat that Jack Tramiel) and latch data move/copy engine!

VGA's 640x480p and 320x200 256 colors use cases was recycled from the 1986 release MCGA.

Both C128's Z80 CP/M mode and Z8001 CPU/Z8010 MMU C900 used MOS Technology 8563 display controller.

C900 supported the optional Z8070 floating point solution running at 24 MHz.

Zilog admitted their mistake was dumping backwards compatibility with Z80 CP/M, which annoyed many business users. Zilog's mistake benefited IBM's PC and Intel's x86.

Intel learns the lesson from Zilog and made sure 80286 and 80386 were backward compatible with 8086/8086.

Intel Itanium IA-64 had IA-32 mode, which has inferior performance when compared to AMD64. AMD/NVIDIA duo made sure PC gaming had preferred X86-64 direction e.g. Crysis X64 with NVIDIA performant drivers for Windows XP X64. AMD/NVIDIA recycles Microsoft's Game for Windows tactics that defeated IBM OS/2 Warp in SOHO markets.

Gaming PC/workstation PC X64's larger production scale would outnumber Itanium (IA-64/IA-32) PC workstations. AMD64's economies of scale killed Itanium.

_________________

@agami

The credit for Steve Jobs' return to Apple is Gil Amelio. It took another Steve Jobs driven MacOS development (via NextSTEP) to replace the classic MacOS.

Steve Jobs wasn't perfect, but his passion for a good GUI end-user experience is the reason for MacOS X and iOS platforms.

Steve Jobs used his personal relationship with Bill Gates for the rescue package.

Bill Gates indirectly owns US$1.8 billion (AU$2.7 billion) worth of Apple shares via Warren Buffett's Berkshire Hathaway.

Last edited by Hammer on 10-Sep-2025 at 02:10 AM.

_________________

@Kronos

Quote:

Kronos wrote: @cdimauro Quote: cdimauro wrote: Please, tell me more about the Slow Mem which was worse than Chip and Fast Mem, a) it wasn't worse, it combined the flaws of the 2

So, it was worse.

You had no advantage of the two types of memory, and only their disadvantages.
Quote:

b) it was a significant cost cutting measure (no need for a separate controller)

You didn't need a full memory controller: the trick was used only to implement the refresh logic of DRAM.

They could have just used this part for the Slow mem, while keeping the separate buses on all other cases.
Quote:

that had minimal performance impact on 68000 systems.

It's the opposite: it had a HUGE impact on performance, because the 68000 was blocked all times that there was a chip memory access.
Quote:

c) it was only meant to be in the low cost A500, the A2000A had real 512k FAST in the MMU slot. The A2000B was a rushed fix for the A2000A failure

I had an A2000 (I don't know if it was a 2000A or 2000B), and there was nothing like that: it only had 512kB of Slow mem, and no way to convert / use it as Fast mem.
Quote:

d) once Agnus went 1MB C= never again shipped a system with SLOW. A500 of that time needed to have just one Jumper cut when upgraded with a trapdoor card.

It was too late: the market was already saturated, and developers had to develop games taking into account Slow mem, and not the extra 512kB Chip Mem for games.
Quote:

Plenty worse WFTs in pretty much every piece of HW of that period.

But this HEAVILY crippled the Amiga games, unfortunately.

@OlafS25

Quote:

OlafS25 wrote: @cdimauro in my view management was most to blame. The main problem was, there was no real longterm plan where to go. Amiga was just one product, 16bit home computer. Most money was earned with C64. And they also sold PCs but they were too small for that. They certainly had losses there. But in the company they used PCs for administration, not amigas. For them the amiga was just a toy system they could sell at young people, not a key technology for Commodore to survive. Additionally they did a lot of costly wrong decisions. certainly engineers have a tendency to overengineer and develop products nobody want to buy being too expensive or set wrong priorities. But even for that is management to blame. There must be competent project managers who oversee a project and report to the top management.

I share dipsomania post on that.

@Hammer

Quote:

Hammer wrote: @cdimauro Quote: As you reported from the book, no: it wasn't the case. Managers knew PC graphics (and Mac graphics as well), and they started meetings for discussing the specs of the Next Big Thing exactly with the purpose of contrasting the competition which became very serious. Managers rejected an immediate solution for PC graphics handling.

On the exact contrary: they immediately started seeking for solutions just after IBM's PS/2 -> VGA and Apple's Mac II.
Quote:

Akiko's C2P is abstracted by an API anyway.

Akiko was completely out of context here: it's from 1993, whereas the context was around 1987.
Quote:

CDTV-CR/CD32's multimedia team doesn't have the mandate to modify the core Amiga graphics design.

As explained on the other thread (and please keep it there the discussions), it's ok: their scope is completely different, and they haven't required any chipset change.
Quote:

The Next Big Thing like the AAA project is tainted with a crazy high-end 64-bit VRAM-equipped workstation when the mainstream SVGA with improved FPM DRAM is the immediate competitive threat.

That was "thanks" to the engineers, which weren't able to deliver a good, realistic / affordable product.

Pay attention that the 1000x800x256 colours on 1987 required the same amount of crazy high-end bandwidth -> completely unrealistic (at least for the Amiga market).
Quote:

Commodore HR failed to hire enough chip engineers when there was a single engineer bottleneck for C65, AAA, and ECS Agnus B. Brain drain is real.

Right. Management had its faults.
Quote:

For CeBit 1988, Commodore PC engineers and Herni Rubin were impressed with ATI's 1988 era SVGA chipsets(e.g. VGA Wonder), so they demoed them instead of the Amiga 500 with a flicker fixer, even though it was Herni Rubin who issued the monochrome hi-res directive in 1986.

And it was Rubin which started the meetings for the Next Big Thing on 1987.

The problem is that, after more than one year of discussions, there's no concrete outcome from the Amiga engineers, which had completely different visions. So, they failed even on this paper work...
Quote:

System engineering leadership ignored the following arguments: 1. 1987 era 32-bit chipset evolution argument = ignored.

Not needed. 256 colours where needed.

14Mhz should have been the next step, and 32-bit the following one.
Quote:

2. 1987 era 8-bitplane chipset evolution argument = ignored.

Correct, and it would have been very easy-
Quote:

3. 1987 era 16-bit (65,536 colors) graphics chipset evolution before moving into 24-bit graphics argument = ignored.

You need double memory bandwidth for that -> 14Mhz clock or 32-bit bus. 14Mhz should have been the logic choice.

However, 65536 colours were not needed for the Amiga platform. AT THE TIME.
Quote:

The Unix camp wanted the high-end workstation. The Amiga Unix (Make AT&T Great Again) project is another C900 mk2 distraction that departed from Commodore's core market and mass production strengths.

That's high-end -> not in the primary Amiga scope / market target.
Quote:

There are many other project distractions e.g. C64GS with C65's cartridge size support, PC genlock, various mice control devices, A2410 TiGA, A2024 monitor with 5000 production scale, x86 sidecart for A1000, system integration Commodore PC clones, and many others.

Right.
Quote:

Mac LC II's best-selling Mac model for 1992 was done with a modest 256-color and 65,536-color display modes.

Which prove my above points.
Quote:

Commodore management failed to replace Jay Miller's Amiga OCS and CSG's C64 core business.

Correct.

@OlafS25
@agami

You're all right guys, I'm not putting the blame mostly on the developers/engineers. Amiga engineers did a lovely job and the Animals were heroic and capable on their own, no doubts.
But I don't longer buy the story that guys like Mehdi Ali, Marshall Smith et similia, killed the company on purpose. In hindsight they did many bad mistakes, of course, as for example the Animals when they decided on their own to build the infamous C16/Plus-4/C128, fighting with the so called "middle management". In this specific case, for me it was the starting point of the defeat of Commodore, and here the management was only culprit to be absent or too weak, in my opinion.
However, in more than 35 years I'm following Commodore story, I read tons of interviews from engineers (as everybody here I think), but to be honest very little from the managers... I would like to hear their point of views, but I believe it wouldn't be possible unfortunately, they almost died long time ago...

About the hiring process, you're right, but in Commodore in the 80s there wasn't a HR department at all. Usually people were hired by guys like Jeff Porter and in some cases even directly by engineers in person, like Bil Herd gave the ok to Dave Haynie if I'm not wrong.

Just saying everybody had their load of shame, somebody more, somebody else a bit less, from the top to the very bottom. The entire company culture in Commodore was sick and mental, compared to its competitors (the speed bumps story just to mention an episode).

Last edited by dipsomania on 10-Sep-2025 at 11:57 AM.
Last edited by dipsomania on 10-Sep-2025 at 11:57 AM.

@cdimauro

Quote:

cdimauro wrote: It's the opposite: it had a HUGE impact on performance, because the 68000 was blocked all times that there was a chip memory access..

No it wasn't, it was a minor impact in normal operation (when the chipset stayed within it's half of the cycles) and only got worse if and when when stressed.

How much would a separate bus cost at the time? 5 maybe 10$, sounds small but on a system with high volume and low margins that was a nogo.

If your A2000 had slow it was a B. And sure slow had no place in such a preices computer but it was the best they could do in short timeframe.

Quote:

But this HEAVILY crippled the Amiga games, unfortunately.

No it didn't, most games were constrained by the size of chip (1MB CHIP for all A500 from the start, now that would made difference), the overall RAM available (the 512+512 requirement really only took hold from 1990 onwards) and the lack of space (and speed) per disk. I'd say spending the above 10$ on giving the A500 a full speed HD floppy would have had a much bigger impact on games.

But again f###ups are a reality, the A20 gate, saving 5ct on C64 giving it the same hyper slow floppy support as the a PET. SinclairQL running an 68k on a 8Bit bus and so on.

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada