The bad evolution of the Amiga chipsets was also one of the reasons why this wonderful platform fell into oblivion.
The engineers systematically put all the blame on the (unsuccessful) management, but they are by no means without sin, as we can see with regard to the successor of the original chipset, the so-called ECS.
English: https://www.appuntidigitali.it/20663/amiga-ecs-and-the-deception-of-read-my-lips-no-new-chips/
Italian: https://www.appuntidigitali.it/20617/amiga-ecs-e-linganno-di-read-my-lips-no-new-chips/

@cdimauro
Thanks. It's another good read. I agree that ECS was not good enough for how late it arrived. The colors needed to be increased at least. Not only was it generally not a target for software enhancements but it was not enough of a hardware upgrade to sell ECS Amigas. AGA was far from perfect and late but at least it increased the colors and chip memory bandwidth. ECS needed to double the OCS chip memory bandwidth in order to add more colors though. You don't really mention how this would be done except for saying that the 68020 can access chip memory every 2 cycles instead of 4 cycles of the CPU with the 68000. Are you advocating for upgrading to a 68020 with ECS+ to give more chip memory accesses? Would this have effectively doubled chip memory bandwidth like it sounds? Are there articles or documentation somewhere that talks about this?

I'm not sold on the conclusion giving partial blame for the disappointing Amiga chipset enhancements to the engineers. It may be true but there is a lack of evidence given for it. C= upper management was so bad that they created an environment which suppressed creativity and innovation. Suggesting using VRAM was likely to get the engineer in the dog house like the Amiga Corporation team and even Jay Miner. Being told "Read my lips - no new chips" is hostile and also suppressive of initiative. Ranger was cancelled, AGA was delayed, the DSP was cancelled. These were engineering initiatives that were suppressed by upper management.

@cdimauro

The Vampire V4 Series of upgrades and stand alone upgraded all the chips (SAGA) in an awesome way even the CPU (68080). Now people need to buy even more of them.

Last edited by DiscreetFX on 26-Oct-2023 at 04:19 AM.

_________________
Sent from my Quantum Computer.

@matthey

Quote:

matthey wrote: @cdimauro Thanks. It's another good read. I agree that ECS was not good enough for how late it arrived. The colors needed to be increased at least. Not only was it generally not a target for software enhancements but it was not enough of a hardware upgrade to sell ECS Amigas. AGA was far from perfect and late but at least it increased the colors and chip memory bandwidth. ECS needed to double the OCS chip memory bandwidth in order to add more colors though. You don't really mention how this would be done except for saying that the 68020 can access chip memory every 2 cycles instead of 4 cycles of the CPU with the 68000. Are you advocating for upgrading to a 68020 with ECS+ to give more chip memory accesses?

That would have been good: it was 1990!

However no: I think that keeping a 68000 for low-end (more game-oriented systems) could still be the way to go IF the chipset was updated accordingly.
Quote:

Would this have effectively doubled chip memory bandwidth like it sounds? Are there articles or documentation somewhere that talks about this?

I had an idea about how to keep the whole system 16-bit (so, still using a 68000 and with the same 16-bit chips), but I don't know if the memories available at the time were able to run at double the frequency (14Mhz instead of 7Mhz).

This is something which I haven't covered on my article because I lacked this information (and I do NOT want to put fantasies on it: I prefer to keep real and technical).

But granted that it was possible, I had some simple ideas, also very easy to implement and taking just a bunch of transistors, to double the system bandwidth AND let the Blitter run at more than double its speed.
This would have put the Amiga on a much better and competitive position with the other systems of the tim.
Quote:

I'm not sold on the conclusion giving partial blame for the disappointing Amiga chipset enhancements to the engineers. It may be true but there is a lack of evidence given for it. C= upper management was so bad that they created an environment which suppressed creativity and innovation. Suggesting using VRAM was likely to get the engineer in the dog house like the Amiga Corporation team and even Jay Miner. Being told "Read my lips - no new chips" is hostile and also suppressive of initiative. Ranger was cancelled, AGA was delayed, the DSP was cancelled. These were engineering initiatives that were suppressed by upper management.

Yes, but the engineers had still the time and resources to work on completely useless stuff.

The same happened to the CD32 with the Akiko: an horrible patch to implement the 8-bit packed/chunky graphic.

Those guys really lacked creativity at all.

@DiscreetFX

Quote:

DiscreetFX wrote: @cdimauro The Vampire V4 Series of upgrades and stand alone upgraded all the chips (SAGA) in an awesome way even the CPU (68080). Now people need to buy even more of them.

Too much expensive. That's why the Amiga Mini sold like cakes...

@cdimauro

Vampire V4 prices are very fair for what you get. Imagine what a AAA Amiga would have cost back in the day.

_________________
Sent from my Quantum Computer.

@DiscreetFX

Hmmm... I'm not sure I can justify Vampire prices when the PiStorm exists! The wife rolled her eyes at the PiStorm32 outlay etc even though I found a reasonable CM4 in the middle of the supply shortage!

_________________
"Art challenges technology. Technology inspires the art."
John Lasseter, Co-Founder of Pixar Animation Studios

PiStorm doesn’t offer anything in terms of Amiga chipset, so…

On the other hand, SAGA…
http://www.apollo-core.com/knowledge.php?note=5768

(SAGA is getting old!)

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

cdimauro Quote:

That would have been good: it was 1990! However no: I think that keeping a 68000 for low-end (more game-oriented systems) could still be the way to go IF the chipset was updated accordingly.

I defiantly would have been wanting a CPU upgrade by 1990. There was a 68000@16.67MHz available by that time but the 68020 offered more performance and value which is probably why C= eventually moved to it with AGA. If chip memory bandwidth could have been doubled with the 68020 by doubling chip memory accesses compared to the 68000, that is very compelling and makes the 68020 appealing earlier. The tiny 68020 instruction cache would have reduced CPU memory traffic too.

cdimauro Quote:

I had an idea about how to keep the whole system 16-bit (so, still using a 68000 and with the same 16-bit chips), but I don't know if the memories available at the time were able to run at double the frequency (14Mhz instead of 7Mhz). This is something which I haven't covered on my article because I lacked this information (and I do NOT want to put fantasies on it: I prefer to keep real and technical). But granted that it was possible, I had some simple ideas, also very easy to implement and taking just a bunch of transistors, to double the system bandwidth AND let the Blitter run at more than double its speed. This would have put the Amiga on a much better and competitive position with the other systems of the time.

I doubt clocking up the memory was a good idea. The Acorn Archimedes did this

https://en.wikipedia.org/wiki/Acorn_Archimedes Quote:

The A540, introduced in late 1990, was an anticipated consequence of Acorn's Unix workstation development, offering the same general specification as Acorn's R260 Unix workstation (running RISC iX) but without built-in Ethernet support and running RISC OS 2 instead of Unix. It was Acorn's first machine to be fitted with the ARM3 processor as standard, supporting up to 16 MB of RAM, and included higher speed SCSI and provision for connecting genlock devices. The memory access frequency was raised to 12 MHz in the A540, compared to 8 MHz in earlier models, thus providing enhanced system performance over earlier models upgraded with ARM3 processors. The hardware design featured memory modules, each providing their own memory controller and 4 MB of RAM, and a processor module providing the ARM3 and a slot for a floating point accelerator (FPA) chip, the latter offering the possibility (subsequently unrealised) of processor upgrades. The FPA, replacing Acorn's previous floating point podule, was scheduled to be available in 1991.[79] Much delayed, the FPA finally became available in 1993. ... These new models utilised the first ARM system-on-chip - the ARM250 microprocessor - a single-chip design including the functionality of an ARM2 (or ARM3 without cache), the IOC1, VIDC1a and MEMC1a chips all "integrated into a single giant chip" and fabricated using a 1 micron process. The ARM250, running at a higher 12 MHz clock frequency and used in conjunction with faster 80ns memory chips, compared to the 8 MHz of the ARM2 and the 125ns memory of the A3000, gave a potential 50% performance increase over such older systems, achieving a reported 7 MIPS. ... By employing a 16 MHz clock signal, as envisaged by Acorn in the design of the A3010, in conjunction with dynamic RAM devices with a 70 ns access time, the upgrade provided a total of 4 MB of RAM and a 40 percent performance improvement. Unlike standard RAM upgrades, the turbo upgrade needed to be fitted at a suitable facility, and the board was priced slightly higher than a standard RAM upgrade at £129 plus VAT. A "super turbo" version of the board with 20 MHz crystal and 45 ns dynamic RAM devices was reviewed and apparently available subject to component availability, reportedly achieving 12.25 MIPS.

At 14MHz, at least 80ns and likely 70ns or 60ns memory would have been necessary judging by this (Acorn was likely more conservative about increasing memory clock speeds than 3rd parties which may have even overclocked memory). I have heard claims that the high memory cost was one of the reasons why the Archimedes failed. C= wasn't completely crazy for wanting to avoid more expensive memory. Even small caches significantly reduced CPU memory traffic and increased performance. CPU caches add jitter compared to faster memory and do not increase chipset performance but increasing memory clock speeds with CPU clock speeds was not practical for long.

cdimauro Quote:

Yes, but the engineers had still the time and resources to work on completely useless stuff. The same happened to the CD32 with the Akiko: an horrible patch to implement the 8-bit packed/chunky graphic. Those guys really lacked creativity at all.

The engineers probably figured out what C= would tolerate after having their hard work cancelled enough times. C= upper management may have asked for the cheapest possible C2P support like C= used the cheapest possible CPUs and memory. Top down authoritarian leadership stifles creativity and replaces it with loyalty.

Last edited by matthey on 26-Oct-2023 at 08:54 PM.
Last edited by matthey on 26-Oct-2023 at 07:18 PM.

@BigD

You won’t be able to play the new Menace, Apollo Invaders and many more games on PiStorm.

_________________
Sent from my Quantum Computer.

@cdimauro

Quote:

cdimauro wrote: The bad evolution of the Amiga chipsets was also one of the reasons why this wonderful platform fell into oblivion. The engineers systematically put all the blame on the (unsuccessful) management, but they are by no means without sin, as we can see with regard to the successor of the original chipset, the so-called ECS. English: https://www.appuntidigitali.it/20663/amiga-ecs-and-the-deception-of-read-my-lips-no-new-chips/ Italian: https://www.appuntidigitali.it/20617/amiga-ecs-e-linganno-di-read-my-lips-no-new-chips/

The ECS design consideration existed with the 1989-era Amiga 500 Rev 6 motherboard's 2 MB Chip RAM jumper configuration via Amiga 3000's 8372AB or 8372B Agnus chips and higher density memory chips.

ECS recycled 68000-era 16-bit memory controller and DMA engine designs from OCS.

Super Denise's very low color display productivity modes were obsolete in 1990.

Besides the 1 MB vs 2 MB Chip RAM difference, 1989-era Amiga 500's 1 MB ECS Agnus can support all the features of 2 MB ECS Agnus e.g. Blitter capable of handling areas up to 32768 x 32768 pixels (instead of the 1024 x 1024 of OCS Agnus).

1989 Amiga 500 Rev 6's ECS 2 MB Chip RAM consideration is a reserve capability.

Your argument is based on released retail products, not on Commodore engineers' point of view.


The Amiga 3000 comes with
1. a Ramsey chip which is a 32-bit memory controller with 25 Mhz clock speed.
2. a Super DMAC that performs DMA transfer functions.

Commodore engineers were not allowed to build accelerated raster graphics functions based on this improved memory controller intellectual property.

AGA's Lisa's raster operations will perform better with this improved memory controller IP.

Meanwhile, the PC world is not limited by Commodore's management limitations e.g. Tseng Labs ET4000AX was released in 1989.

https://dosdays.co.uk/topics/retro_review_et4000_pt1.php

With 1 MB of video memory onboard, an ET4000 could display resolutions up to 1024 x 768 in 256 colours.
...
In both of these, the ET4000 was marginally beaten by two cards that ran 70ns DRAMs - a Western Digital WDC90C30 and a Trident TVGA-8900CL-B


https://en.wikipedia.org/wiki/Tseng_Labs_ET4000
support for up to 1MB of memory with a ~16-bit VRAM or ~32-bit DRAM memory data bus width.

Trident TVGA-8900CL-B was released in 1992.
https://www.vgamuseum.info/index.php/cpu/item/434-trident-tvga8900cl-b


Western Digital WDC90C30 LR was released in 1991. https://www.vgamuseum.info/index.php/cpu/item/484-western-digital-wd90c30-lr



https://archive.org/details/bub_gb_hqQJaNzN9IcC/page/n603/mode/2up
PC Mag 1992-08, page 604 of 664,
Diamond Speedstar 24 (ET4000AX ISA) had $169 USD retail asking price in 1992.

In 1992, I have both a 386DX-33 PC with ET4000AX and an Amiga 3000 25 Mhz with Kickstart 2.04 ROM model.

386DX-33 PC with ET4000AX can run Doom in late 1993 and the Gaming PC market has a larger install base!

Amiga's AGA arrived in Q4 1992 in limited production numbers. AGA's new 32-bit chip is Lisa while Alice recycled OCS/ECS Agnus's obsolete 16-bit Blitter. Alice allows full-bus width access for bitplane DMA.

Remember, the early NTSC Amiga 1000 didn't come with a 6-bitplane 64 color EHB mode, and the 1987-era Amiga Ranger's 128 color is a natural evolution. OCS has color display mode improvements.

Commodore's A2410 TIGA did nothing for Amiga chipset evolution.

I have Amiga 500 Rev 5 and Rev 6 motherboards, along with Amiga 1200 Rev 1D4.

Last edited by Hammer on 27-Oct-2023 at 01:59 AM.
Last edited by Hammer on 27-Oct-2023 at 01:37 AM.
Last edited by Hammer on 27-Oct-2023 at 01:34 AM.
Last edited by Hammer on 27-Oct-2023 at 01:26 AM.
Last edited by Hammer on 27-Oct-2023 at 01:05 AM.
Last edited by Hammer on 27-Oct-2023 at 12:59 AM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@kolla
Quote:

kolla wrote: PiStorm doesn’t offer anything in terms of Amiga chipset, so… On the other hand, SAGA… http://www.apollo-core.com/knowledge.php?note=5768 (SAGA is getting old!)

PiStorm and Emu68 developers are planning standalone PiStorm with multiple low-cost FPGA for the Amiga chipset clones.

Apollo-Core's SAGA hardware extensions are design forks from Commodore's AGA standard just like PC's multiple incompatible SVGA extensions until VESA BIOS's establishment and "Design for Windows" 2D accelerators i.e. Microsoft's RTG solution.

Buffee (Renee Cousins) is planning for AGA-on-ECS drop-in replacement chips.

Last edited by Hammer on 27-Oct-2023 at 03:43 AM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@DiscreetFX

Quote:

DiscreetFX wrote: @BigD You won’t be able to play the new Menace, Apollo Invaders and many more games on PiStorm.

PiStorm-Emu68 can play Star Wars "Dark Force" RTG better than Apollo Core's Vampire Standalone (AC68080 V4, SAGA).

Last edited by Hammer on 27-Oct-2023 at 01:18 AM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@matthey

Quote:

matthey wrote: @cdimauro Thanks. It's another good read. I agree that ECS was not good enough for how late it arrived. The colors needed to be increased at least. Not only was it generally not a target for software enhancements but it was not enough of a hardware upgrade to sell ECS Amigas. AGA was far from perfect and late but at least it increased the colors and chip memory bandwidth. ECS needed to double the OCS chip memory bandwidth in order to add more colors though. You don't really mention how this would be done except for saying that the 68020 can access chip memory every 2 cycles instead of 4 cycles of the CPU with the 68000. Are you advocating for upgrading to a 68020 with ECS+ to give more chip memory accesses? Would this have effectively doubled chip memory bandwidth like it sounds? Are there articles or documentation somewhere that talks about this? I'm not sold on the conclusion giving partial blame for the disappointing Amiga chipset enhancements to the engineers. It may be true but there is a lack of evidence given for it. C= upper management was so bad that they created an environment which suppressed creativity and innovation. Suggesting using VRAM was likely to get the engineer in the dog house like the Amiga Corporation team and even Jay Miner. Being told "Read my lips - no new chips" is hostile and also suppressive of initiative. Ranger was cancelled, AGA was delayed, the DSP was cancelled. These were engineering initiatives that were suppressed by upper management.

Commodore management allowed C65's 256 colors with a 4096 color palette VIC-III chipset R&D to be completed before canceling C65's R&D in December 1990. The project was canceled by Commodore's chairman Irving Gould in 1991.

C65's 256 color mode is a planar (bit-planes) pixel format on Amiga 500's level Chip RAM bandwidth!

The Commodore 65 prototype was later reverse-engineered and released as MEGA65 in 2020.

Commodore's R&D resources were wasted.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@DiscreetFX

Quote:

DiscreetFX wrote: @cdimauro Vampire V4 prices are very fair for what you get.

I don't think so: it's the most expensive solution for what you get.
Quote:

Imagine what a AAA Amiga would have cost back in the day.

Which isn't the case. SuperAGA isn't an AAA (re)implementation, rather a new project (and badly designed, I've to add).


@BigD

Quote:

BigD wrote: @DiscreetFX Hmmm... I'm not sure I can justify Vampire prices when the PiStorm exists! The wife rolled her eyes at the PiStorm32 outlay etc even though I found a reasonable CM4 in the middle of the supply shortage!

PiStorm doesn't "emulate" the chipset. But a good part of the system, yes (RTG, AHI, 3D acceleration is possible).


@kolla

Quote:

kolla wrote: PiStorm doesn’t offer anything in terms of Amiga chipset, so…

Right. It's not its purpose.

However it offers also much more: see above).
Quote:

On the other hand, SAGA… http://www.apollo-core.com/knowledge.php?note=5768 (SAGA is getting old!)

To open source it? Where is the repo with the source codes?

@matthey

Quote:

matthey wrote: cdimauro Quote: That would have been good: it was 1990! However no: I think that keeping a 68000 for low-end (more game-oriented systems) could still be the way to go IF the chipset was updated accordingly. I defiantly would have been wanting a CPU upgrade by 1990.

As a coder, me too, for sure. But we've to consider that the company has to gain money selling the products.

A low-end 16-bit solution at the time was still good enough for Amiga's market (games). With a 32 bit solution for professionals.

So, basically the Amiga 500 + 2000.
Quote:

There was a 68000@16.67MHz available by that time

Exactly. Maybe a 68010 would be a better solution thanks to its DBRA hardware loop, which helped a bit the performances.
Quote:

but the 68020 offered more performance and value which is probably why C= eventually moved to it with AGA.

A 32 bit ECS with my suggestions could have helped, before going to a full 32 bit chipset implementation with the AGA.
Quote:

If chip memory bandwidth could have been doubled with the 68020 by doubling chip memory accesses compared to the 68000, that is very compelling

That was my idea, in fact. Thanks to doubling the clock for memory (and the CPU, of course: see above).
Quote:

and makes the 68020 appealing earlier. The tiny 68020 instruction cache would have reduced CPU memory traffic too.

Better to leave it for the high-end. You also have to consider that a motherboard with 32 bit data busses is much more complicated and expensive compared to one with 16-bit data busses.

If you want to be aggressive on the low-end market, especially competing also with consoles, you absolutely need to cut the production costs.
Quote:

cdimauro Quote: I had an idea about how to keep the whole system 16-bit (so, still using a 68000 and with the same 16-bit chips), but I don't know if the memories available at the time were able to run at double the frequency (14Mhz instead of 7Mhz). This is something which I haven't covered on my article because I lacked this information (and I do NOT want to put fantasies on it: I prefer to keep real and technical). But granted that it was possible, I had some simple ideas, also very easy to implement and taking just a bunch of transistors, to double the system bandwidth AND let the Blitter run at more than double its speed. This would have put the Amiga on a much better and competitive position with the other systems of the time. I doubt clocking up the memory was a good idea. The Acorn Archimedes did this https://en.wikipedia.org/wiki/Acorn_Archimedes Quote: The A540, introduced in late 1990, was an anticipated consequence of Acorn's Unix workstation development, offering the same general specification as Acorn's R260 Unix workstation (running RISC iX) but without built-in Ethernet support and running RISC OS 2 instead of Unix. It was Acorn's first machine to be fitted with the ARM3 processor as standard, supporting up to 16 MB of RAM, and included higher speed SCSI and provision for connecting genlock devices. The memory access frequency was raised to 12 MHz in the A540, compared to 8 MHz in earlier models, thus providing enhanced system performance over earlier models upgraded with ARM3 processors. The hardware design featured memory modules, each providing their own memory controller and 4 MB of RAM, and a processor module providing the ARM3 and a slot for a floating point accelerator (FPA) chip, the latter offering the possibility (subsequently unrealised) of processor upgrades. The FPA, replacing Acorn's previous floating point podule, was scheduled to be available in 1991.[79] Much delayed, the FPA finally became available in 1993. ... These new models utilised the first ARM system-on-chip - the ARM250 microprocessor - a single-chip design including the functionality of an ARM2 (or ARM3 without cache), the IOC1, VIDC1a and MEMC1a chips all "integrated into a single giant chip" and fabricated using a 1 micron process. The ARM250, running at a higher 12 MHz clock frequency and used in conjunction with faster 80ns memory chips, compared to the 8 MHz of the ARM2 and the 125ns memory of the A3000, gave a potential 50% performance increase over such older systems, achieving a reported 7 MIPS. ... By employing a 16 MHz clock signal, as envisaged by Acorn in the design of the A3010, in conjunction with dynamic RAM devices with a 70 ns access time, the upgrade provided a total of 4 MB of RAM and a 40 percent performance improvement. Unlike standard RAM upgrades, the turbo upgrade needed to be fitted at a suitable facility, and the board was priced slightly higher than a standard RAM upgrade at £129 plus VAT. A "super turbo" version of the board with 20 MHz crystal and 45 ns dynamic RAM devices was reviewed and apparently available subject to component availability, reportedly achieving 12.25 MIPS. At 14MHz, at least 80ns and likely 70ns or 60ns memory would have been necessary judging by this (Acorn was likely more conservative about increasing memory clock speeds than 3rd parties which may have even overclocked memory).

The original Amiga had 140ns memories AFAIR but I've checked and in 1990 70ns memories were already available for sure. I don't know the prices anyway.
Quote:

I have heard claims that the high memory cost was one of the reasons why the Archimedes failed. C= wasn't completely crazy for wanting to avoid more expensive memory.

You don't need to add 4MB of 70ns memory to the 16-bit system: 1MB (all Chip RAM) would have been a good price spot, with the usual possibility to expand it further via the trapdoor.
Quote:

Even small caches significantly reduced CPU memory traffic and increased performance. CPU caches add jitter compared to faster memory and do not increase chipset performance but increasing memory clock speeds with CPU clock speeds was not practical for long.

You don't need it if you've a good chipset, like the Amiga had, and the CPU was mostly busy instructing the chips on what to do.

Again, you've to think about games being the primary goal for such low-end platform.
Quote:

cdimauro Quote: Yes, but the engineers had still the time and resources to work on completely useless stuff. The same happened to the CD32 with the Akiko: an horrible patch to implement the 8-bit packed/chunky graphic. Those guys really lacked creativity at all. The engineers probably figured out what C= would tolerate after having their hard work cancelled enough times. C= upper management may have asked for the cheapest possible C2P support like C= used the cheapest possible CPUs and memory. Top down authoritarian leadership stifles creativity and replaces it with loyalty.

Well, from what I know Akiko C2P came out as an idea during one lunch: great example of creativity.

I've written another article some time ago about this topic, showing a native packed/chunky mode implementation for AGA which would have been simple enough to implement and also cheap (not so many transistors added).

@Hammer

Quote:

Hammer wrote: @cdimauro Quote: cdimauro wrote: The bad evolution of the Amiga chipsets was also one of the reasons why this wonderful platform fell into oblivion. The engineers systematically put all the blame on the (unsuccessful) management, but they are by no means without sin, as we can see with regard to the successor of the original chipset, the so-called ECS. English: https://www.appuntidigitali.it/20663/amiga-ecs-and-the-deception-of-read-my-lips-no-new-chips/ Italian: https://www.appuntidigitali.it/20617/amiga-ecs-e-linganno-di-read-my-lips-no-new-chips/ The ECS design consideration existed with the 1989-era Amiga 500 Rev 6 motherboard's 2 MB Chip RAM jumper configuration via Amiga 3000's 8372AB or 8372B Agnus chips and higher density memory chips. ECS recycled 68000-era 16-bit memory controller and DMA engine designs from OCS. Super Denise's very low color display productivity modes were obsolete in 1990. Besides the 1 MB vs 2 MB Chip RAM difference, 1989-era Amiga 500's 1 MB ECS Agnus can support all the features of 2 MB ECS Agnus e.g. Blitter capable of handling areas up to 32768 x 32768 pixels (instead of the 1024 x 1024 of OCS Agnus). 1989 Amiga 500 Rev 6's ECS 2 MB Chip RAM consideration is a reserve capability. Your argument is based on released retail products, not on Commodore engineers' point of view.

My argument is about what was really needed an achievable with the same constraints that engineers had for ECS.
Quote:

The Amiga 3000 comes with 1. a Ramsey chip which is a 32-bit memory controller with 25 Mhz clock speed. 2. a Super DMAC that performs DMA transfer functions. Commodore engineers were not allowed to build accelerated raster graphics functions based on this improved memory controller intellectual property.

And it wasn't needed to do, in fact, with my suggestions: the chipset remained almost the same.
Quote:

AGA's Lisa's raster operations will perform better with this improved memory controller IP.

AGA couldn't compete with my further idea of memory running at double the frequency, because in this case the entire system gained both double frequency and double memory bandwidth.

It means that the Blitter had the chance to run at 14Mhz as well (and Copper too, of course).

And the display subsystem also had the chance to take benefit of it with some tricky ideas, opening the way for 256 colours in high-resolution, for example.
Quote:

Meanwhile, the PC world is not limited by Commodore's management limitations e.g. Tseng Labs ET4000AX was released in 1989. https://dosdays.co.uk/topics/retro_review_et4000_pt1.php With 1 MB of video memory onboard, an ET4000 could display resolutions up to 1024 x 768 in 256 colours. ... In both of these, the ET4000 was marginally beaten by two cards that ran 70ns DRAMs - a Western Digital WDC90C30 and a Trident TVGA-8900CL-B https://en.wikipedia.org/wiki/Tseng_Labs_ET4000 support for up to 1MB of memory with a ~16-bit VRAM or ~32-bit DRAM memory data bus width.

Right, and? 70ns DRAM could have been used with my suggestion giving a HUGE boost to the Amiga platform with MINIMAL changes to the chipset, so don't breaking the "no new chips" dictact.
Quote:

Trident TVGA-8900CL-B was released in 1992. https://www.vgamuseum.info/index.php/cpu/item/434-trident-tvga8900cl-b Western Digital WDC90C30 LR was released in 1991. https://www.vgamuseum.info/index.php/cpu/item/484-western-digital-wd90c30-lr https://archive.org/details/bub_gb_hqQJaNzN9IcC/page/n603/mode/2up PC Mag 1992-08, page 604 of 664, Diamond Speedstar 24 (ET4000AX ISA) had $169 USD retail asking price in 1992. In 1992, I have both a 386DX-33 PC with ET4000AX and an Amiga 3000 25 Mhz with Kickstart 2.04 ROM model. 386DX-33 PC with ET4000AX can run Doom in late 1993 and the Gaming PC market has a larger install base! Amiga's AGA arrived in Q4 1992 in limited production numbers. AGA's new 32-bit chip is Lisa while Alice recycled OCS/ECS Agnus's obsolete 16-bit Blitter. Alice allows full-bus width access for bitplane DMA.

That's already 1992. I was talking about 1990 and my ECS version would have been WAY better than AGA for games (and not only) even staying 16 bit.
Quote:

Remember, the early NTSC Amiga 1000 didn't come with a 6-bitplane 64 color EHB mode, and the 1987-era Amiga Ranger's 128 color is a natural evolution. OCS has color display mode improvements. Commodore's A2410 TIGA did nothing for Amiga chipset evolution. I have Amiga 500 Rev 5 and Rev 6 motherboards, along with Amiga 1200 Rev 1D4.

Sure, and? The topic was about a (much) better ECS possible on 1990 with very minimal changes.

@Hammer

Quote:

Hammer wrote: @matthey Quote: matthey wrote: @cdimauro Thanks. It's another good read. I agree that ECS was not good enough for how late it arrived. The colors needed to be increased at least. Not only was it generally not a target for software enhancements but it was not enough of a hardware upgrade to sell ECS Amigas. AGA was far from perfect and late but at least it increased the colors and chip memory bandwidth. ECS needed to double the OCS chip memory bandwidth in order to add more colors though. You don't really mention how this would be done except for saying that the 68020 can access chip memory every 2 cycles instead of 4 cycles of the CPU with the 68000. Are you advocating for upgrading to a 68020 with ECS+ to give more chip memory accesses? Would this have effectively doubled chip memory bandwidth like it sounds? Are there articles or documentation somewhere that talks about this? I'm not sold on the conclusion giving partial blame for the disappointing Amiga chipset enhancements to the engineers. It may be true but there is a lack of evidence given for it. C= upper management was so bad that they created an environment which suppressed creativity and innovation. Suggesting using VRAM was likely to get the engineer in the dog house like the Amiga Corporation team and even Jay Miner. Being told "Read my lips - no new chips" is hostile and also suppressive of initiative. Ranger was cancelled, AGA was delayed, the DSP was cancelled. These were engineering initiatives that were suppressed by upper management. Commodore management allowed C65's 256 colors with a 4096 color palette VIC-III chipset R&D to be completed before canceling C65's R&D in December 1990. The project was canceled by Commodore's chairman Irving Gould in 1991. C65's 256 color mode is a planar (bit-planes) pixel format on Amiga 500's level Chip RAM bandwidth! The Commodore 65 prototype was later reverse-engineered and released as MEGA65 in 2020. Commodore's R&D resources were wasted.

As usual. This is clearly another example of dumb decisions.

And here I'm talking about the technical side as well: engineers made their own mistakes: C64 was chunky and they... switched to the MUCH WORSE planar graphics with the C65!

Only Commodore engineers make it possible...

Last edited by cdimauro on 27-Oct-2023 at 07:51 PM.

@cdimauro

Quote:

My argument is about what was really needed an achievable with the same constraints that engineers had for ECS.

Look at Commodore's IP not used for Amiga graphics. The building blocks exist within Commodore to develop competitive hardware graphics solutions.

Quote:

That's already 1992. I was talking about 1990 and my ECS version would have been WAY better than AGA for games (and not only) even staying 16 bit.

FYI, the Amiga 3000+ AGA design was completed around Q1 1991 and it was delayed since Commodore's "IBM PC Jr" management was focusing on 020+ with ECS development, hence losing more than 6 months.

Amiga 4000 is based on Amiga 3000+ AGA design.

Amiga 1200's Budgie is effectively a cost-reduced 32-bit Super Buster, 32-bit Fast RAM controller (for additional 32-bit Fast RAM like A3000/A4000's Ramsey), and CPU's access to 32-bit Chip RAM.

Commodore UK wanted a CPU-accelerated A1200 games bundle in mass production. CPU performance was a major factor in 1993.

Last edited by Hammer on 27-Oct-2023 at 06:20 AM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

When the original team left the next chipset was finished they didn’t use it.

@cdimauro

Quote:

PiStorm doesn't "emulate" the chipset. But a good part of the system, yes (RTG, AHI, 3D acceleration is possible).

In Emu68 the drivers are accessed natively, no emulation needed

_________________
Indigo 3D Lounge, my second home.
The Illusion of Choice | Am*ga