@cdimauro

Quote:

On FUTURE (well above 1993) machines, according to Eggebrecht. And the cost was $20-$30.

DSP3210's inclusion is too late for Mehdi Ali's Feb 1992 A1200 directive.

DSP3210 is being bundled with AA+ chipset improvements.

DSP3210 wouldn't be useful with Chip RAM-only configuration.

It would take A1200 redesign for a cost optimize 1 MB Fast RAM and 2 MB Chip RAM configuration.

A1200 is modified from A300/A600 with AA being placed in from A3000 Plus.

Due to PCMCIA's "memory only" mode, A600's Gayle needs to be modified for 68EC20, hence Ramsey and Bridgette were modified into Budgie. This is was done within two months.

A1200 was completed around end of March 1992 and the production line needs to be re-tooled for A1200. About six months was available until Mehdi Ali's Sep 1992 release directive. About 44,000 A1200s reached European markets for Xmas Q4 1992.

"Six+ months" lost is on management.

Unlike A1200, CD32 project had extra time.

With CD32's extra time, Commodore selected $50 C-Cube CL450 SoC with 512 KB ($13?) "Fast RAM" for the FMV module due to Irving Gould's focus on digital videos which is premature for MPEG1-based VCD format.

Unlike MPEG2-based DVD, there's not enough video quality difference between VHS and MPEG1-based VCD.

Irving Gould didn't give up on Philips CDi's FMV VCD concept until the end.

The Amiga has a very weak presence in East and South-East Asia. VCD was popular in the mentioned regions.

Quote:

Crap and alien stuff because of the incompetence of Commodore's engineers. .

1. Wrong. Its management should be offering proper leadership.

2. Without AGA software support, AGA is largely useless.

3. 3rd party IP blocks can be licensed.

Ex-Commodore engineers were hired by companies like 3DO, SGI, AMD and 'etc'.

Commodore engineer for C65's CPU was hired by AMD and joined K7 Athlon team. MOS/CSG 65xx CPU family has double rate processing design. K7 Athlon has DDR.

CSG was bought out by ex-Commodore management and GMT Microelectronics survived to the year 2000.

Commodore The Final Years book mentions SGI as a popular destination for ex-Commodore engineers. SGI has 3DFX (graphics business purchased by NVIDIA) and ArtX (purchased by ATI/AMD) breakaway companies.

Intel has assimilated major parts of 3DLabs. https://www.phoronix.com/news/MTIzMjk

Last edited by Hammer on 09-Aug-2024 at 06:41 AM.
Last edited by Hammer on 09-Aug-2024 at 06:34 AM.
Last edited by Hammer on 09-Aug-2024 at 06:32 AM.
Last edited by Hammer on 09-Aug-2024 at 06:28 AM.
Last edited by Hammer on 09-Aug-2024 at 06:26 AM.
Last edited by Hammer on 09-Aug-2024 at 06:22 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

@Hammer

Quote:

Hammer wrote: @cdimauro Quote: On FUTURE (well above 1993) machines, according to Eggebrecht. And the cost was $20-$30. DSP3210's inclusion is too late for Mehdi Ali's Feb 1992 A1200 directive.

It was just an experiment, and plans for using it were on future products.
Quote:

DSP3210 is being bundled with AA+ chipset improvements.

? AA+ was dropped well before the above Eggbrecht's interview.
Quote:

DSP3210 wouldn't be useful with Chip RAM-only configuration.

Of course: it's an alien component on an Amiga.

Whereas I've proved that even an OCS Amiga could get 14 audio channels... operating on Chip ram. Can you explain how is it possible?
Quote:

It would take A1200 redesign for a cost optimize 1 MB Fast RAM and 2 MB Chip RAM configuration.

Absolutely not: Fast Ram was NOT required for an A1200.

Instead, the A1200 required capable engineers which at least could haven't crippled the Chip RAM access from the CPU.
Quote:

A1200 is modified from A300/A600 with AA being placed in from A3000 Plus. Due to PCMCIA's "memory only" mode, A600's Gayle needs to be modified for 68EC20, hence Ramsey and Bridgette were modified into Budgie. This is was done within two months. A1200 was completed around end of March 1992 and the production line needs to be re-tooled for A1200. About six months was available until Mehdi Ali's Sep 1992 release directive. About 44,000 A1200s reached European markets for Xmas Q4 1992. "Six+ months" lost is on management.

Six, actually, according to what you've reported above.
Quote:

Unlike A1200, CD32 project had extra time.

Do you have dates for that? Because after the A1200 there was little time for the CD32 development.
Quote:

With CD32's extra time, Commodore selected $50 C-Cube CL450 SoC with 512 KB ($13?) "Fast RAM" for the FMV module due to Irving Gould's focus on digital videos which is premature for MPEG1-based VCD format. Unlike MPEG2-based DVD, there's not enough video quality difference between VHS and MPEG1-based VCD. Irving Gould didn't give up on Philips CDi's FMV VCD concept until the end. The Amiga has a very weak presence in East and South-East Asia. VCD was popular in the mentioned regions.

But this solution is expensive. And not needed for a console. In fact, the above module is an extra component, which was a good decision.
Quote:

Quote: Crap and alien stuff because of the incompetence of Commodore's engineers. . 1. Wrong. Its management should be offering proper leadership.

The same for the engineers: they lacked technical leadership. And not even the leadership.
Quote:

2. Without AGA software support, AGA is largely useless.

?
Quote:

3. 3rd party IP blocks can be licensed.

Why? Commodore has its own IPs that never shared.
Quote:

Ex-Commodore engineers were hired by companies like 3DO, SGI, AMD and 'etc'. Commodore engineer for C65's CPU was hired by AMD and joined K7 Athlon team. MOS/CSG 65xx CPU family has double rate processing design. K7 Athlon has DDR. CSG was bought out by ex-Commodore management and GMT Microelectronics survived to the year 2000. Commodore The Final Years book mentions SGI as a popular destination for ex-Commodore engineers. SGI has 3DFX (graphics business purchased by NVIDIA) and ArtX (purchased by ATI/AMD) breakaway companies. Intel has assimilated major parts of 3DLabs. https://www.phoronix.com/news/MTIzMjk

So, did bad only when worked with Commodore?

@cdimauro

It’s as if you are arguing against yourself… you see no point in implementing the hardware changes since there’s no software that would support it, all though it would be super easy to add support…

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@kolla

Quote:

kolla wrote: @cdimauro It’s as if you are arguing against yourself… you see no point in implementing the hardware changes since there’s no software that would support it, all though it would be super easy to add support…

? Those things are related, but NOT one the consequence of the other.

kolla, that's just elementary logic, eh!

What AGA needed was chipset a d chip ram design that did not cripple the CPU acces. However, that's not the AGA we got, so as a system, the A1200 should have had fast ram to compensate. No AGA system should ever have shipped without it, really.

We can argue all day about the pros and cons and in the end it doesn't make any difference. We got what we got. Still a great little machine.

_________________
Doing stupid things for fun...

Karlos Quote:

What AGA needed was chipset a d chip ram design that did not cripple the CPU acces. However, that's not the AGA we got, so as a system, the A1200 should have had fast ram to compensate. No AGA system should ever have shipped without it, really.

With one memory for the CPU and chipset, the easiest option when the Amiga was designed was to share (take turns). Another option became available shortly after the Amiga launched which was to have two ports to the memory. This was dual ported VRAM. Jay Miner foresaw VRAM as a natural fit for the Amiga which he added with the Ranger chipset but CBM foresaw too high of prices for the Amiga to turn into the next C64.

Jay Miner designed the Amiga with unified video and audio memory as the chipset memory unlike many systems back then where these memories were separate. The unified chip memory provided a larger chipset memory for video and audio and simplified programming but separate video and audio memory would have made it cheaper to have VRAM for the video memory and memory contention would have been lowered in some cases. I prefer the unified chipset memory, especially with the option to let the CPU have its own fast memory. CBM just needed to supply the fast memory. Even as an after thought, CBM could have supplied the Amiga 1200 with 1MiB fast memory trapdoor expansions like CBM UK did with the A501 512kiB expansion for the Amiga 500. Then again, memory was still expensive and the price could have easily been twice that of the A501 in the early 1990s which was still not cheap (£50-£100?). The case could be made for retaining fast memory as a pre-installed option to make the price of entry for the Amiga lower. CBM should have not just allowed but worked with the CBM UK and others to pre-install accelerators and memory upgrades though.

@Karlos

Quote:

Karlos wrote: What AGA needed was chipset a d chip ram design that did not cripple the CPU acces. However, that's not the AGA we got, so as a system, the A1200 should have had fast ram to compensate. No AGA system should ever have shipped without it, really. We can argue all day about the pros and cons and in the end it doesn't make any difference. We got what we got. Still a great little machine.

We got what we got even because the A1200 was an inexpensive machine, and this was because it has NOT embedded Fast RAM on it.

Memory was the most expensive thing on this machine. AFAIR around $50, whereas the entire chipset was $12.

Adding even 1MB of Fast RAM would have increased its price (and certainly not of $25: Commodore should have needed to amortise its cost) or reduced the profit margin (which the company desperately needed).

A solution would have been to bundle 1MB of Chip + 1MB of Fast. Then be prepared to see a drastic drop in quality for the games, since most of them benefit from the former and not from the latter.

In fact, and on the other hand, this is a 2D machine. And on such machines CPUs are basically slaves of the custom chips, because most of the work should be done by those and the CPU has almost only the duty to program them and let them execute those activities. You don't need Fast RAM for that.


@matthey unfortunately VRAM is expensive. And Ranger was not addressing what was really needed for Commodore's market.

@cdimauro

Quote:

@cdimauro It was just an experiment, and plans for using it were on future products.

Mehdi Ali's A1200 directive started from Feb 1992 which overridden Bill Sydnes' and Jeff Franks' product stack approach.

Lew wasn't in the hot seat in Feb 1992.


Quote:

@cdimauro ? AA+ was dropped well before the above Eggbrecht's interview.

Lew's "AA+" follows AA.

http://www.bambi-amiga.co.uk/amigahistory/leweggebrecht.html
Lew's interview is dated Jan 1993.

Lew's "AA+" includes chunky pixels mode, higher memory bandwidth for 72 hz, improved serial port, improved high-density floppy drive and DSP. When compared to AAA, Lew's AA+ are modest.


Quote:

@cdimauro Of course: it's an alien component on an Amiga. Whereas I've proved that even an OCS Amiga could get 14 audio channels... operating on Chip ram. Can you explain how is it possible?

Go implement the 14-channels Fat Paula in silicon.

For 1987 context, Commodore's LSI group is resistant against ECS Denise with 8 color registers with an existing 4096 (12-bit) palette and have to make do with an extra 4 color registers and 64 color palette. More than two years wasted on #metoo monochrome vs four-color high-res Denise.

From AAA project for AA, 24-bit (16.7 million color) palette is shared across 15 kHz and 31 kHz modes. This is for 1989-1991 date range.

3DO "Clio" has 16-bit custom DSP for audio with 16-bit stereo @ 44.1 kHz and 4 channel Dolby Surround.

Quote:

@cdimauro Absolutely not: Fast Ram was NOT required for an A1200. Instead, the A1200 required capable engineers which at least could haven't crippled the Chip RAM access from the CPU.

You're out of touch with the 1990s "3MB 32-bit game consoles" group with discrete memory pools.

The "3MB 32-bit game consoles" group's PS1, 3DO and Saturn have separated system memory and video memory pools.

Sega Mega Drive has 68000's 16-bit system memory and custom chip's 8-bit high-speed VRAM. Mega Drive is not gimped like A500's shared 16-bit Chip RAM.

Similar to 3DO's MADAM, Amiga Hombre Nathaniel has discrete 32-bit system and 32-bit/64-bit video memory pools. 25 Mhz MADAM includes matrix math co-processor and CEL engine's Data unpacker (DUP), Pixel decoder (PDC), Pixel processor (scale, merge, translucence) and Projector (pixel values written on quadrilateral area).

You're thinking like 1985. The CPU needs its memory bandwidth due to 3D-related game logic/game state processing.

3DO has the lessons from Super A500/A3000's large expensive VRAM i.e. MADAM's hybrid access between FP DRAM and VRAM.

https://www.youtube.com/watch?v=1zpuQOCF9oY
Arcade quality Super Street Fighter II Turbo (1994) on 3DO. This is "so-what" since other 3MB 32-bit consoles can do the same.

A1200's 140 ns read/write cycle FP DRAM has 7.1 Mhz effective which is 28.4 MB/s and strong enough CPU can grab 7 MB/s from it. 68EC020-14 doesn't deliver 28 MIPS let alone 7 MIPS.

CD32's FMV module's MIPS-X @ 40Mhz's arithmetic intensity can maximize the use with 16-bit 80 ns access FP DRAM.

Quote:

@cdimauro Six, actually, according to what you've reported above.

Actually, six+ i.e. "six plus".

https://landley.net/history/mirror/commodore/haynie.html
Quote:

Dave Haynie: When he got to Engineering, he hired a human bus error called Bill Sydnes to take over. Sydnes, a PC guy, didn't have the chops to run a computer, much less a computer design department. He was also an ex-IBMer, and spent much time trying to turn C= (a fairly slick, west-coast-style design operation), into the clunky mess that characterized the Dilbert Zones in most major east-coast-style companies. He and Ali also decided that AA wasn't going to work, so they cancelled both AA projects (Amiga 3000+ and Amiga 1000+, either one better for the market than the A4000 was), and put it all on the backburner, intentionally blowing the schedule by six+ months.

"AA task force" has it's talk fest wastage.

Quote:

@cdimauro The same for the engineers: they lacked technical leadership. And not even the leadership.

Similar ranking engineers without teamwork leads to nowhere, disjoined R&D, unfocused R&D and 'etc'. Needs proper leadership.

Quote:

@cdimauro Do you have dates for that? Because after the A1200 there was little time for the CD32 development.

Hint: Akiko is dependent on A1200's Budgie and AA-Gayle.

Quote:

From Commodore -The Final Years, July 1992, Jeff Porter made a first attempt to cost out a stripped-down CD Game System, which he called CD Stripper (pronounced “seedy stripper�). The new system would have no keyboard, no disk drive, and limited expansion. This first proposal also included a whopping 8 MB of RAM, a lot of memory at the time for a game system. The total bill of materials for the AGA system came to $232.67. (snip) On September 4, Jeff Porter began writing a formal product overview for the system, which he dubbed the Amigo. Once again, just like the CDTV, Porter promised the machine could expand into a full Amiga computer with keyboard and mouse. “We then went back and wrote a design specification in late September, and also established that we could achieve the price point,� says Lew Eggebrecht. (snip) In order to speed up development on the system, much of the technology in the Amiga would be inherited from George Robbins’ A1200. “The chief engineer on each project had a lot of discretion to build what made sense,� says Porter. “So for the CD32, we took an even cheaper CD mechanism without a motorized drawer (both Sony and Philips models) and wrapped an A1200 circuit around it. (snip) Engineering efforts towards a prototype motherboard began, and Jeff Porter delivered the final draft of his spec on November 2, 1992. He considered the Amigo to be a super-cost-reduced CDTV that resembled a fat Walkman. (snip) Hedley Davis and an engineer named Chris Coley would work on the Akiko chip design (called Arizona early in development)

Commodore didn't execute proper market intelligence in relation to the "3MB 32bit game console group" i.e. 3DO, PS1 and Saturn development.

CD32 project initially has 8 MB RAM design.

Last edited by Hammer on 11-Aug-2024 at 12:25 PM.
Last edited by Hammer on 11-Aug-2024 at 12:18 PM.
Last edited by Hammer on 11-Aug-2024 at 12:14 PM.
Last edited by Hammer on 11-Aug-2024 at 12:04 PM.
Last edited by Hammer on 11-Aug-2024 at 11:43 AM.
Last edited by Hammer on 11-Aug-2024 at 10:00 AM.
Last edited by Hammer on 11-Aug-2024 at 09:59 AM.
Last edited by Hammer on 11-Aug-2024 at 09:29 AM.
Last edited by Hammer on 11-Aug-2024 at 09:25 AM.
Last edited by Hammer on 11-Aug-2024 at 09:00 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

@cdimauro

Quote:

cdimauro wrote: @Karlos Quote: Karlos wrote: What AGA needed was chipset a d chip ram design that did not cripple the CPU acces. However, that's not the AGA we got, so as a system, the A1200 should have had fast ram to compensate. No AGA system should ever have shipped without it, really. We can argue all day about the pros and cons and in the end it doesn't make any difference. We got what we got. Still a great little machine. We got what we got even because the A1200 was an inexpensive machine, and this was because it has NOT embedded Fast RAM on it. Memory was the most expensive thing on this machine. AFAIR around $50, whereas the entire chipset was $12.

Not correct. It's higher i.e. "It cost us $14 for all three chips" for A500.

Quote:

From Commodore - The Final Years, "C65 was supposed to be super affordable to the low-end user. The bill of materials came to $39 US in models without disk drives." The 3.5 inch drive alone cost $26.78. With other parts, the bill of materials came to $120.82, including labor, duty and shipping. This was a lot compared to the C64, which cost Commodore around $50 to manufacture. At retail it would cost customers around $300, three times the proposed retail price of the C65. Bowen would lead the project and expected to be able to demonstrate a prototype by December 1987. Pre-production units would be completed in March 1988, and at least 5000 units could be built by June 1988. Now the race was on to see if Commodore would succeed in releasing a new and improved low-cost computer to carry on the C64 legacy.

C65 used CSG's 2 micron process node.

Quote:

cdimauro wrote: Adding even 1MB of Fast RAM would have increased its price (and certainly not of $25: Commodore should have needed to amortise its cost) or reduced the profit margin (which the company desperately needed). A solution would have been to bundle 1MB of Chip + 1MB of Fast. Then be prepared to see a drastic drop in quality for the games, since most of them benefit from the former and not from the latter. In fact, and on the other hand, this is a 2D machine. And on such machines CPUs are basically slaves of the custom chips, because most of the work should be done by those and the CPU has almost only the duty to program them and let them execute those activities. You don't need Fast RAM for that. @matthey unfortunately VRAM is expensive. And Ranger was not addressing what was really needed for Commodore's market.

8MB RAM adds $20 on the BOM cost.

Quote:

From Commodore - The Final Years, On June 27, 1992 (snip), Jeff Porter made a first attempt to cost out a stripped-down CD Game System, which he called CD Stripper (pronounced “seedy stripper�). The new system would have no keyboard, no disk drive, and limited expansion. This first proposal also included a whopping 8 MB of RAM, a lot of memory at the time for a game system. The total bill of materials for the AGA system came to $232.67. (snip) Porter had previously proposed a machine with 8 MB of memory, which added over $20 to the bill of materials When the developers found out it added over $50 to the retail price, they settled for 2 MB. (skip) Engineering efforts towards a prototype motherboard began, and Jeff Porter delivered the final draft of his spec on November 2, 1992. He considered the Amigo to be a super-cost-reduced CDTV that resembled a fat Walkman. The device would have connectors for two joysticks, a television, and a proprietary expansion connector similar to that of the Amiga 500. Each unit would cost Commodore $226 to manufacture.

For 1993-1995, the right balance is between 2 MB and 8 MB i.e. 3 MB to 4 MB.

Psygnosis disagreed with CD32's low ball specs. During CD32's development, Psygnosis already giving hints about a certain "far east" game console to Commodore.


Quote:

From Commodore the Inside Story - The Untold Tale of a Computer Giant by David John Pleasance, As a result of this, I was asked by Ian Hetherington (cofounder, with Jonathan Ellis, of Psygnosis) to arrange a meeting with Mehdi Ali at their studios in Liverpool. (skip) Ian explained to Mehdi that with a few seemingly quite modest design changes, the CD32 could have an incredible boost in performance at very marginal additional cost. He also pointed out the benefits it would give developers like Psygnosis and other major players in the industry, who would find it easier to produce even better-quality products and enhance the reputation of the CD32 and the games publishers – a genuine ‘win-win’. (skip) Well, it went exactly as expected. Mehdi was rude and ignorant, and clearly had no idea what Ian was talking about. But instead of just admitting that, he more or less turned on Ian, as though he ‘must be crazy telling us how to design our computers!’ I ushered Mehdi out of the building feeling very ashamed, and it was quite a while before I plucked up the courage to talk to Ian again. Luckily for me, Ian had realised what kind of a person Mehdi Ali was and held no bad feelings towards me.

For Sony, it's a good thing Commodore's leadership didn't have "only the paranoid survive" mindset.

PS; my RPi 3A+, 4B and CM4 are manufactured by Sony UK.

Last edited by Hammer on 11-Aug-2024 at 11:46 AM.
Last edited by Hammer on 11-Aug-2024 at 10:40 AM.
Last edited by Hammer on 11-Aug-2024 at 10:38 AM.
Last edited by Hammer on 11-Aug-2024 at 10:37 AM.
Last edited by Hammer on 11-Aug-2024 at 10:25 AM.
Last edited by Hammer on 11-Aug-2024 at 10:22 AM.
Last edited by Hammer on 11-Aug-2024 at 10:20 AM.
Last edited by Hammer on 11-Aug-2024 at 10:15 AM.
Last edited by Hammer on 11-Aug-2024 at 10:09 AM.
Last edited by Hammer on 11-Aug-2024 at 10:07 AM.
Last edited by Hammer on 11-Aug-2024 at 10:06 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

cdimauro Quote:

@matthey unfortunately VRAM is expensive. And Ranger was not addressing what was really needed for Commodore's market.

VRAM was too expensive in the mid to late 1980s for the low end but CBM needed a higher margin higher end "workstation" to solidify its reputation as a graphics leader and there weren't as many options to increase chip bandwidth early. The 1990 ECS Amiga 3000 was a nice machine other than the disappointing graphics and old chipset bottlenecks. The Amiga 3000UX offered an expensive graphics card that was more expensive than integrated graphics. Starting around 1990, dual ported VRAM could have been an option for the high end and a few years later even for the low end. The competition often had faster graphics and more graphics bandwidth by using a smaller amount of VRAM while the Amiga unified chipset memory made this more expensive. AAA supported both DRAM and VRAM options. Dual ported VRAM was a better fit for the Amiga than the memory alignment restricted AGA method to increase chip bandwidth that was less compatible. Options to increase chip bandwidth were expensive whether the memory was wider, faster or dual ported. The chosen solution for AGA was a cheaper option that still required moderately faster and more expensive memory with reduced compatibility. AGA failed to fully take advantage of the chip bandwidth increase, defaulting to the older, slower and more compatible bandwidth. Some of this may have been due to the rushed AGA design but the increased bandwidth was not assumed and fully utilized.

The time period is important. Even with a standard amount of Amiga memory, what is the correct split of chip and fast memory?

1985-1986 512kiB
1987-1988 1MiB
1989-1990 2MiB
1991-1992 3MiB
1993-1994 4MiB

All chip memory with more chip bandwidth is desired. Even AGA failed to deliver enough bandwidth increase to eliminate the need for fast memory though. AGA seemed pretty slow compared to VRAM GPUs moving into the mid-1990s. Chunky would have helped but 16 bit chunky would have pushed the need for more chipmem leaving less budget for fast memory. Then 32/24 bit chunky would have done it again but there was getting to be more memory options by that time. The tradeoff of having less high performance memory, that can act like a high level cache, or more cheap slow memory is still around today while most desktops choose the latter with insane amounts of caches. Some systems integrate higher performance eDRAM on chip and many MCUs use SRAM for main memory where wider, faster and more ports are possible.

Last edited by matthey on 14-Aug-2024 at 08:12 PM.

@cdimauro

Quote:

A solution would have been to bundle 1MB of Chip + 1MB of Fast. Then be prepared to see a drastic drop in quality for the games, since most of them benefit from the former and not from the latter.

As long as there was an inexpensive option to upgrade the Chip Ram it wouldn't have been much of an issue.

@Hammer

Quote:

Hammer wrote: @cdimauro Quote: cdimauro wrote: @Karlos We got what we got even because the A1200 was an inexpensive machine, and this was because it has NOT embedded Fast RAM on it. Memory was the most expensive thing on this machine. AFAIR around $50, whereas the entire chipset was $12. Not correct. It's higher i.e. "It cost us $14 for all three chips" for A500.

I recall that you reported $12 for the Amiga 1200 chipset.

Regarding the $14 for A500, was it on 1987 (e.g.: its introduction)?
Quote:

Quote: From Commodore - The Final Years, "C65 was supposed to be super affordable to the low-end user. The bill of materials came to $39 US in models without disk drives." The 3.5 inch drive alone cost $26.78. With other parts, the bill of materials came to $120.82, including labor, duty and shipping. This was a lot compared to the C64, which cost Commodore around $50 to manufacture. At retail it would cost customers around $300, three times the proposed retail price of the C65. Bowen would lead the project and expected to be able to demonstrate a prototype by December 1987. Pre-production units would be completed in March 1988, and at least 5000 units could be built by June 1988. Now the race was on to see if Commodore would succeed in releasing a new and improved low-cost computer to carry on the C64 legacy. C65 used CSG's 2 micron process node.

That's a very important information, because it means that CGS had this process around 1987.

However, it's very strange that it took another 3 years (from 1988) to continue the development of the C65 (until it was cancelled).
Quote:

Quote: cdimauro wrote: Adding even 1MB of Fast RAM would have increased its price (and certainly not of $25: Commodore should have needed to amortise its cost) or reduced the profit margin (which the company desperately needed). A solution would have been to bundle 1MB of Chip + 1MB of Fast. Then be prepared to see a drastic drop in quality for the games, since most of them benefit from the former and not from the latter. In fact, and on the other hand, this is a 2D machine. And on such machines CPUs are basically slaves of the custom chips, because most of the work should be done by those and the CPU has almost only the duty to program them and let them execute those activities. You don't need Fast RAM for that. @matthey unfortunately VRAM is expensive. And Ranger was not addressing what was really needed for Commodore's market. 8MB RAM adds $20 on the BOM cost. Quote: From Commodore - The Final Years, On June 27, 1992 (snip), Jeff Porter made a first attempt to cost out a stripped-down CD Game System, which he called CD Stripper (pronounced “seedy stripper�). The new system would have no keyboard, no disk drive, and limited expansion. This first proposal also included a whopping 8 MB of RAM, a lot of memory at the time for a game system. The total bill of materials for the AGA system came to $232.67. (snip) Porter had previously proposed a machine with 8 MB of memory, which added over $20 to the bill of materials When the developers found out it added over $50 to the retail price, they settled for 2 MB. (skip) Engineering efforts towards a prototype motherboard began, and Jeff Porter delivered the final draft of his spec on November 2, 1992. He considered the Amigo to be a super-cost-reduced CDTV that resembled a fat Walkman. The device would have connectors for two joysticks, a television, and a proprietary expansion connector similar to that of the Amiga 500. Each unit would cost Commodore $226 to manufacture.

OK, then $50 was the added value to the prices for consumers and not the BOM. I've mixed the two things.

$20 for 8MB was quite cheap.
Quote:

For 1993-1995, the right balance is between 2 MB and 8 MB i.e. 3 MB to 4 MB.

Doom required 4MB, which was a big leap frog from the past (even Wolfestein3D required 640kB, AFAIR).
Quote:

Psygnosis disagreed with CD32's low ball specs. During CD32's development, Psygnosis already giving hints about a certain "far east" game console to Commodore. Quote: From Commodore the Inside Story - The Untold Tale of a Computer Giant by David John Pleasance, As a result of this, I was asked by Ian Hetherington (cofounder, with Jonathan Ellis, of Psygnosis) to arrange a meeting with Mehdi Ali at their studios in Liverpool. (skip) Ian explained to Mehdi that with a few seemingly quite modest design changes, the CD32 could have an incredible boost in performance at very marginal additional cost. He also pointed out the benefits it would give developers like Psygnosis and other major players in the industry, who would find it easier to produce even better-quality products and enhance the reputation of the CD32 and the games publishers – a genuine ‘win-win’. (skip) Well, it went exactly as expected. Mehdi was rude and ignorant, and clearly had no idea what Ian was talking about. But instead of just admitting that, he more or less turned on Ian, as though he ‘must be crazy telling us how to design our computers!’ I ushered Mehdi out of the building feeling very ashamed, and it was quite a while before I plucked up the courage to talk to Ian again. Luckily for me, Ian had realised what kind of a person Mehdi Ali was and held no bad feelings towards me. For Sony, it's a good thing Commodore's leadership didn't have "only the paranoid survive" mindset.

Sure, however Psygnosis requests about the CD32 aren't know.

2MB of Chip RAM were enough for this console, thanks to the fast CD (300kB/s is more than 10 times faster than the usual floppy), which allowed to load data on-demand (e.g.: streaming of data).

And the rich sound track can stay on the CD, freeing space and time for it.

For 2D games it was very nice.

However, if Psygnosis wanted to target 3D, then it's different, and it was right (but, again, depends on the requests, which we don't know).

@matthey

Quote:

matthey wrote: cdimauro Quote: @matthey unfortunately VRAM is expensive. And Ranger was not addressing what was really needed for Commodore's market. VRAM was too expensive in the mid to late 1980s for the low end but CBM needed a higher margin higher end "workstation" to solidify its reputation as a graphics leader and there weren't as many options to increase chip bandwidth early. The 1990 ECS Amiga 3000 was a nice machine other than the disappointing graphics and old chipset bottlenecks. The Amiga 3000UX offered an expensive graphics card that was more expensive than integrated graphics.

Two chipset were required to address both markets, but Commodore wasn't able to understand it.
Quote:

Starting around 1990, dual ported VRAM could have been an option for the high end and a few years later even for the low end. The competition often had faster graphics and more graphics bandwidth by using a smaller amount of VRAM while the Amiga unified chipset memory made this more expensive. AAA supported both DRAM and VRAM options. Dual ported VRAM was a better fit for the Amiga than the memory alignment restricted AGA method to increase chip bandwidth that was less compatible. Options to increase chip bandwidth were expensive whether the memory was wider, faster or dual ported.

VRAM could have been a good addition, but only for the high-end systems.
Quote:

The chosen solution for AGA was a cheaper option that still required moderately faster and more expensive memory with reduced compatibility. AGA failed to fully take advantage of the chip bandwidth increase, defaulting to the older, slower and more compatible bandwidth. Some of this may have been due to the rushed AGA design but the increased bandwidth was not assumed and fully utilized.

AGA brought a very good bandwidth increase, but its main problem was that it was done only on the display subsystem and not on all areas, as I've proved on my last articles.

A LOT could have been done with very minimal changes...
Quote:

The time period is important. Even with a standard amount of Amiga memory, what is the correct split of chip and fast memory? 1985-1986 512kiB 1988-1989 1MiB 1990-1991 2MiB 1992-1993 3MiB

IMO no Fast RAM was required because 2D was the primary target.

But I can tell you my idea about how much memory (Chip-only) was needed:
1985-1986: 512kB
1987-1989: 512kB + 512kB (NOT Slow RAM!!!) via trapdoor
1990-1992: 2MB
1993-: 4MB
Quote:

All chip memory with more chip bandwidth is desired. Even AGA failed to deliver enough bandwidth increase to eliminate the need for fast memory though. AGA seemed pretty slow compared to VRAM GPUs moving into the mid-1990s.

Double bandwidth (e.g.: 14Mhz clock) on 1990 / ECS would have helped a lot to bridge the gap.

AGA with 4 x bandwidth on low-end systems was very good.
Quote:

Chunky would have helped but 16 bit chunky would have pushed the need for more chipmem leaving less budget for fast memory.

8-bit packed/chunky was important even before AGA and while sticking to 2D. Planar graphics was really the worst choice for the chipset, since the beginning, because it wasted too much bandwidth (on cookied-cut / masking) and complicated a lot drawing the BOBs.

16-bit bit chunky for 2D games on 1992 / AGA could have been a very nice upgrade, IMO, and super-compelling.
Quote:

Then 32/24 bit chunky would have done it again but there was getting to be more memory options by that time.

That's more important on GUI / Workbench / applications.
Quote:

The tradeoff of having less high performance memory, that can act like a high level cache, or more cheap slow memory is still around today while most desktops choose the latter with insane amounts of caches. Some systems integrate higher performance eDRAM on chip and many MCUs use SRAM for main memory where wider, faster and more ports are possible.

That's completely different from the usual chipset work / usage. Maybe on more advanced 3D-systems (like the XBox360's GPU has shown. But that's very late in time).


@Rob

Quote:

Rob wrote: @cdimauro Quote: A solution would have been to bundle 1MB of Chip + 1MB of Fast. Then be prepared to see a drastic drop in quality for the games, since most of them benefit from the former and not from the latter. As long as there was an inexpensive option to upgrade the Chip Ram it wouldn't have been much of an issue.

Yes, but see above: I still prefer having 2MB of Chip RAM on board and add Fast RAM in the trapdoor.

The Amiga's primary target usage is 2D, and for this only the Chip RAM was required / needed.

Fast RAM is more for 3D, and applications, which was less important.


@all: it would be good to know how Doom ran on an A1200 + 2MB Fast Mem, and nothing else, while displaying a 320x200 256 colours screen.

I mean, the time demo WITHOUT the C2P routine called for each frame, even if there's garbage displayed on the screen.

@cdimauro

Quote:

@all: it would be good to know how Doom ran on an A1200 + 2MB Fast Mem, and nothing else, while displaying a 320x200 256 colours screen. I mean, the time demo WITHOUT the C2P routine called for each frame, even if there's garbage displayed on the screen.

Not exactly A1200 + 2MB setup but still you get some relative values on Bruce Abbott's post on EAB.

Blizzard 1230IV (50MHz 030) and 60ns RAM

"c2p_optimized" (normal AGA c2p routine) - 10.1 fps
"fake chunky" (copy FastRAM to ChipRAM) - 10.85 fps, 7% faster.
"fake RTG" (copy FastRAM to FastRAM) - 12.65 fps, 25% faster
"Fake RTG" direct (just rendering to FastRAM) - 13.27 fps, 31% faster

How much CPU does C2P consume?

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

@cdimauro

Quote:

@all: it would be good to know how Doom ran on an A1200 + 2MB Fast Mem, and nothing else, while displaying a 320x200 256 colours screen.

Is it even possible to run original Doom on a PC with so little RAM?

I have a vague memory of having to upgrade the old family 486SX25 to 8MB before we could actually run original Doom.

For what it's worth...

I just tried ADoom on bare A1200 with 4MB RAM and 3.0 kickstart, it won't start.

"Error: Unable to allocate at least 2097152 fastmem for zone management while leaving 2Mb fastmem free" (SIC)

I used all tricks to free fast ram and tried different versions, but no success. Other ports of Doom (BOOM, Chocolate etc) have much higher requirements.

Last edited by kolla on 14-Aug-2024 at 01:59 PM.
Last edited by kolla on 14-Aug-2024 at 01:57 PM.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@pixie

Quote:

pixie wrote: @cdimauro Quote: @all: it would be good to know how Doom ran on an A1200 + 2MB Fast Mem, and nothing else, while displaying a 320x200 256 colours screen. I mean, the time demo WITHOUT the C2P routine called for each frame, even if there's garbage displayed on the screen. Not exactly A1200 + 2MB setup but still you get some relative values on Bruce Abbott's post on EAB. Blizzard 1230IV (50MHz 030) and 60ns RAM "c2p_optimized" (normal AGA c2p routine) - 10.1 fps "fake chunky" (copy FastRAM to ChipRAM) - 10.85 fps, 7% faster. "fake RTG" (copy FastRAM to FastRAM) - 12.65 fps, 25% faster "Fake RTG" direct (just rendering to FastRAM) - 13.27 fps, 31% faster

He used a 030: not useful for what I need.

I was thinking about some tests which could show the potential of the Amiga platform (AGA. OCS/ECS is irrelevant on 1992+).

This starting with a base A1200 + 2MB di Fast Mem (because Doom used 4MB.However, kolla's experiment shows that more than 4MB of Fast Mem is needed) and with rendering on Fast RAM and then copying the content to Chip RAM (for being displayed, even if it's pure garbage since it's packed/chunky).

Another interesting config is A1200 with 4MB Chip RAM, so that it directly renders the graphics on Chip RAM.
It doesn't exist in the reality, but WinUAE can emulate it. Unfortunately, the emulation cannot be cycle-exact in this case, but it can be good enough to have a rough idea.

Other configurations (A1200 + 020Mhz + 2MB Fast RAM, ...) can be useful to see how the platform scales with more powerful processors.

I'm not interested on other tests (C2P or RTG) other than those.
Quote:

How much CPU does C2P consume?

Not relevant for what I need.


@kolla

Quote:

kolla wrote: @cdimauro Quote: @all: it would be good to know how Doom ran on an A1200 + 2MB Fast Mem, and nothing else, while displaying a 320x200 256 colours screen. Is it even possible to run original Doom on a PC with so little RAM? I have a vague memory of having to upgrade the old family 486SX25 to 8MB before we could actually run original Doom.

DOOM required 4MB of RAM.
Quote:

For what it's worth... I just tried ADoom on bare A1200 with 4MB RAM and 3.0 kickstart, it won't start. "Error: Unable to allocate at least 2097152 fastmem for zone management while leaving 2Mb fastmem free" (SIC) I used all tricks to free fast ram and tried different versions, but no success. Other ports of Doom (BOOM, Chocolate etc) have much higher requirements.

Thanks. Albeit very very strange, because an A1200 with 4MB of Fast RAM you have a total of 6MB, which is well beyond what Doom needed on PCs.

Maybe ADoom needs to move some assets to Chip RAM, to free more space in Fast RAM. However, this requires some changes to the code, and I don't think that the one(s) which did the port could consider to implement it.

The lack of chunky support in the Amiga chipset was not just a performance impediment but wasted valuable memory again that CBM needed to minimize to keep costs low. Chipset and 68k ISA improvements could have provided a significant competitive cost advantage for an insignificant increase in chipset cost.

Doom was a tight squeeze on 4MiB PCs. Id recommended disabling everything using memory including memory bank handlers/extensions. The DOOM.EXE size increased over time making the fit tighter with later versions.

version | size in bytes
1.0 523,221
1.1 579,187
1.2 575,767
1.25 560,547
1.4 635,361
1.5 656,329
1.6 658,469
1.666 687,001
1.8 709,865
1.9 715,493

https://doomwiki.org/wiki/DOOM.EXE

The DOOM1.WAD likely provides most of the data and is ~4MiB (4,196,020 for last versions).

https://doomwiki.org/wiki/DOOM1.WAD

Either the DOOM1.WAD only partially loaded into memory or it is compressed to fit into memory. The DOOM.EXE likely loads into memory and remains in memory. The 68k has about 20% better code density than x86 but x86 compilers are often better so this is not always realized. A 20% code savings could save 104,644 bytes for v1.0 to 143,098 bytes for v1.9. Motorola, or CBM if they had licensed the 68k, could have added ColdFire instructions and other optimizations that may have reduced the code size another 5%.

A 320x200x8 bitmap requires 64,000 bytes and this would be doubled if using double buffering. The Amiga may be able to use a chunky bitmap and planar bitmap using the same bitmap memory as double buffering or may require a 3rd bitmap. The Amiga would use a few kiB more memory for the c2p code. The 68k Amiga should still have a memory savings advantage over the PC version of Doom. An Amiga with 2MiB chip mem and 2MiB fast mem instead of unified memory would reduce the max mem chunk to 2MIB which could keep this spec of Amiga from being able to use Doom. A 4MiB chip mem 68k Amiga would have a better chance even with the c2p overhead while a 68020+ Amiga with 4MiB chip mem and 8 bit chunky/CLUT mode would likely have an easier time than the PC. AA+ was planned to only have 16 bit chunky which would have doubled the bitmap mem cost to 128,000 bytes for double buffering but this may be payed for with the better code density of the 68k. AA+ was planning to increase the chip memory max to 8MiB so a 4MiB AA+ chipset 68k Amiga could plausibly run Doom with 4MiB of chip mem even in 16-bit chunky mode. Some optimizations and adaption for the 68k Amiga hardware would certainly have helped as well.

The RPi Pico/RP2040 version of Doom gives an idea of what is possible to run Dooom as Thumb ISAs have similar code density to the 68k Amiga. The RP2040 has 264 kiB SRAM and the RPi Pico has 2MiB flash yet the Pico version of Doom is not cut down and plays at "30-35+ FPS". Impressive result that cheap hardware makes possible. A 68060+AA+ uses fewer transistors than the $1 USD RP2040 ASIC SoC but the Amiga peasants must eat expensive and fattening PPC cake which is why the Amiga is dead.

https://kilograham.github.io/rp2040-doom/



Is the Amiga curse perpetually horrible management? Do Amiga users have to wear a bag over their head forever?

Last edited by matthey on 15-Aug-2024 at 12:25 AM.
Last edited by matthey on 14-Aug-2024 at 11:58 PM.
Last edited by matthey on 14-Aug-2024 at 11:56 PM.

@matthey


Quote:

Either the DOOM1.WAD only partially loaded into memory or it is compressed to fit into memory. The DOOM.EXE likely loads into memory and remains in memory. The 68k has about 20% better code density than x86 but x86 compilers are often better so this is not always realized. A 20% code savings could save 104,644 bytes for v1.0 to 143,098 bytes for v1.9. Motorola, or CBM if they had licensed the 68k, could have added ColdFire instructions and other optimizations that may have reduced the code size another 5%.

1. Commodore only has a few engineers that handle 65xx CPUs and PA-RISC's complexity let alone 68K. TG68's ALTERA Cyclone II did not exist in the late 1980s.

2. For 68030-25's cost, AMD offered Am386-40.

3. Code density is less useful when 68020/68030's instruction cycle latency is inferior to MIPS R2000 and R3000 series. 68020/68030 lost to RISC competitors in the UNIX workstation market. 3D requires good math-intensity processing.

4. For CD32 FMV module's custom MIPS-X (C-Cube CL450 SoC) beats 68020/68030 on integer math intensity workload.

For Sega's Saturn project, 68030 lost to Hitach's SuperH2 in September 1993. Saturn has SuperH1 as a CD-ROM controller.

5. The original ColdFire core was launched in 1994. Commodore's AA R&D started around 1989. You're a few years late. You have a timeline problem.

ColdFire multiply-accumulate (MAC) is optional instead of being standard.

68EC020 was available in early 1991.

Before the 68EC020 offer in early 1991, C= AA was partnered with 68000 @ 14 Mhz. AA Lisa is the only 32-bit chip with 16-bit 68000-14 Mhz CPU, 16-bit AA Alice, and 16-bit Paula.

For PlayStation-X, Sony selected the LSI's MIPS-based IP for the CPU and GTE. PSX was running its 1st texture-mapped 3D Ridge Racer game in Dec 1993. Most of PS1's heavy engineering was done by LSI for CPU/GTE and Toshiba for GPU.

SGI's 1990 MIPS R3000A selection for Indigo. Indigo workstation was released in July 1991. Sony's R3000A selection followed SGI's example.


Timeline for 68040 delays,
http://kpolsson.com/micropro/proc1990.htm
September 1990, Motorola announces it has begun general sampling of the 68040 processor.
November 1990, Motorola begins volume shipments of the 25 MHz 68040 processor.
October 1991, the 1st Mac Quadra 700 with 68040 was released.


Timeline for 80486,
1988, 80486 started production,
June 1989, 80486 launched for i486DX-20, i486DX-25. The first i486-based PCs were announced in late 1989.
May 1990, i486DX-33 launched,
April 1991, i486SX-16 launched,
June 1991, i486DX-50 launched,
Sep 1991, i486SX-20 and i486SX-25 launched.


68040's delays caused other 68K Unix workstation vendors to look at RISC alternatives e.g. MIPS, DEC Alpha, and HP PA-RISC (Weitek co-design its FPU).

Last edited by Hammer on 15-Aug-2024 at 02:55 AM.
Last edited by Hammer on 15-Aug-2024 at 02:53 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

@cdimauro

Quote:

@cdimauro, I recall that you reported $12 for the Amiga 1200 chipset. Regarding the $14 for A500, was it on 1987 (e.g.: its introduction)?

From "Commodore - The Final Years" book,
Quote:

From former A500 engineer, Bob Welland, CSG was able to produce Agnus, Denise, and Paula for $5.49, $5.19, and $7.91 respectively.

Total: $18.59 for media chips. This doesn't include CSG's cost for two CIA and Gary support chips.

Bob Welland offered two cost values for A500 i.e. $18.59 (1987 context) and $14 (with context against Bill Sydnes' administration).

Quote:

@cdimauro That's a very important information, because it means that CGS had this process around 1987. However, it's very strange that it took another 3 years (from 1988) to continue the development of the C65 (until it was cancelled).

While the 65CE02 (1 IPC with 8-bits) CPU was completed around 1988, engineering resources were lacking with C65's media custom chips.

With the PC world during the entire 1980s, Intel wasn't responsible for SVGA clones and MS-DOS/Windows. The PC world has the advantage of R&D cost being partitioned across multiple companies following clear PC standards.

For the C65 platform, Commodore is responsible for the entire platform's R&D.

Commodore's Amiga hardware R&D is populated with Amiga-related add-on R&D instead of focusing on Amiga's core graphics R&D. Time wasting with monochrome vs color high-res Denise and post OCS/ECS debates.

Quote:

From Commodore - The FInal Years, Henri Rubin, the instigator of the new project, wanted to go after the Macintosh by adding noninterlaced monochrome video for the Amiga. (skip) Rubin wanted his engineers to build a flat paperwhite monochrome monitor, similar to the Macintosh, for the Amiga 500. (skip) Rubin had little computer experience and no way to tell if the scheme was possible. “So he looks over at Bob and says, ‘Bob, is that possible? Does this even make sense?’ And Bob said, ‘It’s an insanely good idea.’ That's when I got the go-ahead.�

#metoo R&D road map from Henri Rubin.

Henri Rubin had little computer experience leading a computer tech company. LOL

Commodore committed R&D resources on the custom MMU for 68020 before Motorola released its 68851 MMU. Early A2620 has 68020 with Commodore's custom MMU.

Commodore committed R&D resources on Amiga SX-500, Osborne style portability for A500.

Commodore senior management dislikes GVP for a reason.

Sony's PSX R&D effort was spread across LSI (MIPS IP for CPU and GTE) and Toshiba (for GPU).

Quote:

@cdimauro OK, then $50 was the added value to the prices for consumers and not the BOM. I've mixed the two things. $20 for 8MB was quite cheap.

The maximum profit is extracted from CD32 when Commodore management listened to lowball 3rd party UK game studios. Commodore management wasn't paranoid enough.

If CD32 had 2 MB Chip RAM and 6 MB Fast RAM as originally intended, what can you do with it?


Quote:

@cdimauro 2MB of Chip RAM were enough for this console, thanks to the fast CD (300kB/s is more than 10 times faster than the usual floppy), which allowed to load data on-demand (e.g.: streaming of data). And the rich sound track can stay on the CD, freeing space and time for it. For 2D games it was very nice. However, if Psygnosis wanted to target 3D, then it's different, and it was right (but, again, depends on the requests, which we don't know).

I recall from one of David Pleasance's youtube videos, I recall BullFrog also disagreed with CD32's configuration. Bull Frog had Magic Carpet R&D.

My cited Street Fighter 2 AGA tech demo examples wouldn't be portable with OCS/ECS Amigas. Reshoot Proxima 3 is only for AGA Amigas.

Fightin Spirit runs on both OCS/ECS and AGA Amigas.

There's an effort on Final Fight port for OCS/ECS Amigas with Metro Siege's style color with Copper shading optimizations which removes Atari ST's influence. https://www.timeextension.com/news/2024/06/super-final-fight-promises-a-more-authentic-arcade-experience-for-amiga-fans

From https://eab.abime.net/showthread.php?t=118275 there is an effort to port Mortal Kombat 2 for AGA Amigas.

These are some of the "What If" itch that annoyed the Amiga community.

Last edited by Hammer on 15-Aug-2024 at 05:15 AM.
Last edited by Hammer on 15-Aug-2024 at 05:11 AM.
Last edited by Hammer on 15-Aug-2024 at 05:02 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

@Hammer

Quote:

Hammer wrote: @cdimauro Quote: @cdimauro, I recall that you reported $12 for the Amiga 1200 chipset. Regarding the $14 for A500, was it on 1987 (e.g.: its introduction)? From "Commodore - The Final Years" book, Quote: From former A500 engineer, Bob Welland, CSG was able to produce Agnus, Denise, and Paula for $5.49, $5.19, and $7.91 respectively. Total: $18.59 for media chips. This doesn't include CSG's cost for two CIA and Gary support chips. Bob Welland offered two cost values for A500 i.e. $18.59 (1987 context) and $14 (with context against Bill Sydnes' administration).

OK, so those are the numbers for first and last production period.

Do you know how much costed the three chips for the A1200?
Quote:

Quote: @cdimauro OK, then $50 was the added value to the prices for consumers and not the BOM. I've mixed the two things. $20 for 8MB was quite cheap. The maximum profit is extracted from CD32 when Commodore management listened to lowball 3rd party UK game studios. Commodore management wasn't paranoid enough. If CD32 had 2 MB Chip RAM and 6 MB Fast RAM as originally intended, what can you do with it?

I would have complained, because such amount of memory, especially the presence of Fast RAM, is non-sense, as I've already explained.

Unless you want to also address the 3D. Then either you needed an AGA + 8-bit packed/chunky (impossible, due to the incompetent engineering team) or 4MB of Fast RAM + an Akiko with C2P using the DMA (and also this is impossible, due to the same, incompetent team).
Quote:

Quote: @cdimauro 2MB of Chip RAM were enough for this console, thanks to the fast CD (300kB/s is more than 10 times faster than the usual floppy), which allowed to load data on-demand (e.g.: streaming of data). And the rich sound track can stay on the CD, freeing space and time for it. For 2D games it was very nice. However, if Psygnosis wanted to target 3D, then it's different, and it was right (but, again, depends on the requests, which we don't know). I recall from one of David Pleasance's youtube videos, I recall BullFrog also disagreed with CD32's configuration. Bull Frog had Magic Carpet R&D.

Again, if it's about the 3D, then see above.

But for 2D, the CD32 is very good as it is.