@cdimauro

Quote:

cdimauro wrote: @agami Quote: agami wrote: Yes. That's the concept @Karlos has proposed a bunch of times, and I have to say that it makes more sense than most Amigans are prepared to accept. I don't see the problem: Amigans already use (only) 68k. By definition (Amiga is a precise and well defined platform).

There are those that are still infatuated with PowerPC or the idea that an Amiga has to be different at it's core(s). Even Trevor voiced how he doesn't want the Amiga to be just a "PC".
Then you also have the ARM-port and x86-port camps, ignoring the fact that there there are no teams left to port AmigaOS to either, and MorphOS has been silent on the subject for too many years.

The concept put forward is that everything moves to 68k for the foreseeable future, and high performance needs are then met with JIT on x64 or ARM64, with some OS components progressively written for the native hardware for additional speed/features.

Write for 68k, execute (almost) anywhere.
AGA and machine specifics for older software can be emulated in software (UAE) or alternatively via FPGA.

That's a kind of "different" that is just right for me.

_________________
All the way, with 68k

@pixie

Quote:

pixie wrote: @cdimauro Quote: The video is too long. I've tried to go forward to reach that specific part, but I haven't found it. So, I can't evaluate the idea. It's around 02h01m20s

Thanks!

The changes that they have made to FS-UAE are just to make it easy to develop its demo.

So, I assume (because there was no specific mention about it) and the memory protection was used to intercept bad/illegal accesses to memory areas.

Which should be something similar to what is already possible with tools like Enforce.


@agami

Quote:

agami wrote: @cdimauro Quote: cdimauro wrote: @agami [quote]agami wrote: Yes. That's the concept @Karlos has proposed a bunch of times, and I have to say that it makes more sense than most Amigans are prepared to accept. I don't see the problem: Amigans already use (only) 68k. By definition (Amiga is a precise and well defined platform).

There are those that are still infatuated with PowerPC or the idea that an Amiga has to be different at it's core(s).[/quote]
Because they sleep with their eyes open: reality is completely different.
Quote:

Even Trevor voiced how he doesn't want the Amiga to be just a "PC".

He doesn't even have the AmigaOne brand for its computers...
Quote:

Then you also have the ARM-port and x86-port camps, ignoring the fact that there there are no teams left to port AmigaOS to either, and MorphOS has been silent on the subject for too many years. The concept put forward is that everything moves to 68k for the foreseeable future, and high performance needs are then met with JIT on x64 or ARM64, with some OS components progressively written for the native hardware for additional speed/features. Write for 68k, execute (almost) anywhere. AGA and machine specifics for older software can be emulated in software (UAE) or alternatively via FPGA. That's a kind of "different" that is just right for me.

Exactly. "Just" accept it...

@cdimauro

Oh come on - it was accepted a couple of decades ago already!

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@agami

Quote:

agami wrote: There are those that are still infatuated with PowerPC or the idea that an Amiga has to be different at it's core(s). Even Trevor voiced how he doesn't want the Amiga to be just a "PC".

We all grew up with an 'Amiga' that was a real physical machine, something with own very specific hardware, a nice OS and a blooming software market providing the best possible games.

Most of us would like to see a new Amiga with its own hardware and OS and some are even dreaming about the return of AAA titles to such an Amiga.

Own hardware will cost a lot of money especially in small numbers and I don't talk about such rubbish like CPU architectures, just make own x86-64 or ARM boards with integrated high class GPUs and you will end up in a price range of an A1-5000.

The Amiga OS is not really something, that could compete with modern free systems like Linux, just the well integrated GFX and GUI could help to compete but the default File-Managers, WWW-Browsers or drivers are no match for other systems.

And when it comes to games, we would be glad if we could have a port of something that runs better on Linux or Windows.

So in the end, UAE will be the best way. If you are a hardcore fan you can run a native system like AROS or MorphOS on other hardware but when it comes to old or Amiga compatible software look for an integrated UAE.

@OneTimer1

Quote:

OneTimer1 wrote: @agami We all grew up with an 'Amiga' that was a real physical machine, something with own very specific hardware, a nice OS and a blooming software market providing the best possible games. Most of us would like to see a new Amiga with its own hardware and OS and some are even dreaming about the return of AAA titles to such an Amiga.

Yes, and we forget that the unique part was the custom chipset giving it multimedia capabilities and the pre-emptive multitasking OS.
In the second half of the '80s the M68000 was a mainstream CPU also in use by Apple for the LISA and Macintosh, and by Atari for the ST line. If anything, during the mid '80s, the x86 CPU architecture was the outlier.

To put it in today's terms, what made Amiga different was a decent sound card and graphics card as standard, and the OS. That is what created a positively differentiated user experience.

Quote:

Own hardware will cost a lot of money especially in small numbers and I don't talk about such rubbish like CPU architectures, just make own x86-64 or ARM boards with integrated high class GPUs and you will end up in a price range of an A1-5000.

This is exactly my point.
There are those that don't want the Amiga NG computing experience to just be an OS running on commodity hardware. They believe it needs to be exotic, so they want to use an exotic CPU, even if ti means the hardware ends up costing 2x-3x more than a really good personal computer for Linux, Windows or macOS, while only being able to accomplish a 5th of the other platforms.

Despite Apple fancying up their x86 line of computers, during that phase macOS was mostly just an OS running on commodity hardware, which the whole Hackintosh movement proved to be true. But it still ended up growing the overall market for the platform.

Most of the desktop and laptop Linux computing experience is just an OS running on commodity hardware. And that platform has seen more growth in the past 5 years than in the 20 years before that.

I too would like a nice commercial offering of computing hardware which is sold and branded with some Amiga NG computing experience, but I am very happy for that to be just commodity hardware. x64 or ARM64, I don't mind either way.

Personal computing is about the user experience, so the focus should be the user experience for developers and general users alike. A good user experience for developers means they write/port apps and games to the platform, which in turn makes a good user experience for the general user. As Linux and Apple have shown, you no longer need exotic hardware to create a positively differentiated user experience.

_________________
All the way, with 68k

@bhabbott

Quote:

bhabbott wrote: @cdimauro Quote: cdimauro wrote: No, the management wanted to have CP/M compatibility as a possibility (via cartridge). So, it was an optional add-on. Whereas the engineers inserted the required hardware in the machine. YES, that's what I said! Quote: The "funny" thing is that they didn't talk each other... I sense your goal here is to bash Commodore for not micro-managing hardware development from the top. Like many companies they gave their engineers a degree of flexibility in meeting design goals. Bil Herd's innovative CP/M solution was one of those examples. You may say it proves that Commodore's management was dysfunctional, but this is how a lot of companies produced great stuff. For example:- Ken Kutaragi (designer of the Sony PlayStation) Quote: In 1989, he was watching his daughter play a Famicom and realized the potential that existed within video games. At that particular time, Sony's executives had no interest in video games. Thus, when Nintendo expressed the need for a sound chip for its upcoming new 16-bit system, Kutaragi accepted the offer. Working in secret, he designed the chip, the SPC700. When they found out, Sony's executives were furious. Only with Sony CEO Norio Ohga's approval was Kutaragi able to complete the chip and keep his job. Even while working with Nintendo, within Sony, gaming was still regarded as a fad. Despite this hostile atmosphere to video games, Kutaragi managed to persuade Ohga into working with Nintendo to develop a CD-ROM add-on for the Super NES which would be released alongside a Sony branded console which could play both Super NES cartridges and CD games. These efforts resulted in a device called the "Play Station"... Despite being considered a risky gamble by other Sony executives, Kutaragi once again had the support of Ohga and several years later the company released the original PlayStation. The commercial success of the PlayStation franchise makes Sony Computer Entertainment the most profitable business division of Sony. The PlayStation was the result of one engineer whose vision clashed with management. He wasn't given any specs to work with, and had to create a prototype of his own design just to convince them to go ahead with the project. The rest, as they, is history, but without Ken Kutaragi's tenacity and willingness to work 'in secret' against management's wishes the PlayStation wouldn't have happened - a loss for Sony and for the world.

For PS1, Ken Kutaragi's design worked with about 1 million transistors budget.

https://www.eetimes.com/playstation-creator-courts-serious-fun/

Out of that frustration an opportunity was born that ultimately led Kutaragi to his current spot as “the father” of the Playstation. Kutaragi and a Sony salesperson met with executives from Nintendo to propose Sony apply its signal-processing prowess to Nintendo's next-generation console. The two companies quickly struck a deal.

“We designed a small chip and made an offer to Nintendo, and they picked it up in their 16-bit system, the Super NES, which offered PCM [pulse code modulation] audio,” said Kutaragi. The work gave birth to a small team of about five designers, including Masakazu Suzuoki, who later became the core of the Playstation team.

“We realized that this was a nice growth area for us in digital entertainment, and driven by the evolution in semiconductors "Moore's Law"— there would be a new level in entertainment,” he said. Indeed, emboldened by his success, Kutaragi made another proposal to Nintendo in 1989: the two should work toward developing the first CD-ROM-based console. A year later, Nintendo agreed and the two were off to the races.

'Beat Nintendo'
“But at Summer CES [Consumer Electronics Show] in 1991, I had a surprise when Nintendo announced a realignment, and said they would work with Philips on the console and stop our project,” Kutaragi recalled. “Our engineers had a good relationship [with theirs], but management decided to go another way.”

Ironically, the Nintendo/Philips console never got off the ground, but the deal had an unintended effect. “We decided to start our own [console] development, and we gathered up a team at Sony to create a new gaming system to beat Nintendo,” said Kutaragi.

Sony's ambitious goal in May 1992 was to create the first CD-ROM console with real-time computer graphics powered by a 1-million-transistor system-on-a-chip. Kutaragi talked to every semiconductor company who would accept a meeting to find a partner for his plans. Some weren't interested, others said it couldn't be done. Ultimately, Halla seized the opportunity for LSI Logic and helped deliver the MIPS-based chip.

“Almost every night for two-and-a-half years we had conference calls on the project, with Kutaragi in attendance at most of them, mainly to go over engineering trade-offs,” Halla said. “The rest of the meetings were highly animated philosophical discussions about pricing.”


----
https://en.wikipedia.org/wiki/PlayStation_(console)

Namco's research managing director Shegeichi Nakamura met with Kutaragi in 1993 to discuss the preliminary PlayStation specifications, with Namco subsequently basing the Namco System 11 arcade board on PlayStation hardware and developing Tekken to compete with Virtua Fighter.[64] The System 11 launched in arcades several months before the PlayStation's release, with the arcade release of Tekken in September 1994

Namco's Ridge Racer was one of the most popular arcade games at the time, and it was already confirmed behind closed doors that it would be the PlayStation's first game by December 1993.

1. For specification discussions, Sony's CEO allowed Kutaragi to work with game developers such as Namco.

2. Many conference calls about engineering trade-offs and pricing.

After the failed meeting between Psygnosis and Commodore International, Psygnosis was later acquired by Sony in 1993 to work on the PlayStation console.

According to David Pleasance, Commodore International's CEO Mehdi Ali insulted the 3rd party developers to fukoff in front of their faces.

I extremely dislike Mehdi Ali.



From https://www.afr.com/companies/commodore-falls-on-troubled-times-19900126-kalsw
Gould has 20% of Commodore's shares and Gould calls the shots at this troubled company and has for the past 23 years.


-----
In modern times, LSI Logic is part of s Broadcom Inc. and has good partnerships with Raspberry Pi Foundation and Sony Pencoed UK.

Last edited by Hammer on 21-Nov-2023 at 12:26 PM.
Last edited by Hammer on 21-Nov-2023 at 12:22 PM.
Last edited by Hammer on 21-Nov-2023 at 12:17 PM.
Last edited by Hammer on 21-Nov-2023 at 12:15 PM.
Last edited by Hammer on 21-Nov-2023 at 12:13 PM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@agami

Quote:

agami wrote: @cdimauro Quote: cdimauro wrote: @Kronos However emulation and VMs are quite different. Mostly when it comes to the CPU. In other areas they may equally emulate a range of hardware and in this regard they are much the same.

With a documented workaround, I can run NVIDIA GPU drivers on Hyper V's client Windows with the GPU resources being shared between multiple VMs and host OS.

QEmu supports PCIe passthrough and my Ryzen 9 7900X/7950X has unused RDNA 2 IGPs while the host OS has NVIDIA RTX 4080/4090 GPUs. I don't recall A-EON's System 54 supporting RDNA 2 GPUs.

Last edited by Hammer on 21-Nov-2023 at 12:37 PM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Hammer

Quote:

Hammer wrote: “But at Summer CES [Consumer Electronics Show] in 1991, I had a surprise when Nintendo announced a realignment, and said they would work with Philips on the console and stop our project,” Kutaragi recalled. “Our engineers had a good relationship [with theirs], but management decided to go another way.” Ironically, the Nintendo/Philips console never got off the ground, but the deal had an unintended effect. “We decided to start our own [console] development, and we gathered up a team at Sony to create a new gaming system to beat Nintendo,” said Kutaragi.

So Kutaragi only developed the PlayStation because Nintendo dropped their plans to produce a CD-ROM based console. But this was not a Sony management decision, it was just Kutaragi - with support from the CEO who had backed him on the previous occasion when he developed a chip in secret. Sony management thought games were just a fad not worthy of Sony's involvement, which is hardly surprising given the history of games consoles and Nintendo's 'surprise' backing out of its CD-ROM deal.

The important point is, the PlayStation was not a well thought-out corporate plan to produce a world-beating console, it was a result of Kutaragi's skill and tenacity combined with a fair bit of luck. Had things gone a bit differently Sony might not have gotten involved in the video game industry at all.

Quote:

1. For specification discussions, Sony's CEO allowed Kutaragi to work with game developers such as Namco.

Kutaragi was lucky to get the CEO's ear.

My understanding is that Japanese companies are very hierarchical with every employee knowing their place and not questioning those above them. In technology companies I have worked in New Zealand it was very difficult to convince the boss to go with my ideas, even when I was specifically hired to provide technical advice. I can't imagine that it was easier in Japan.

Quote:

I extremely dislike Mehdi Ali.

I don't know enough about him to have such an extreme view. I have just ordered the book "Commodore: The Final Years" by Brian Bagnall, hoping to get some insight into what Mehdi Ali did to deserve such dislike.

But this is all a bit off-topic for this thread. The OP says:-

"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."

And I tend to agree. Much of the narrative regarding development of the Amiga chipset is rather one-sided, having come from engineers or fans with unrealistic expectations. Apart from suspect use of the word 'deception' in the title, cdimauro doesn't fall into that trap with his ideas about what could have been provided in ECS.

I think without intimate knowledge of the chipset and appreciation of the engineering, production and marketing parameters it is difficult to say what could realistically could have been done better. I am happy just getting a better understanding of what was achieved, which was still pretty awesome even if not every Amiga fan's wet dream. I am also interested in what we might be able to do to the chipset today while sticking to the spirit of those times.

In the next post I will explore this in detail...

In his blog cdimauro suggests that the changes made to ECS were "paltry improvements" that "did not justify the creation of ad hoc games". This is certainly true. However the question is should it have been more and if so why? Just assuming that a chipset with better gaming capabilities must be worth it is not enough. The biggest problem the Amiga faced was maintaining a large enough user base to justify continued game development, and the biggest threat to that was the PC - which had 10 times the market share and more affluent owners who could afford better technology. Any discussion of what the Amiga chipset should have had must consider these factors paramount if it is to be taken seriously.

In 1990, when ECS was released, the A500 was at the peak of popularity with sales growing even as PC sales were (temporarily) declining. That was largely down to its low price and higher quality games compared to other home computers (and most PCs). IMO it was better to capitalize on this rather than introduce new features that would raise the price and fragment the market. Other home computer makers got a poor response to doing that because games still targeted the base model due to the large installed base. We couldn't ignore that effect. Any enhancements would need to complement the already excellent chipset without significantly raising the price or creating incompatibilities.

However by 1990 we could see that VGA and Sound Blaster (introduced in July of that year) were making the PC a force to be reckoned with, and with such a huge number of users there was far more incentive to port games from the Amiga to the PC than vice versa. Therefore we really needed something to make it easier to get those ports across (as well as assuaging Amiga fans' PC envy). IMO two things needed to be done urgently if the Amiga was not to be perceived as slipping behind:-

1. 256 colors in lo-res, preferably with with at least an 18 bit palette to match VGA. This was needed to make porting VGA graphics as easy as possible with negligible quality loss. PC owners were swooning over the new VGA games that looked so much better than EGA - and we needed to be there too.

To do 256 colors with acceptable speed bitplane DMA bandwidth had to be increased. There were two obvious ways to do this - a 32 bit ChipRAM bus, or double CAS. 32 bit would be expensive and not much use unless a 32 bit CPU was also used. Double CAS could be done cheaply, but would still require significant changes to the chipset. The enhanced palette could be provided with an 'industry standard" RAMDAC like those used in VGA cards (and [i]was[/i used for AGA). Alternatively it could stick with 4096 colors and just increase the number of palette registers inside Denise from 32 to 256.

The result would be graphics half way between ECS and AGA. It could provide 16 colors in hires without DMA contention to improve 'productivity' apps. Sprites could also be doubled in width or quantity. Forget about double-scan resolutions though - an external flicker fixer could be used for that with better compatibility. Without the complications of double scan and 32 bits the chipset would be much easier to design than AGA, and the OS wouldn't need as much work either.

2. Synth sound. The music produced by Paula's 4 PCM channels was much more organic and realistic than the Adlib's cheap synth chip. But PC fans quickly grew to love that toy-synth sound, which was what they were expecting since arcade machines and gaming consoles also used it.

This could very easily be added to the Amiga simply by putting an OPL chip on the motherboard. An 'industry standard' external mixer chip could be used to combine the synth sound with Paula. If all 4 channels of Paula were brought out separately it could be used to pan them too.

One good thing about ECS was that it could be retrofitted to existing Amigas. It would be awesome if the same thing could be done with a 256 color chipset, perhaps combined with a daughterboard to add synth sound (which might plug into the Paula socket). This way existing Amiga owners would be able to upgrade without having to buy a new machine, making the transition easier and assuaging their fear of being left behind.

Of course this wouldn't be the end of it. AGA could still be introduced in 1992, but now with a solid backing of 256 color games. AGA might also introduce even better performance with its 32 bit architecture, such as a faster blitter, chunky pixels and higher CPU bandwidth. However again it might be best to stick with 15 kHz scan rates. High-end Amigas would now be using RTG, and the AGA chipset could be provided for them on a 32 bit plugin card. Games could then be produced to work on RTG or AGA, like they often are now.

Last edited by bhabbott on 21-Nov-2023 at 09:42 PM.
Last edited by bhabbott on 21-Nov-2023 at 09:41 PM.
Last edited by bhabbott on 21-Nov-2023 at 09:37 PM.
Last edited by bhabbott on 21-Nov-2023 at 09:36 PM.
Last edited by bhabbott on 21-Nov-2023 at 09:35 PM.
Last edited by bhabbott on 21-Nov-2023 at 09:33 PM.

@bhabbott: you bite off more than you can chew. Remember that the management forced the engineers to make little changes to the chipset (that's why we got the ECS instead of the AGA): "no new chips". For this reason I bet that some of your proposals wouldn't be accepted.

However and if you read both my articles (this is the first one: the next one addressed the performance improvement) you can see that with my proposal (which were feasible at the time) the ECS could have got:
- 8 bitplanes / 256 colours (without the expensive 256 CLUT);
- 8 bitplanes / 16 + 15 colours dual playfields (exactly like AGA);
- panning for the 4 audio channels (e.g. much more immersive sound. WITHOUT requiring two channels for it);
- better CPU access (as you know, the chipset crippled the CPU accesses for 68020+);
- maybe BOBs' mask caching (this would have improved the performance, allowing to do/move more graphics).
That's for the first article.

On the second article the topic / suggested improvements was just doubling the chipset frequency (without double CAS: just using faster memories would have been enough), which could have allowed doubling the available memory accesses slots and then get this:
- low-res 256 colours or 4 + 4 DP mode using the same bandwidth of the OCS low-res with 4 bitplanes;
- high-res 256 colours or 4 + 4 DP mode (using the full bandwidth. Like the OCS high-res with 4 bitplanes) or even HAM mode (using only 6 bitplanes);
- faster Copper (two times);
- faster Blitter (from two up to three times, based on the available free slots).

Everything 100% compatible with OCS, of course.

With all those changes my ECS would have been even much better than AGA and a serious competitor with PCs and consoles (for 2D games, of course. But 3D games would have gained a great improvement compared to the Amiga 68000/OCS).

@bhabbott

Quote:

So Kutaragi only developed the PlayStation because Nintendo dropped their plans to produce a CD-ROM based console. But this was not a Sony management decision, it was just Kutaragi - with support from the CEO who had backed him on the previous occasion when he developed a chip in secret. Sony management thought games were just a fad not worthy of Sony's involvement, which is hardly surprising given the history of games consoles and Nintendo's 'surprise' backing out of its CD-ROM deal. The important point is, the PlayStation was not a well thought-out corporate plan to produce a world-beating console, it was a result of Kutaragi's skill and tenacity combined with a fair bit of luck. Had things gone a bit differently Sony might not have gotten involved in the video game industry at all.

The company's CEO is the ultimate manager.

Kutaragi reminded Ohga of the humiliation he suffered from Nintendo, Ohga retained the project and became one of Kutaragi's most staunch supporters. Pride is a major factor.

Quote:

"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."

1989 released ECS Agnus has existed with A500 Rev 6A.

With A3000's ECS Agnus and higher-density RAM chips, the A500 Rev 6A motherboard has the necessary jumper settings for 2MB Chip RAM configuration. A500 Rev 6A's 2MB Chip RAM design is a reserve capability.

The "Read my lips – no new chips" context is for A3000's R&D phase.

Except for the 1 MB to 2 MB Chip RAM difference, all A3000's ECS Agnus updates are in 1989 era A500 Rev6's ECS Agnus hence, hence statement is correct.

C65's VIC-III (256-color display with 4096 color palette on A500 chipram bandwidth) was completed around December 1990.

Commodore could have 32-bit 25 Mhz Chip RAM with 256-color (8 bit) display with 4096 color palette chipset for the A3000.

Last edited by Hammer on 22-Nov-2023 at 02:52 PM.
Last edited by Hammer on 22-Nov-2023 at 02:51 PM.
Last edited by Hammer on 22-Nov-2023 at 02:49 PM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Hammer

Quote:

Hammer wrote: Quote: "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." 1989 released ECS Agnus has existed with A500 Rev 6A. With A3000's ECS Agnus and higher-density RAM chips, the A500 Rev 6A motherboard has the necessary jumper settings for 2MB Chip RAM configuration. A500 Rev 6A's 2MB Chip RAM design is a reserve capability. The "Read my lips – no new chips" context is for A3000's R&D phase.

Which is exactly the context of what we're discussing here.
Quote:

Except for the 1 MB to 2 MB Chip RAM difference, all A3000's ECS Agnus updates are in 1989 era A500 Rev6's ECS Agnus hence, hence statement is correct.

See above: it wasn't about the (existing) Amiga 500 (or 2000) platform.

The context was about using the Amiga 3000 development for the new ECS chipset for this machine and the subsequent ones (500+, 600). So, for the new generation of machines with REALLY Enhanced capabilities (as per my two articles).
Quote:

C65's VIC-III (256-color display with 4096 color palette on A500 chipram bandwidth) was completed around December 1990.

Here the "no new chips" didn't apply for this crap...
Quote:

Commodore could have 32-bit 25 Mhz Chip RAM with 256-color (8 bit) display with 4096 color palette chipset for the A3000.

16-bit 14Mhz Chip RAM + chipset should have been more than enough for completely revamping the Amiga platform, as I've suggested.

@cdimauro

Quote:

16-bit 14Mhz Chip RAM + chipset should have been more than enough for completely revamping the Amiga platform, as I've suggested.

My argument is based on proven Commodore's Intellectual property results. Ramsey's 25 Mhz 32-bit memory controller intellectual property design blocks serve as the proposed Agnus/Alice replacement and the Denise/Lisa raster chip is based on the memory controller's and DMA's performance.

Ramsey is responsible for addressing RAM and producing DMA addressing.

All A3000 motherboards are 25 Mhz capable including 16 Mhz version i.e. jumper set to 25 Mhz over clock.

J100 has three possible settings:
1-2 = 25 Mhz
2-3 = 16 Mhz
3-4 = EXT

I have the 25 MHz model.

It's too bad Commodore didn't design a new graphics hardware IP on Ramsey's 25 Mhz 32-bit memory controller/DMA addressing foundation.

The 16 million color palette is not a big deal when compared to sustained color display capability e.g. fast 640x480p with 256 colors or 4096 (12 bit) colors HAM with +25 Mhz clock speed.

The legacy timings for the raster display's 3.5 Mhz/ 7.1 Mhz/ 14 Mhz requirements can be slotted into 25 Mhz or 28 Mhz.

16 million color palette consumes a certain amount of transistors and Commodore's chip fabrication capability may not be ready for it.

---------
For 1994, both Sony's PS1 and Commodore's Amiga Hombre aimed for a 1 million transistors budget.

Sony's PS1 CPU has a 33 Mhz clock speed, hence the 33 MIPS CPU and 66 MIPS co-processor (Geometry Transformation Engine) and fixed function 32-bit 3D GPU operate at 53 Mhz.

Last edited by Hammer on 23-Nov-2023 at 07:54 AM.
Last edited by Hammer on 23-Nov-2023 at 07:47 AM.
Last edited by Hammer on 23-Nov-2023 at 07:45 AM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Hammer

Quote:

Hammer wrote: It's too bad Commodore didn't design a new graphics hardware IP on Ramsey's 25 Mhz 32-bit memory controller/DMA addressing foundation. The 16 million color palette is not a big deal when compared to sustained color display capability e.g. fast 640x480p with 256 colors or 4096 (12 bit) colors HAM with +25 Mhz clock speed.

The A3000 was an expensive mistake. But both Commodore and many Amiga fans didn't understand that the high end wasn't worth chasing.

Fast 640x480p with 256 colors was useless without a fast CPU and lots of RAM to go with it. But that would guarantee a high price and very low sales. What games would support it? Almost none. PCs weren't even using that resolution because it was too taxing. This hardware could not* be put in a low-end machine that would sell in large numbers. Without large numbers the new graphics mode would not be used in games. Result - fail!

* before you start telling me how it could be done, remember that 'low-end' meant displaying on a TV - which can't do 640x480 except in flickering interlace and doesn't have the resolution for it either. A complete non-starter!

@cdimauro

Quote:

cdimauro wrote: @bhabbott: you bite off more than you can chew. Remember that the management forced the engineers to make little changes to the chipset (that's why we got the ECS instead of the AGA): "no new chips". For this reason I bet that some of your proposals wouldn't be accepted.

Maybe not, but they did accept some changes - which suggests they might have accepted others.

If they weren't going to do what I suggested then it would be best to just concentrate on making cheaper machines to flood the market until they were ready for new chips. No A3000, no CDTV, just heaps of A500s and A2000s at prices nobody could turn down. Then deliver AGA in early 1992.

Quote:

However and if you read both my articles (this is the first one: the next one addressed the performance improvement) you can see that with my proposal (which were feasible at the time) the ECS could have got: - 8 bitplanes / 256 colours (without the expensive 256 CLUT);

I don't like this idea. EHB wasn't that popular, and extending it to 256 colors would suck almost as bad as the Archimedes did. Would not be compatible with VGA so PC developers would hate it.

Quote:

- 8 bitplanes / 16 + 15 colours dual playfields (exactly like AGA);

Nice. Wouldn't need a 256 color CLUT for dual playfields. However I still think it would need it for 256 colors. AGA used an external RAMDAC and so should this.

Quote:

- panning for the 4 audio channels (e.g. much more immersive sound. WITHOUT requiring two channels for it);

A very minor improvement that many users wouldn't notice, But hard to do inside Paula. Better to bring out each channel separately and do it externally. That way you could do surround sound too! 48 pin DIP doesn't have any spare pins for this, but a 52 pin PLCC should.

Quote:

- better CPU access (as you know, the chipset crippled the CPU accesses for 68020+);

Only useful if you have an 020+, not low-end model. And why do you need fast CPU access anyway? I would rather have a faster blitter. However I suspect both of these ideas would be difficult to implement without 'new' chips.

Quote:

- maybe BOBs' mask caching (this would have improved the performance, allowing to do/move more graphics).

Another nice idea that doesn't sound easy to implement.

Quote:

On the second article the topic / suggested improvements was just doubling the chipset frequency (without double CAS: just using faster memories would have been enough), which could have allowed doubling the available memory accesses slots and then get this: - low-res 256 colours or 4 + 4 DP mode using the same bandwidth of the OCS low-res with 4 bitplanes; - high-res 256 colours or 4 + 4 DP mode (using the full bandwidth. Like the OCS high-res with 4 bitplanes) or even HAM mode (using only 6 bitplanes); - faster Copper (two times); - faster Blitter (from two up to three times, based on the available free slots).

Sounds good. Not sure how easy it would be to do though.


Quote:

Everything 100% compatible with OCS, of course.

Of course, this is essential. Puts even more burden on the designers though.

Quote:

With all those changes my ECS would have been even much better than AGA and a serious competitor with PCs and consoles (for 2D games, of course. But 3D games would have gained a great improvement compared to the Amiga 68000/OCS).

3D games needed a very fast CPU. Once you have that it changes the game - now fast ChipRAM access is important, but you also want chunky. Actually you need chunky or John Carmark et al won't be interested.

@Hammer

Quote:

Hammer wrote: @cdimauro Quote: 16-bit 14Mhz Chip RAM + chipset should have been more than enough for completely revamping the Amiga platform, as I've suggested. My argument is based on proven Commodore's Intellectual property results. Ramsey's 25 Mhz 32-bit memory controller intellectual property design blocks serve as the proposed Agnus/Alice replacement and the Denise/Lisa raster chip is based on the memory controller's and DMA's performance. Ramsey is responsible for addressing RAM and producing DMA addressing.

The point is that Ramsey is not useful for the chipset. Even because... it's NOT needed: Agnus has already everything inside.

BTW, the chipset was THE topic on this series of articles.
Quote:

All A3000 motherboards are 25 Mhz capable including 16 Mhz version i.e. jumper set to 25 Mhz over clock. J100 has three possible settings: 1-2 = 25 Mhz 2-3 = 16 Mhz 3-4 = EXT I have the 25 MHz model. It's too bad Commodore didn't design a new graphics hardware IP on Ramsey's 25 Mhz 32-bit memory controller/DMA addressing foundation.

Because it's not needed: see above.
Quote:

The 16 million color palette is not a big deal when compared to sustained color display capability e.g. fast 640x480p with 256 colors or 4096 (12 bit) colors HAM with +25 Mhz clock speed.

That made sense only for high-end systems: not for the low-end / mainstream ones.
Quote:

The legacy timings for the raster display's 3.5 Mhz/ 7.1 Mhz/ 14 Mhz requirements can be slotted into 25 Mhz or 28 Mhz.

25Mhz is not enough. You need 28Mhz to be perfectly in sync with the raster.
Quote:

16 million color palette consumes a certain amount of transistors and Commodore's chip fabrication capability may not be ready for it.

In fact, my proposal was about keeping the existing chipset with its 4096 colours.


@bhabbott

Quote:

bhabbott wrote: @Hammer Quote: Hammer wrote: It's too bad Commodore didn't design a new graphics hardware IP on Ramsey's 25 Mhz 32-bit memory controller/DMA addressing foundation. The 16 million color palette is not a big deal when compared to sustained color display capability e.g. fast 640x480p with 256 colors or 4096 (12 bit) colors HAM with +25 Mhz clock speed. The A3000 was an expensive mistake. But both Commodore and many Amiga fans didn't understand that the high end wasn't worth chasing. Fast 640x480p with 256 colors was useless without a fast CPU and lots of RAM to go with it. But that would guarantee a high price and very low sales. What games would support it? Almost none. PCs weren't even using that resolution because it was too taxing. This hardware could not* be put in a low-end machine that would sell in large numbers. Without large numbers the new graphics mode would not be used in games. Result - fail! * before you start telling me how it could be done, remember that 'low-end' meant displaying on a TV - which can't do 640x480 except in flickering interlace and doesn't have the resolution for it either. A complete non-starter!

I totally agree.
Quote:

bhabbott wrote: @cdimauro Quote: cdimauro wrote: @bhabbott: you bite off more than you can chew. Remember that the management forced the engineers to make little changes to the chipset (that's why we got the ECS instead of the AGA): "no new chips". For this reason I bet that some of your proposals wouldn't be accepted. Maybe not, but they did accept some changes - which suggests they might have accepted others. If they weren't going to do what I suggested then it would be best to just concentrate on making cheaper machines to flood the market until they were ready for new chips. No A3000, no CDTV, just heaps of A500s and A2000s at prices nobody could turn down. Then deliver AGA in early 1992.

That was the point, in fact.

However the A3000 was designed: then the engineers should have taken the chance to propose a more advanced ECS chipset (instead of the crap that they delivered), to be used on the low-end machines and make the Amiga again very competitive with the other platforms (consoles included).
Quote:

Quote: However and if you read both my articles (this is the first one: the next one addressed the performance improvement) you can see that with my proposal (which were feasible at the time) the ECS could have got: - 8 bitplanes / 256 colours (without the expensive 256 CLUT); I don't like this idea. EHB wasn't that popular,

Not popular because it demanded a lot to the system, but the few games that used were nice.
Quote:

and extending it to 256 colors would suck almost as bad as the Archimedes did.

No, because Archimedes was worse: it had a CLUT with only 16 entries, whereas the Amiga had 32. So, you required only 8 tones instead of 16.
Quote:

Would not be compatible with VGA so PC developers would hate it.

It doesn't matter: the important thing is to have 256 colours. You don't get all shades, but the palette was limited to 4096 colours anyway.
Quote:

Quote: - 8 bitplanes / 16 + 15 colours dual playfields (exactly like AGA); Nice. Wouldn't need a 256 color CLUT for dual playfields. However I still think it would need it for 256 colors.

Not needed. But maybe you can tinker yourself and convert some VGA / 256 colours images to an 8-bit EHB mode @ 4096 colours and check the results.
Quote:

AGA used an external RAMDAC and so should this.

It's expensive. Remember: "no new chips". You've to think about your proposals which should be subject to minimal changes AKA minimal transistors added.
Quote:

Quote: - panning for the 4 audio channels (e.g. much more immersive sound. WITHOUT requiring two channels for it); A very minor improvement that many users wouldn't notice, But hard to do inside Paula.

Why? It's the exact opposite: trivial. It's enough to duplicate the low byte of the volume channel to the high byte, and connect the new output to the opposite stereo channel. Done!
Quote:

Better to bring out each channel separately and do it externally. That way you could do surround sound too! 48 pin DIP doesn't have any spare pins for this, but a 52 pin PLCC should.

Again: "no new chips".

And with my proposal you already have a much better surround sound.
Quote:

Quote: - better CPU access (as you know, the chipset crippled the CPU accesses for 68020+); Only useful if you have an 020+, not low-end model.

Exactly, but we also had high-end models.
Quote:

And why do you need fast CPU access anyway?

For high-end systems.
Quote:

I would rather have a faster blitter.

Indeed and already addressed.
Quote:

However I suspect both of these ideas would be difficult to implement without 'new' chips.

I don't think that the bus arbitration logic requires so many changes only for this. Anyway, it wasn't the most important thing to implement.
Quote:

Quote: - maybe BOBs' mask caching (this would have improved the performance, allowing to do/move more graphics). Another nice idea that doesn't sound easy to implement.

No, it's super-easy. It requires two bits: one which enables the caching and another one which selects if the cache should be read or written.

With the first blit you enable it and select to write the fetched masks to the mask cache. On subsequent blits you read the mask from the cache.

When a blit starts, the internal index for the cache is reset. When a mask is written or read then the index is just incremented.

Fairly simple, isn't it? It's a stupid buffer implementation (just sequential writes or reads).

The only issue is represented by the transistor budget: a cache for 64x64 images costs 512 bits = 256 x 16-bit words, which is a lot (the same number of entries used for the chipset bank register). Maybe a 32x32 cache could be more feasible (1/4 of the size = only 128 bits needed = 64 16-bit words).
Quote:

Quote: On the second article the topic / suggested improvements was just doubling the chipset frequency (without double CAS: just using faster memories would have been enough), which could have allowed doubling the available memory accesses slots and then get this: - low-res 256 colours or 4 + 4 DP mode using the same bandwidth of the OCS low-res with 4 bitplanes; - high-res 256 colours or 4 + 4 DP mode (using the full bandwidth. Like the OCS high-res with 4 bitplanes) or even HAM mode (using only 6 bitplanes); - faster Copper (two times); - faster Blitter (from two up to three times, based on the available free slots). Sounds good. Not sure how easy it would be to do though.

It's easy. You can check the second article, where I explain how it works.
Quote:

Quote: With all those changes my ECS would have been even much better than AGA and a serious competitor with PCs and consoles (for 2D games, of course. But 3D games would have gained a great improvement compared to the Amiga 68000/OCS). 3D games needed a very fast CPU. Once you have that it changes the game - now fast ChipRAM access is important, but you also want chunky. Actually you need chunky or John Carmark et al won't be interested.

No, I wasn't talking about 3D with texture mapping, rather the polygonal 3D which we were used to see on our machines. This could have been around 2.5/3 times faster with my proposals, which I think is a huge improvement.

@bhabbott

Quote:

bhabbott wrote: The A3000 was an expensive mistake. But both Commodore and many Amiga fans didn't understand that the high end wasn't worth chasing.

A3000 is indeed expensive since its chip integration levels were inferior to the competition.

A3000's Amber flicker flicker has its own video memory chips i.e. three relatively low nanosecond access 256K chips.

ECS's 16-bit design wasted A3000's 32-bit Chip RAM.

No graphics hardware exploits Ramsey's 32-bit 25 Mhz memory controller infrastructure.

Define high-end.

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

PC's VGA and SVGA's Amiga 1000 moment started from 1987's IBM VGA and 8514.

IBM VGA and 8514 standards served as the basis for many cost-reduced SVGA clones like the mentioned ET4000AX example.


A1200's Budgie includes a 32-bit memory controller designed for a 68020-type CPU's Fast RAM. Commodore engineers have cost-reduced Buster's and Ramsey's essential functions into Budgie.


Quote:

Fast 640x480p with 256 colors was useless without a fast CPU and lots of RAM to go with it. But that would guarantee a high price and very low sales. What games would support it? Almost none. PCs weren't even using that resolution because it was too taxing. This hardware could not* be put in a low-end machine that would sell in large numbers. Without large numbers the new graphics mode would not be used in games. Result - fail! * before you start telling me how it could be done, remember that 'low-end' meant displaying on a TV - which can't do 640x480 except in flickering interlace and doesn't have the resolution for it either. A complete non-starter!

For desktop applications, my ET4000AX's 640x480p 256 colors are faster than my A1200 AGA's productivity mode 640x480p 256 colors.

ET4000AX has excellent 320x240 256 color gaming performance.

Last edited by Hammer on 24-Nov-2023 at 03:42 AM.
Last edited by Hammer on 24-Nov-2023 at 03:39 AM.
Last edited by Hammer on 24-Nov-2023 at 03:12 AM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@cdimauro

Quote:

The point is that Ramsey is not useful for the chipset. Even because... it's NOT needed: Agnus has already everything inside.

FYI, Agnus/Alice has 68000 bus behavior while Ramsey has 68020/68030 bus behavior.

Agnus/Alice needs to be resigned for 68020/68030 CPUs.

Ramsey proves Commodore engineers can design 25 Mhz memory controllers.

A1200's Budgie includes a 32-bit memory controller designed for a 68020-type CPU's Fast RAM. Commodore engineers have cost-reduced Buster's and Ramsey's essential functions into Budgie.

Quote:

25Mhz is not enough. You need 28Mhz to be perfectly in sync with the raster.

FYI, Apollo's SAGA from V4 has a maximum pixel clock of around 85 Mhz.

Tseng Labs ET4000AX has a 14.31818 MHz crystal oscillator with an ICS2494N video clock generator, which runs at 89.8 MHz. The clocks generated are 25.175 MHz (for VGA mode), 28.322 MHz (for VGA and CGA modes), 32.514 MHz (for EGA mode), and 40.0 MHz (for "Extended" modes).

ET4000AX supports 50 to 70 Hz vertical refresh rates and screen resolutions up to 1024 x 768.

25 Mhz partly covers VGA mode.

Last edited by Hammer on 24-Nov-2023 at 03:11 AM.
Last edited by Hammer on 24-Nov-2023 at 03:06 AM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@bhabbott

Quote:

If they weren't going to do what I suggested then it would be best to just concentrate on making cheaper machines to flood the market until they were ready for new chips. No A3000, no CDTV, just heaps of A500s and A2000s at prices nobody could turn down. Then deliver AGA in early 1992.

FYI, AA3000 (A3000+) was mostly completed around February 1991. Commodore management wasted more than 6 months on full 32-bit 68K ECS designs.

In recent times, AA3000 (A3000+) design was given to the Amiga community for the retro market.

https://bigbookofamigahardware.com/bboah/product.aspx?id=23

The Amiga 3000+ was the first computer based on the Pandora chipset (which was later dubbed AA, then AGA). Revision 0 of this system first booted successfully in February of 1991, thanks due to a chip revision that got the display logic actually working. This is revision 1, which completed the audio subsystem, and moved to surface-mount components.


Xmas 1991 could be a possibility if Commodore allocated the full R&D resources for Q4 1991 AGA release.

AGA itself was on top of Commodore management's "no new chips" time-wasting directive during A3000's R&D phase.

For Doom Q4 1993, the PC has a longer time to create a full 32-bit X86 CPU with a VGA configuration install base since it started in 1987.

Full 32-bit CPU equipped Amiga 500/2000/3000 couldn't participate in Amiga's AGA gaming platform target. The game console chip topology doesn't work on desktop computers.

Commodore doesn't sell AGA chipsets to other vendors to cover the motherboard upgrade path for existing A500/A2000/A3000 users.

Last edited by Hammer on 24-Nov-2023 at 03:55 AM.
Last edited by Hammer on 24-Nov-2023 at 03:54 AM.
Last edited by Hammer on 24-Nov-2023 at 03:51 AM.
Last edited by Hammer on 24-Nov-2023 at 03:35 AM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Hammer

Quote:

Hammer wrote: @bhabbott A3000 is indeed expensive since its chip integration levels were inferior to the competition. A3000's Amber flicker flicker has its own video memory chips i.e. three relatively low nanosecond access 256K chips.

"read my lips: no new chips"...
Quote:

ECS's 16-bit design wasted A3000's 32-bit Chip RAM.

Surely. Like the OCS for systems with 68020+ processors
Quote:

No graphics hardware exploits Ramsey's 32-bit 25 Mhz memory controller infrastructure.

Because, and as I've already stated other times, the chipset has ITS OWN, independent, memory controller.

Could you please tell me when you'll study how the Amiga chipset works?
Quote:

Define high-end.

not low-end.
Quote:

Quote: Fast 640x480p with 256 colors was useless without a fast CPU and lots of RAM to go with it. But that would guarantee a high price and very low sales. What games would support it? Almost none. PCs weren't even using that resolution because it was too taxing. This hardware could not* be put in a low-end machine that would sell in large numbers. Without large numbers the new graphics mode would not be used in games. Result - fail! * before you start telling me how it could be done, remember that 'low-end' meant displaying on a TV - which can't do 640x480 except in flickering interlace and doesn't have the resolution for it either. A complete non-starter! For desktop applications, my ET4000AX's 640x480p 256 colors are faster than my A1200 AGA's productivity mode 640x480p 256 colors. ET4000AX has excellent 320x240 256 color gaming performance.

320x240@256 colours is A LITTLE BIT different (and MUCH LESS demanding) compared to 640x480@256. Isn't it?
Quote:

Hammer wrote: @cdimauro Quote: The point is that Ramsey is not useful for the chipset. Even because... it's NOT needed: Agnus has already everything inside. FYI, Agnus/Alice has 68000 bus behavior while Ramsey has 68020/68030 bus behavior.

Behaviour? Do you know / understand how a 68000 platform (so, not specifically an Amiga) works? I don't think so.
Quote:

Agnus/Alice needs to be resigned for 68020/68030 CPUs.

If you mean the access pattern for CPUs, yes: I've already written on my article.

If it was something else, then please clarify it.
Quote:

Ramsey proves Commodore engineers can design 25 Mhz memory controllers.

Clap clap clap...
Quote:

A1200's Budgie includes a 32-bit memory controller designed for a 68020-type CPU's Fast RAM. Commodore engineers have cost-reduced Buster's and Ramsey's essential functions into Budgie.

Clap clap clap...
Quote:

Quote: 25Mhz is not enough. You need 28Mhz to be perfectly in sync with the raster. FYI, Apollo's SAGA from V4 has a maximum pixel clock of around 85 Mhz.

And? What's the point here?
Quote:

Tseng Labs ET4000AX has a 14.31818 MHz crystal oscillator with an ICS2494N video clock generator, which runs at 89.8 MHz. The clocks generated are 25.175 MHz (for VGA mode), 28.322 MHz (for VGA and CGA modes), 32.514 MHz (for EGA mode), and 40.0 MHz (for "Extended" modes). ET4000AX supports 50 to 70 Hz vertical refresh rates and screen resolutions up to 1024 x 768. 25 Mhz partly covers VGA mode.

And? What's the point here?