@matthey

Main reason for ARM SoCs being relatively cheap is the enormous amount of chips being produced for a wide variety of 'systems'. So for that 68k SoC the production numbers should be extremely high to get in the ARM SoC price region. I think for those not owning classic or OS4 Amiga hardware who want to get back for the casual game or two, emulation on ARM is the most cost effective way.

_________________
Back home...

@Amiga_3k

Quote:

Amiga_3k wrote: @matthey Main reason for ARM SoCs being relatively cheap is the enormous amount of chips being produced for a wide variety of 'systems'. So for that 68k SoC the production numbers should be extremely high to get in the ARM SoC price region. I think for those not owning classic or OS4 Amiga hardware who want to get back for the casual game or two, emulation on ARM is the most cost effective way.

And the experience from most people is that the emulation is "good enough"! While demos seem to suffer from slowdown on THE A500 Mini its not a deal breaker for most. My worst peeve presently is graphics glitching on Antheads, but not many would pay £hundreds to play that better on a 68k based SOC!

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

matthey Quote:

A 68k SoC ASIC could bring a lower price

kolla Quote:

Doubtful!

THEA500Mini (Allwinner H6 SoC)
4 core ARM Cortex-A53@1.8GHz
Mali-T720 MP2@600Mhz 3D GPU
512MiB memory
256MiB NAND Flash

Is all this hardware needed to emulate a 68000@7MHz Amiga with 512kiB to 1MiB of memory?

The SoC likely only costs $5-$10 in quantity which may seem cheap at first glance but there is a huge amount of area wasted compared to the area of an Amiga 500 or even a 68060+AGA Amiga. The Cortex-A53 512kiB L2 is the same size as the main memory of the original Amiga 500. This is likely SRAM and could be zero wait state memory on custom hardware with a programmable wait state. A 68000 CPU core is roughly 68,000 transistors. ARM cores are small though right?

Transistor counts
ARM1 25,000
ARM2 27,000 - hardware multiply, fast interrupts
68000 68,000
68020 190,000 - 256B I cache, 32 bit address bus
68030 270,000 - 256B I+256B D cache
ARM3 310,000 - 4kiB unified cache
ARM6 360,000 - 4kiB unified cache, 32 bit address bus
ARM7 579,000 - 8kiB unified cache, Thumb ISA
68040 1,170,000 - 8kiB I+8kiB D cache
68060 2,530,000 - 4kiB I+4kiB D cache
ARM11 17,370,000 - Thumb 2 ISA, 8 stage pipe, OoO completion, split caches, FPU and SIMD units
Cortex-A53 20,000,000-25,000,000? - 64 bit in order, AArch64 ISA, L2
Cortex-A9 26,000,000 - OoO 32 bit, 8 stage pipe, L2
Atom Silverthorne 47,212,207 - 64 bit in order 16-19 stage pipe, low power x86-64

There are 4 ARM Cortex-A53 cores in the SoC though L2 is shared so less than 4 times the above numbers. Half the CPU area could be for CPU caches and then the GPU may be larger than the area used for CPU cores and caches (64 bit and fat AArch64 requires more caches too). A 68k Amiga SoC ASIC, including custom chips, could use a fraction of the area of this generic SoC which reduces cost. The generic SoC is using a more expensive 28nm process than needed because of inefficient emulation. All that for China backdoors with Allwinner.

https://en.wikipedia.org/wiki/Allwinner_Technology#Backdoor_controversy
https://arstechnica.com/information-technology/2016/05/chinese-arm-vendor-left-developer-backdoor-in-kernel-for-android-pi-devices/?comments=1

The Raspberry Pi Foundation knows to avoid but not the competition like Zidoo H6 Pro, Orange Pi One Plus, Orange Pi Lite 2, Orange Pi 3, PINE H64 model A and B.

matthey Quote:

better performance

kolla Quote:

For what? The old 68k games that don't need better performance? For the carousel front end?

The FleaFPGA Ohm was $45 and potentially had more accurate simulation of an A500 than THEA500 Mini. What did it lack? An eye candy case, mouse and controller, included games, out of box ease of use, support and performance were lacking. Performance is actually a big attraction of existing Amiga customers though it does imply general purpose use and not just game use like THEA500 Mini.

matthey Quote:

more accurate 68k execution

kolla Quote:

What does that even mean? Accurate how, accurate compared to what?

Original hardware is the accuracy baseline.

1. original hardware
2. ASIC recreation of original hardware
3. FPGA hardware
4. emulation
5. JIT emulation

There was different original hardware with some later hardware being less compatible to the earlier hardware. Today, technology allows to improve compatibility without sacrificing as much performance. Examples include adjustable CPU clocks with fully static core designs (actually the 68060 and 68040V are fully static designs allowing clock speeds to be reduced from max to zero but only more modern Amiga accelerators have adjustable clock speeds), bus snooping for self modifying code, eased alignment of data in AGA, etc.

matthey Quote:

better compatibility

kolla Quote:

With what? Games made for 7.14 MHz 68000? Games made for 68020+? Or better compatibility with WHDLoad?

It would be possible to create cycle exact hardware of a particular model but that would limit performance. I believe it would be better to create flexible hardware with compatibility features as mentioned above.

matthey Quote:

and better I/O

kolla Quote:

What i/O? Better as in faster? Better as in more modern? Better as in "not emulated"? Well, that is NOT the CPU, that is the chipset - so if you want to new 68k SoC, you also need a whole new Amiga chipset implementation with new peripherals, and last but certainly not least - software support! How will you do that? And how are old games supposedly going to support all this new stuff? And once all this is in ASIC, there's no going back - you are stuck with it, bugs and all.

Digital video (HDMI and DVI) has been successfully added to FPGA Amiga hardware. Solid state cards and drives have been added to FPGA Amiga hardware. Ethernet and Wi-Fi has been added to FPGA Amiga hardware. USB has been added to FPGA Amiga hardware even though it may be lacking in some cases. This is how a SoC is created. Yes, it is faster and more modern than original hardware. Yes, there is likely more work to be done to create a good SoC. Currently, efforts are divided in modern Amiga fashion.

kolla Quote:

Why all this hassle when software emulation is PLENTY good enough, and allow for easy updates and improvements?

Emulation is inefficient and the sign of a dead platform. Modern affordable 68k Amigas would be useful and there are many people who would buy them. It is about improvement rather than "640k ought to be enough for anybody" attitude. The visionary Jay Miner understood the importance of integrating and enhancing the Amiga which he pushed CBM to do for the Amiga. Technology is what created the Amiga but it has passed the Amiga by even as a single chip Amiga could be produced for about $1. A 68k Amiga SoC ASIC should be produced for historical homage and posterity if nothing else. It is affordable and doable if people worked together. If done professionally, it may even make money and create the user base the Amiga needs.

Amiga_3k Quote:

Main reason for ARM SoCs being relatively cheap is the enormous amount of chips being produced for a wide variety of 'systems'. So for that 68k SoC the production numbers should be extremely high to get in the ARM SoC price region. I think for those not owning classic or OS4 Amiga hardware who want to get back for the casual game or two, emulation on ARM is the most cost effective way.

Economies of scale applies more to chip production where ARM has no particular advantage except for their small low performance cores that use less area. ARM has little if any advantage for higher performance cores. Development costs are lower for ARM ala carte cores but this may be offset by licensing costs with higher production. For a long time, many Amiga users didn't think there was enough interest for an ASIC until THEA500 Mini. I thought there was interest but I had a backup plan anyway. There are embedded chips created in the millions for just one embedded use like IoT sensors. There are CEOs producing embedded chips who don't like ARM and like the 68k. I was talking to one such CEO although he wanted more open hardware than closed hardware Gunnar wanted.

@matthey

Quote:

Economies of scale applies more to chip production where ARM has no particular advantage except for their small low performance cores that use less area. ARM has little if any advantage for higher performance cores. Development costs are lower for ARM ala carte cores but this may be offset by licensing costs with higher production. For a long time, many Amiga users didn't think there was enough interest for an ASIC until THEA500 Mini. I thought there was interest but I had a backup plan anyway. There are embedded chips created in the millions for just one embedded use like IoT sensors. There are CEOs producing embedded chips who don't like ARM and like the 68k. I was talking to one such CEO although he wanted more open hardware than closed hardware Gunnar wanted.

Absolutely.
Based on what I know about the hardware game, there are several potential market use-cases for new 68k-based silicon. Any new 68k ASIC would essentially go beyond merely serving the Amiga retro market and would need to straddle two or more use-cases to justify the investment. The Amiga use-case is unlikely to be the primary. It would be a bonus benefit to other more profitable use-cases.

While the cost per SoC or integrated board in 1M+ quantities can be relatively low, it is in no way commentary on the design effort and development costs that would be required before the first prototypes rolled off the line.

Were a starting gun to eventually fire on such a sufficiently funded project, it would realistically take at least 2 years before such a 68k ASIC integrated board running Linux would be ready for the market.

_________________
All the way, with 68k

agami Quote:

Absolutely. Based on what I know about the hardware game, there are several potential market use-cases for new 68k-based silicon. Any new 68k ASIC would essentially go beyond merely serving the Amiga retro market and would need to straddle two or more use-cases to justify the investment. The Amiga use-case is unlikely to be the primary. It would be a bonus benefit to other more profitable use-cases.

Right. Potential customers can't be expected to show up on the doorstep wanting products either. There is competition and potential customers have to be engaged to find out their requirements, desires and reluctance to switch to another product. There should be product pipeline planning, longevity planning and 2nd source production options or open hardware which can be given to customers. There needs to be developer support as well although compatibility reduces the needed effort. Motorola had to deal with these issues too as can be heard in the following video.

Motorola 68000 Oral History Panel
https://www.youtube.com/watch?v=UaHtGf4aRLs

The embedded market is huge and the 68k has not only potential but experience as the market leader. There are few 68k products on the market where there is likely some customers who need replacements for old 68k and ColdFire products. There are developers who remember the 68k legend and how easy it was to use. There are potential customers who would be attracted to the best code density 64 bit ISA (68k64). There are potential customers who want more open hardware than ARM but don't like RISC-V.

In my opinion, there is enough potential to spend a relatively small amount of R&D money to explore fabless semiconductor development. Emulators and virtual machines are unlikely to lead anywhere like Amiga Nowhere. ARM grew out of Acorn Computers while the PC side died. ARM CPUs were not a easy sell either despite the low power potential for small weak cores. Early ARM CPUs were primitive and had poor code density for the embedded market. It was nearly a decade and a switch to Thumb, a better code density ISA, before they started to become popular for embedded. As Motorola abandoned the 32 bit embedded market leading 68k and converted to PPC for embedded, it was SuperH which took over as the 32 bit embedded CPU leader in the '90s before ARM. ARM took over as 32 bit embedded market leader only when they converted to Thumb and Thumb 2 ISAs which improved the code density to be better than SuperH and finally roughly on par with the 68k. For the Thumb ISAs, ARM licensed the SuperH from Hitachi who was the 2nd source 68000 producer and was sued by Motorola for using 68000 technology in new unlicensed products. It is ironic that Motorola abandoned the 68k technology for PPC yet didn't want competitors using it to beat PPC in the embedded market.

agami Quote:

While the cost per SoC or integrated board in 1M+ quantities can be relatively low, it is in no way commentary on the design effort and development costs that would be required before the first prototypes rolled off the line.

It's not rocket science to develop a mostly in-order core but the result needs to be professional to be competitive. Professionalism requires experience which comes at a cost but is worth it.

agami Quote:

Were a starting gun to eventually fire on such a sufficiently funded project, it would realistically take at least 2 years before such a 68k ASIC integrated board running Linux would be ready for the market.

It may be possible to develop a 68k core in less that 2 years if existing HDL code could be used. It's not as important to rush an ASIC as it was during the Moore's Law heyday. Capital is burning when development is in full swing but extra time for verification and testing could result in a more professional initial product. FPGAs are much cheaper than in the old days and a great development tool for designing an ASIC.

Last edited by matthey on 05-May-2022 at 07:23 PM.

@matthey

Quote:

A 68k SoC ASIC could bring a lower price

How much would it cost to get your 68k SOC production ready, and how much would it cost to produce all the chips needed to hit the price point of a few dollars?

Quote:

better performance, more accurate 68k execution, better compatibility and better I/O

The existing product is adequate for the target market. There's no need to sink multiple millions into a new product that won't be noticeably better to the average customer.

@matthey

It’s not emulation of the 68k CPU that is so resource hungry, it is emulation of the chipset and in particular chipset+68k. And especially with software that really bangs the chipset, like legacy games and demos. On a system like PiStorm, where chipset on motherboard is still used, the emulated 68k flies. We will see how PiStorm32 will perform with 32bit access to AGA motherboards soon, and when AROS/68k eventually gets running on “Emu68 SA”.

You speak of “modern interfaces” as solved problems on FPGA systems, but in reality they are not solved, they are in a state of continuously and constantly being worked on. And how are they implemented? Typically by having a dedicated ARM processor taking care if I/O (MiST and clones). The most complete system in terms of “modern interfaces”, hands down, is MiSTer. And why? Because the entire I/O framework runs on Linux/ARM, so all inputs (keyboard, mouse, joysticks) go through Linux USB stack and input streams.

On the FleaFPGA, the only “modern interface” is HDMI out, but you still need to be lucky with the monitor, and if you go interlaced, you must also set a jumper - very inconvenient for games that jump in and out of interlaced modes.

And the Apollo V4 (vampire no more, it’s all rocket science references from here) struggles with modern interfaces as well, despite years of development.

If you put 68k in ASIC, how will you solve it? ARM controller? FPGA? First 68k thread running I/O “firmware”, second thread running OS? How?

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

Creating the RP2040 costs $5 million

https://www.theverge.com/22966155/raspberry-pi-ceo-interview-eben-upton-computer-chip-shortage-diy

Quote:

[Creating the RP2040] probably cost like $5 million or something. The answer is you sell a lot of them. That’s the answer with all Raspberry Pi products. They are murderously expensive to design, particularly the modern big Raspberry Pis that have radios on them. One thing we do is we go and do radio conformance everywhere. Raspberry Pi is a legal conformant radio emitting product in Tanzania. We take every product everywhere, and that actually piles up a few thousand dollars at a time. Just the radio conformance for a modern Raspberry Pi product is the best part of half a million dollars.

Creating a 68k SOC will probably cost more(?)

kolla Quote:

It’s not emulation of the 68k CPU that is so resource hungry, it is emulation of the chipset and in particular chipset+68k. And especially with software that really bangs the chipset, like legacy games and demos. On a system like PiStorm, where chipset on motherboard is still used, the emulated 68k flies. We will see how PiStorm32 will perform with 32bit access to AGA motherboards soon, and when AROS/68k eventually gets running on “Emu68 SA”.

The Amiga chipset emulation is of an old slow chipset. The chipset bottleneck that existed because CBM was slow to develop and integrate the Amiga can now be removed and that is even more difficult to emulate. Even the FPGA Apollo hardware would require very high performance emulation to duplicate. I would like to see the Amiga chipset advance to at least AA+ specs as CBM was planning to do with modern memory bandwidths.

kolla Quote:

You speak of “modern interfaces” as solved problems on FPGA systems, but in reality they are not solved, they are in a state of continuously and constantly being worked on. And how are they implemented? Typically by having a dedicated ARM processor taking care if I/O (MiST and clones). The most complete system in terms of “modern interfaces”, hands down, is MiSTer. And why? Because the entire I/O framework runs on Linux/ARM, so all inputs (keyboard, mouse, joysticks) go through Linux USB stack and input streams. On the FleaFPGA, the only “modern interface” is HDMI out, but you still need to be lucky with the monitor, and if you go interlaced, you must also set a jumper - very inconvenient for games that jump in and out of interlaced modes. And the Apollo V4 (vampire no more, it’s all rocket science references from here) struggles with modern interfaces as well, despite years of development.

The problem is several separate groups with limited budgets slowly duplicating their work in FPGA instead of combining resources and creating a proper SoC ASIC. The Apollo hardware USB limitations are likely caused by poor hardware planning and limited development time. The FleaFPGA Ohm digital video output incompatibilities are likely caused by limited development resources and time. These are perfectly solvable problems that require effort not normally found for niche products. Do it once and do it right.

kolla Quote:

If you put 68k in ASIC, how will you solve it? ARM controller? FPGA? First 68k thread running I/O “firmware”, second thread running OS? How?

ARM cores are not needed for I/O and neither is Linux. The original Amiga handled I/O elegantly in hardware. Ever heard of DMA?

AmigaNoob Quote:

Creating the RP2040 costs $5 million https://www.theverge.com/22966155/raspberry-pi-ceo-interview-eben-upton-computer-chip-shortage-diy Quote: [Creating the RP2040] probably cost like $5 million or something. The answer is you sell a lot of them. That’s the answer with all Raspberry Pi products. They are murderously expensive to design, particularly the modern big Raspberry Pis that have radios on them. One thing we do is we go and do radio conformance everywhere. Raspberry Pi is a legal conformant radio emitting product in Tanzania. We take every product everywhere, and that actually piles up a few thousand dollars at a time. Just the radio conformance for a modern Raspberry Pi product is the best part of half a million dollars. Creating a 68k SOC will probably cost more(?)

Good article which I had not read. Raspberry Pi Foundation CEO Eben Upton and my philosophies are similar. Several points are what I have been saying.

Quote:

They’re also some of the only readily available computers that are designed to be tinkered with — unlike a smartphone or even really modern desktops, they’re not heavily locked down, and using one requires learning how a computer actually works. And that’s the entire point: Eben told me the idea of the Raspberry Pi was to create a product that enticed kids into studying computer science at the University of Cambridge, where he used to work. Just like the Commodore 64 or the Apple IIe taught a generation of kids how to tinker with computers, Eben wanted to give people an open computer that rewarded experimentation. The goal was simply to sell enough to increase the number of CS applicants by — this is true — 100 students.

Modern computers are often closed locked down hardware and the Raspberry Pi success has come from making "tiny hackable computers that are extremely inexpensive". Has there ever been a more open and easily hackable personal computer than the Amiga? Is the 68k Amiga as easy to use and program as the Raspberry Pi or easier?

Quote:

The Chromebook is an interesting example of an appliance that is very successful at replacing a general purpose computer with an appliance computer. The existence of general-purpose computers is not a God-given right. It’s a historical accident, a path-dependent historical accident. What we came to realize is the thing that had absolutely gone away was the cost-effective general purpose computers. There were lots of cost-effective appliances: cheap games, consoles, and mobile phones. But there weren’t any more low-cost general purpose computers. There probably had never been general purpose computers as low-cost as a Raspberry Pi. Really, the idea was that cost had become a barrier to access to devices that might beguile you into being a computer programmer.

I have been talking about the importance of general purpose computing even as certain features allow to penetrate certain markets like embedded. General purpose open enough hardware allows cross hardware like being able to turn a console into a personal computer, a personal computer into a console, a console into an embedded device and a personal computer into an embedded device. The Amiga pioneered this while the Raspberry Pi Foundation is duplicating this success. Like CBM, the Raspberry Pi Foundation is bringing low cost personal computers to market with multimedia features. In many ways, the Raspberry Pi is more of a CBM and Amiga successor than any Amiga NG.

Quote:

The phrase “software company with recurring revenue” is an in-house joke at Raspberry Pi. I talked to an investor once, and I described all the cool stuff we do. It’s all very well being a software company with recurring revenue, but sooner or later, somebody needs to be making hardware — actually, there’s a business opportunity. If everyone wants to become a software company, there’s a huge business opportunity in not doing that, in being a hardware company. We went to see this investor, and he’s like, “Yeah, you guys are great, but you should totally pivot to being a software company with recurring revenue.”

The markets for easy low risk business activities are saturated. A software only business is easy but the competition for personal computer operating systems is insane and using emulation or virtual hardware only is more insane. The Raspberry Pi Foundation would be nothing without hardware just like the Amiga will continue to be nothing without competitive hardware.

The $5 million that the Raspberry Pi Foundation spent to develop the RP2040 is reasonable. They likely didn't have the people or tools they needed to create it to begin with. They were able to use proven ARM ala carte IP to keep costs down. Existing HDL code for a 68k Amiga SoC could likely be obtained cheaply and there are many motivated Amiga fans who are coders. Experience and leadership are missing which is usually expensive but important. I want a professional ASIC paying homage to the 68k and Amiga chipsets which costs money. Even small businesses can afford $5 million which isn't that much anymore. Ideally, I would like to see 2-3 times that raised before starting. ASICs are affordable for small businesses.

Quote:

Using a custom integrated circuit, or “ASIC”, or “custom chip”, can be a very cost-effective way for small and medium-sized companies to reduce the cost and improve the performance of their products. However, we often encounter the misconception that only large companies can afford a custom ASIC design.

https://www.sigenics.com/page/asics-c

Simple ASICs can be created for less than $1 million as shown by real examples at the bottom of the link above. I want a more complex SoC ASIC that would be expensive to develop but it is not rocket science as exemplified by this basically being done already in FPGA, albeit at a less professional and more limited fashion than desirable. Replacing many chips which allows the board to be simplified to a few dollars is a game changer. Even the ASIC itself can be cheap if small as shown by the RP2040 ASIC which is sold for $0.70 (Raspberry Pi Foundation production cost would be lower). It is likely possible that an Amiga 68k+AA+ SoC ASIC could be reduced to that size although I would like to see a little more put in it because transistors in an ASIC are very cheap and more performance would increase the value. Power is not as much of a concern although it would be advantageous to stay fanless.

@matthey

Can you PLEASE stop jumping around all over the place when discussing with others than yourself? So often your “defense” argument has nothing at all to do with original statement.

So can we now agree that modern I/O, like USB and HDMI, is NOT a solved issue on FPGA systems in general, and even when they exist, it’s with a lot of limitations.

Can we agree that the native Amiga chipset
1 - is necessary for a product like the THEA500 Mini, which whole purpose is to run legacy games?
2 - is the reason why the THEA500 Mini needs the CPU and resources it has! NOT the 68k CPU?
3 - is something you need to implement on any 68k SoC if you intend to run legacy Amiga OS?

Yes, the old Amiga handled I/O elegantly in hardware. Sometimes! And quite often NOT! Serial port being a total PITA. Gayle IDE sucking CPU like nuts. DMA you say, yes, when you’re lucky! How often doesn’t DMA work on zorro3 because of X, Y and Z? Why does my A3000 sometimes need to run the Deneb in PIO mode? All these are rethorical questions, you do not need to answer them in lengthy mostly irrelevant prosa and links… I already know why things are how they are, just want you to accept and admit that Amiga hardware isn’t as rosy as you often claim.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@matthey

One of the key advantages of an FPGA is that can be readily updated to address bugs, improve performance or add new features. An ASIC may be much faster but I think for a hardware solution, given the costs, risk of bugs, etc that your design better be perfect. Which, with the best will in that world, it won't be.

_________________
Doing stupid things for fun...

kolla Quote:

Can you PLEASE stop jumping around all over the place when discussing with others than yourself? So often your “defense” argument has nothing at all to do with original statement. Relax. I was just commenting on the good article AmigaNoob linked. It's nice to read about someone with some business sense and vision for a change. There's not much of that in Amiga Neverland. kolla [quote] So can we now agree that modern I/O, like USB and HDMI, is NOT a solved issue on FPGA systems in general, and even when they exist, it’s with a lot of limitations.

I disagree. I believe the problem with HDMI is not licensing and implementing the IP but rolling a custom implementation. SerDes may be required but are available in more expensive FPGAs. USB may be more reliable too with SerDes and 3.0 may practically require it. These shouldn't be problems for a hardware guy experienced in I/O like Amiga Corporation/Hi-Toro badge #3, Joe Decuir.

https://en.wikipedia.org/wiki/Joseph_C._Decuir

Joe talks about creating a cheap easy to use computer for retro creative use.

The Birth of the Commodore Amiga - Interview with Engineer Joe Decuir 4K UHD
https://youtu.be/9TWuTKJNZIk?t=1075

Too bad that is a Raspberry Pi instead of the Amiga.

kolla Quote:

Can we agree that the native Amiga chipset 1 - is necessary for a product like the THEA500 Mini, which whole purpose is to run legacy games? 2 - is the reason why the THEA500 Mini needs the CPU and resources it has! NOT the 68k CPU? 3 - is something you need to implement on any 68k SoC if you intend to run legacy Amiga OS?

Sure. CPU cores are good at handling sequential data and the chipset is parallel logic. FPGAs are good at parallel logic. GPUs are good at parallel logic too but often have considerable latency.

kolla Quote:

Yes, the old Amiga handled I/O elegantly in hardware. Sometimes! And quite often NOT! Serial port being a total PITA. Gayle IDE sucking CPU like nuts. DMA you say, yes, when you’re lucky! How often doesn’t DMA work on zorro3 because of X, Y and Z? Why does my A3000 sometimes need to run the Deneb in PIO mode? All these are rethorical questions, you do not need to answer them in lengthy mostly irrelevant prosa and links… I already know why things are how they are, just want you to accept and admit that Amiga hardware isn’t as rosy as you often claim.

The lack of serial buffering was due to a transistor budget and cost limitations of the original chipset. It's not like high speed modems were popular when the Amiga 1000 was introduced. Sadly, it was only due to be enhanced in AA+.

https://en.wikipedia.org/wiki/Commodore_AA+_Chipset

The lack of IDE DMA and buffering was due to CBM cheapness but was fixed in the 4000T. It's too bad that the majority of Amiga users never experience an Amiga chipset with bottlenecks removed. Jay Miner wanted to uncork it but the Amiga ARM chair visionaries today are happy with emulation on oppressive monstrosities of PCs.

Karlos Quote:

One of the key advantages of an FPGA is that can be readily updated to address bugs, improve performance or add new features. An ASIC may be much faster but I think for a hardware solution, given the costs, risk of bugs, etc that your design better be perfect. Which, with the best will in that world, it won't be.

Hardware compatibility is easier to test today in FPGA as Amiga compatible FPGA devices demonstrate. Logic simulation and verification is easier than ever on computers. ASIC creation has become a cheap commodity service. IP is readily available for licensing from eDRAM to GPUs. An Amiga on a single chip can be produced for about $1 today and a whole computer for about $5. Compare all this to the hand laid out logic, wire wrapped prototypes, lack of computer simulation and the immense cost of the development and the Amiga custom chips themselves. Joe Decuir talks about some of it in the video I posted above.

The Birth of the Commodore Amiga - Interview with Engineer Joe Decuir 4K UHD
https://www.youtube.com/watch?v=9TWuTKJNZIk

Everybody can do software, emulation and virtual hardware today though. It's just fine for Amiga Nowhere.

Last edited by matthey on 07-May-2022 at 01:48 AM.
Last edited by matthey on 07-May-2022 at 01:43 AM.
Last edited by matthey on 07-May-2022 at 01:42 AM.

@AmigaNoob

Quote:

AmigaNoob wrote: Creating the RP2040 costs $5 million ... Creating a 68k SOC will probably cost more(?)

Doubtful, though of course it would depend just how much of S is on the C.

Based on what I would think of as the first 68k-based SoC which could be used for, among other things, a retro Amiga target, I'd say it would be $2M (+/- 25%).

While the SoC can be developed for under $5M, creating an integrated board with the SoC, and getting it market ready, would push up against the $5M mark (+/- 20%)

Last edited by agami on 07-May-2022 at 02:39 AM.
Last edited by agami on 07-May-2022 at 02:38 AM.
Last edited by agami on 07-May-2022 at 02:36 AM.

_________________
All the way, with 68k

@agami

A 68k SoC with super fast new ASIC 68k CPU, modern RAM (no SDRAM?), AA+ chipset, USB, storage, NIC, wifi, bluetooth, HDMI… no ARM chip in sight, all necessary native AmigaOS drivers and protocol stacks integrated in AmigaOS… and all this to run old legacy games.

Yes, makes total sense.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@kolla

It doesn't make the slightest sense, but if it existed I'd totally buy it.

_________________
Doing stupid things for fun...

A 68k SoC with easy to do bare-metal programming would be nice to have, specifically the GPU which is probably the most closed of them all. Might actually be an untapped market.

Saw this tweets a while back and Raspberry Pi, touted as an open platform, is not actually completely bare-metal programmable when it comes to its GPU and even the framebuffer.

https://nitter.poast.org/cmuratori/status/1470221623307345920#m

https://nitter.poast.org/cmuratori/status/1470451289918033922#m

@matthey

Oh, crap:

Quote:

Hardware compatibility is easier to test today in FPGA as Amiga compatible FPGA devices demonstrate. Logic simulation and verification is easier than ever on computers. ASIC creation has become a cheap commodity service. IP is readily available for licensing from eDRAM to GPUs. An Amiga on a single chip can be produced for about $1 today and a whole computer for about $5. Compare all this to the hand laid out logic, wire wrapped prototypes, lack of computer simulation and the immense cost of the development and the Amiga custom chips themselves. Joe Decuir talks about some of it in the video I posted above.

Again and again with this totally unrealistic statement about your dream $1 ASIC 68K Amiga chip? How many times I had to read it? Repeating this a million of times will not turn it into a true statement. You don't seem to have any clue about scale economics, really.

kolla Quote:

A 68k SoC with super fast new ASIC 68k CPU, modern RAM (no SDRAM?), AA+ chipset, USB, storage, NIC, wifi, bluetooth, HDMI… no ARM chip in sight, all necessary native AmigaOS drivers and protocol stacks integrated in AmigaOS… and all this to run old legacy games. Yes, makes total sense.

Such hardware could be used for many things as it would be general purpose and cheap like the Raspberry Pi. One of the potential uses which seems more viable because of THEA500 Mini success is for legacy gaming.

Karlos Quote:

It doesn't make the slightest sense, but if it existed I'd totally buy it.

I don't think you are alone and with enough people who would buy it it makes sense.

paolone Quote:

Again and again with this totally unrealistic statement about your dream $1 ASIC 68K Amiga chip? How many times I had to read it? Repeating this a million of times will not turn it into a true statement. You don't seem to have any clue about scale economics, really.

The Raspberry Pi Foundation RP2040 ASIC is available for $.70.

https://en.wikipedia.org/wiki/RP2040

Let's look at high transistor cost pieces in the SoC.

2x ARM M0+ 25,000-75,000 transistors
264kiB SRAM 12,976,128 transistors (SRAM uses at least 6 transistors per bit)
---
~13 million transistors from 2 cores and 264kiB of SRAM

A hypothetical minimal 68000+ECS SoC may use roughly 1% of the transistors of the RP2040.

68000 68,000 transistors
ECS chipset 60,000 transistors
---
128,000 transistors

A hypothetical 68060+AA+ SoC may use 21% of the transistors of the RP2040.

68060 2,530,000 transistors
AA+ chipset 200,000 transistors
---
2,730,000 transistors

68k transistor counts come from Motorola sales materials and Amiga chipset transistor counts come from the following link.

https://en.wikipedia.org/wiki/Commodore_AA+_Chipset

The RP2040 has no video output or sound while I think everyone knows what the Amiga has. AA+ even comes with 16 bit chunky graphics and 16 bit sound. The 68060 which was expensive to produce in the '90s even looks relatively small today. There weren't many wasted transistors back in the Amiga day because they couldn't afford to waste transistors. Bring that technology forward and it is tiny and cheap to produce.

@kolla

Quote:

kolla wrote: @agami A 68k SoC with super fast new ASIC 68k CPU, modern RAM (no SDRAM?), AA+ chipset, USB, storage, NIC, wifi, bluetooth, HDMI… no ARM chip in sight, all necessary native AmigaOS drivers and protocol stacks integrated in AmigaOS… and all this to run old legacy games.

Hey now, I clearly said “ it would depend just how much of S is on the C.”, and I never said anything about Amiga OS drivers and protocol stacks.

The first version of the integrated board would most likely use off-the-shelf MCUs/modules for WiFi/Bluetooth, Ethernet, and other bits that make sense, which could very well be ARM-based. Not trying to boil the ocean.

Also, the primary use case for the integrated board would not be the Amiga retro/revival market. It would be aimed at a more profitable use case, and would be made market ready with Linux.

The secondary/tertiary use case would be Amiga 68k retro/revival. Development of Amiga OS/AROS drivers and protocol stacks would be a completely separate endeavor.

_________________
All the way, with 68k

@agami

So you’re saying you and Matthey cannot even agree on an imaginary SoC… and you wonder why others cannot join efforts on real projects?

Here’s your chance, both of you - figure out together here in the open how this imaginary SoC of yours will work, which hardware to use, what features the 68k CPU must have, what bus technologies to use, which operating systems and software you want it to run, and how that can accomplished in terms of drivers etc.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC