@kolla

in 2024 television is digital and normal tv don't display analog output from amiga chipset
in 2024 amiga chipset is worth nothing
in 2024 cpu other than pc is one and only thing from classic amiga that still matter

@Kronos

Quote:

where the casual user

Who? And why do this "casual user" matter more than _actual_ users?
What value does "casual user" add?

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@kolla

And who are those "_actual_users"?

People who insist that something that makes no difference whatsoever adds value.

A.k.a. fanboys.


Remember you are the one claiming that an obsolete chipset (or a fake one that never existed) emulated in a FPGA somehow creates value over emulating that same stuff in SW.

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

@hardwaretech

This is meaningless debate. first of all, what’s the point?

Create platform that has future? Or create platform to be retro arcade.
The users here do not agree on even this.

Emulation is good for emulating old systems. But is imprecise, its not good for high accuracy timing, and costly, if your software is emulating us pixel clocks.

this why I’m a fan of hardware should do hardware thigs, and software should do software things. Most ideally you have native code, but emulation is option, when you have no other options.

Also comparing the FPGA to a real CPU also does not make much sense, a FPGA does not have blocks that represent micro-opcodes. The block in a FPGA is really basic, compared to micro-opcodes, this is why FPGAs don’t scale well for CPU emulation, its good for emulation of other hardware, its not multipipeline out of order, with branch prediction. That’s the domain of real CPU.

JIT is middle ground, a transformation of alien code, into something the CPU can understand, but is not cost free, its not native in the sense its optimized code for CPU, its brute force recompilation of code, the source format is not optimized for the native CPU. as such its not a future for a platform that wants to be relevant.

CPU is optimized for what it does, chips are optimized for there task, and FPGA and software emulation is whys suboptimal simulate or emulate. A imagined or dead platform.

As for AI it has no relevance in this debate, its hosted mostly on a server, large language models are not designed to run computer code, it’s a prediction model, with a ability to grow logical networks that can simulate different aspects, you might think this the future of everything, but its not optimized for running code, it’s a bit like FPGA, its basic network of logical blocks, that designed to perform general tasks. As such the basic ASIC will outperform AI on dumb operations.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@NutsAboutAmiga

Quote:

This is meaningless debate. first of all, what’s the point?

And as per habit the usual contributors chime in, extending the thread into umpteen-and-a-half page with 'facts' and the inevitable difference in interpretations thereof. Up until now only Karlos is absent. He's not unfit somehow, is he?

OldFart

_________________
More then three levels of indigestion and you're scroomed!

@hardwaretech

Everyone has their unique way thinking about it, for most part you can go emulation, or you can go with real hardware, the speed of X5000 is one of the few computers you can’t emulate at this moment, you can emulate speed of X1000.

I believe this is because X1000 has DDR2 memory while X5000 has DDR3 memory, in many ways the PA-Semi is better, but is an older design, the difference in internal clock speed is not that different. It really comes down to memory and bus speeds.

P1022 has DDR3 memory, it can be that in some case A1222 be hard to beat in some cases, actually, FPU emulation issue, is not clear how big is, well it depends on if SPE can be supported better, and I believe even ifs not support latest C++, as long as you can compile opcodes, you can build a C++ class that sets and reads from stack, so should be possible work around this problems.

We know QEMU is not efficient at FPU emulation, it might be that P1022 will outperform QEMU, on some games.

The debate if emulation is better or worse, can be as simple as can you do what want or not, and in case of emulation right now, I believe can’t do some of things, a real X1000 or X5000 or even a A1222 can do, and it comes down to driver support what graphic cards you can use.

Last edited by NutsAboutAmiga on 11-May-2024 at 08:17 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

Kronos Quote:

You are confusing 2 very different and distinct markets: a) people who just wanna play some old games on a cheap&cute device that they gonna throw into the attic soon after b) people who want to use über fast 68k Amigas (Ataris, or whatever) without the need for SW emulation ... For group a) an ASIC might be viable, but than SW or HW emulation at that spec can be had for $1 (when bought in volume) so it hardly makes sense neither.

I'm not confusing markets, I'm combining them. The same 68k ASIC SoC can target multiple markets like the RPi did but more.

RPi markets
hobby
education
embedded
retro (Acorn RISC OS)

ASIC 68k SoC markets
retro (68k Amiga, Atari, Mac, x68000, NeoGeo, Genesis/Mega Drive, etc.)
hobby
embedded
education

A cycle exact fully static 68000 core has a negligible cost in silicon of an ASIC and can also be used as a deterministic I/O processor. Even higher performance cores like 68060 cores should only be a few cents of silicon. A 68060+AA+ SoC could be mass produced for under $1.

agami Quote:

And the Amiga OS was the other half of what made an Amiga system. That said, since AmigaOne systems only have the OS and no custom chipset, shouldn't they actually be branded as AmigaHalf? Or Amiga.5, or AMIGA / 2

The History of the Commodore Amiga - Rare Jay Miner Speech AmiExpo 1990
https://youtu.be/n-MqC35aWrQ

Listen to Jay Miner and you will find that the Amiga consists of 3 parts.

1. 68k
2. chipset
3. AmigaOS

Remove any of the 3 parts and Amiga compatibility is lost. Listen to the YouTube video above to find out how important the 68k and compatibility are to Jay Miner or just look at how small the AmigaNOne market is that could be branded Amiga/3 instead.

NutsAboutAmiga Quote:

Also comparing the FPGA to a real CPU also does not make much sense, a FPGA does not have blocks that represent micro-opcodes. The block in a FPGA is really basic, compared to micro-opcodes, this is why FPGAs don’t scale well for CPU emulation, its good for emulation of other hardware, its not multipipeline out of order, with branch prediction. That’s the domain of real CPU.

There are complex and specialized blocks in FPGAs like SRAM blocks, multipliers, memory controllers, SerDes blocks, etc. Complex blocks can dramatically improve performance. The reason FPGAs have less performance than ASICs is inefficient routing and mismatched resources.

NutsAboutAmiga Quote:

As for AI it has no relevance in this debate, its hosted mostly on a server, large language models are not designed to run computer code, it’s a prediction model, with a ability to grow logical networks that can simulate different aspects, you might think this the future of everything, but its not optimized for running code, it’s a bit like FPGA, its basic network of logical blocks, that designed to perform general tasks. As such the basic ASIC will outperform AI on dumb operations.

AI capable hardware is being deployed today from Apple CPU Neural Engines to edge of network CPUs using AI to filter and reduce IoT data that needs to be transferred. There are real applications for AI but it will never take over most of computing for reasons I gave in my first post on AI.

@matthey

Quote:

I'm not confusing markets, I'm combining them. The same 68k ASIC SoC can target multiple markets like the RPi did but more.

Sure if you can build an ASIC that is so powerful that it can compete with rPI or atleast current Vampire offerings while also being cheap enough to fit into the chip budget of $20 emu toys (which really is in the $1 range).

Wishful thinking.

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

@OldFart

Quote:

OldFart wrote: @NutsAboutAmiga Quote: This is meaningless debate. first of all, what’s the point? And as per habit the usual contributors chime in, extending the thread into umpteen-and-a-half page with 'facts' and the inevitable difference in interpretations thereof. Up until now only Karlos is absent. He's not unfit somehow, is he? OldFart

Not as such, but I'm basically in agreement with the sentiment here. It's all been said umpteen times before. Previously there was some additional angle or facet to tease apart, but not lately.

So. Did anyone watch Fallout?

_________________
Doing stupid things for fun...

@Kronos

Quote:

Kronos wrote: @kolla And who are those "_actual_users"?

I’m one, and I’m certainly not alone, I why would I waste so much desk space on Amiga and FPGA systems if I wasn’t using them? I’m not a casual user, I use my computers quite systematically, the FPGA systems remain in place (since they are small) while real Amiga systems are on rotation, one or two at a time.

Quote:

Remember you are the one claiming that an obsolete chipset (or a fake one that never existed) emulated in a FPGA somehow creates value over emulating that same stuff in SW.

Well, it clearly does. Or why else would the FPGA systems be so popular?

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@ppcamiga1

Quote:

ppcamiga1 wrote: start the same shit again again again and again

Uh huh.

Quote:

we already have good pcs every one here has at least one and dont need another one

*Nods*

Quote:

we buy amiga because we want something different not boring pc Amiga is hobby

*Warning signs of impending stroke from being in agreement this far in to one of your posts*

Quote:

performance no matter

Phew. I thought I was a gonner for a min. It matters if you want that hobby to be sustainable. There's room for expensive vanity systems but they shouldn't be the only game in town, especially when the vanity aspect is all that is left of the USP.

Last edited by Karlos on 12-May-2024 at 09:02 AM.

_________________
Doing stupid things for fun...

@kolla

Quote:

kolla wrote: . Or why else would the FPGA systems be so popular?

But are they really that popular?

All the xxxxMini products seem to based on SW EMU (on an ARM) and I do think they outsell anything FPGA.

Then there are those who just run the EMUs on non-dedicated HW and of course the billions who don't give a damn.

So in reality "the value" is mostly sentimental which really is "no value" for anybody else.

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

Kronos Quote:

Sure if you can build an ASIC that is so powerful that it can compete with rPI or at least current Vampire offerings while also being cheap enough to fit into the chip budget of $20 emu toys (which really is in the $1 range). Wishful thinking.

Like the $1 USD (less in quantity) RP2040 ASIC SoC that has been used in at least 26 products with over 10 million chips produced in a year?

https://en.wikipedia.org/wiki/RP2040
https://www.tomshardware.com/news/raspberry-pi-10-million-rp2040s

Yes, it has been used in primitive video game toys like the Thumby handheld game console. A 68060+68000+AA+ ASIC SoC would use fewer transistors than the RP2040 SoC but would require external memory (the RP2040 SoC is a MCU with on chip SRAM that uses many transistors). The 68k SoC performance could easily be increased by adding 32kiB I+D L1 caches and a 2MiB L2 which could also be configured as a 2MiB SRAM MCU. The StarFive JH7110 SoC has a similar configuration.

JH7110
4xU74 cores 32kiB I+D L1
1xS7 core 16kiB I, 8kiB DTIM (monitor/IO/embedded core)
2MiB L2 cache (configurable as 2MiB SRAM)
IMG BXE-4-32 MC1 GPU
chipset and I/O

68k SoC
2x68060 cores 32kiB I+D L1
1x68000 core (monitor/IO/embedded,retro support core)
2MiB L2 cache (configurable as 2MiB SRAM)
GPU to be determined
chipset and I/O

The JH7110 SoC sold for $3 USD in low volume for development purposes so may not reflect actual production costs. It has been used in sub $100 SBCs though. It makes a very nice MCU with 2MiB of SRAM which is enough to do a lot of computing. It just happens to be the 68020+AGA standard memory so we know how useful 2MiB of memory is on Amiga. A turtle mode may be needed to slow some 68k Amiga software down though as SRAM is blazing fast.

Kronos Quote:

But are they really that popular? All the xxxxMini products seem to based on SW EMU (on an ARM) and I do think they outsell anything FPGA. Then there are those who just run the EMUs on non-dedicated HW and of course the billions who don't give a damn. So in reality "the value" is mostly sentimental which really is "no value" for anybody else.

The Minis tend to get the eye candy nostalgia right but miss the boat on the hardware. Retro customers would love to have more faithful hardware than the Mini money grabs. Some enjoy hacking the hardware to do more despite the limited potential of many cheap Minis. Give them more faithful hardware that is hackable and versatile so they turn into returning developers.

FPGA hardware tends to be more expensive and lower production. Most of the focus has been on the hardware although there have been a few professional FPGA Mini like products.

https://www.analogue.co/mega-sg Quote:

We set out to design the definitive way to explore Sega's 16-bit and 8-bit era. A reimagining of the underdog that led a 16-bit revolution. Engineered with an FPGA. No emulation. 1080p. Zero lag. Total accuracy. Mega Sg is not a plug n' play toy. Compatible with the 2,180+ Sega Genesis, Mega Drive and Master System game cartridge library. ... Mega Sg has the same unparalleled compatibility as Super Nt. The core functionality of each system is engineered directly into an Altera Cyclone V, a sophisticated FPGA. We spent thousands of hours engineering each system via FPGA for absolute accuracy. Unlike the knock off and emulation systems that riddle the market today, you'll be experiencing Sega's 16-bit and 8-bit era free of compromises. Mega Sg is designed to preserve video game history, with the respect it deserves.

SNES version:
https://www.analogue.co/super-nt

These are sold out already at $199.99 USD (controller not included).

MiSTer hardware is more versatile multi-system FPGA simulation hardware that is more expensive. The base Intel/Altera FPGA developer boards ran out during the COVID pandemic. The base FPGA boards have likely sold at least hundreds of thousands and more likely millions but it is difficult to break out the embedded and MiSTer sales. I believe there are at least tens of thousands of MiSTer users and more likely hundreds of thousands. There are also MiST, FPGA Arcade, Flea FPGA Ohm, Chameleon, etc. versatile FPGA users. It's not a huge market because CPU simulation of more advanced CPUs requires a more expensive FPGA making the high end versatile hardware expensive. Simulating just the chipsets allows a cheaper FPGA. For example, the Flea FPGA Ohm SBC sold for $45 USD using a small FPGA more appropriate for chipset only simulation.

There is also Amiga specific FPGA hardware like the MiniMig and Vamp/AC hardware, etc. I don't think the MiniMig was too popular as it was more of a low end Amiga replacement than an upgraded Amiga. The Vamp/AC hardware aims to be an upgrade which is popular but FPGA CPU performance is limited even with more expensive FPGAs and the hardware is not cheap so the value (performance/$) is not competitive with ASICs or even emulation on an ASIC. The PiStorm is taking a bite out of Vamp sales. The PiStorm32-Lite also uses a FPGA for easier firmware upgrades.

@matthey

Quote:

matthey wrote: Like the $1 USD (less in quantity) RP2040 ASIC SoC that has been used in at least 26 products with over 10 million chips produced in a year?

Well call me when you transformed 68k+AGA into a design ready for mass production (with some revisions till you get it right) and have prepaid that 6 or 7 digit production run so economics of scale can kick in.

Now provide SW support as wide and good as for rPI (or ARM in general) and you have an "also ran".


You are off course free to believe what you want, but even in the retro market many just don't care about FPGA or even ASIC (unless they are the real NOS thing) and outside the retro market you need to much better in at least one way to compete with rPI, rPI clones and x86/RISCV SBCs.
Or Arduino and microcontrollers if you aim low.

And don't forget, that once you go mass market patent and IP owners/trolls will come out of the woodworks.


Still wishful thinking

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

ASIC is a fantasy. At least with FPGA and emulation you can fix the inevitable bugs you end up with in your solution and refine it.

Pi exists and Emu68 is already far faster than any physical 68K implementation. It's affordable enough for the toy segment and fast enough for the Uber 68 fans. Well, those not addicted to UAE on some overclocked x64 monster anyway.

_________________
Doing stupid things for fun...

Karlos Quote:

ASIC is a fantasy. At least with FPGA and emulation you can fix the inevitable bugs you end up with in your solution and refine it.

There is an option to accept obsolescence with no more Amiga hardware and EOL emulation of the Amiga. That is the easy and low risk route with guaranteed success to make the Amiga disappear forever. An ASIC is feasible and less fantasy than dropping 6 figures on hardware like the A1222 hoping to bring the AmigaNOne back with bastard PPC hardware after 20 years.

Karlos Quote:

Pi exists and Emu68 is already far faster than any physical 68K implementation. It's affordable enough for the toy segment and fast enough for the Uber 68 fans. Well, those not addicted to UAE on some overclocked x64 monster anyway.

The 68k is on ancient silicon which is THE problem. A 100MHz FPGA CPU is not a solution. Emulation is EOL as the Amiga fades away.

@matthey

Quote:

There is an option to accept obsolescence with no more Amiga hardware and EOL emulation of the Amiga. That is the easy and low risk route with guaranteed success to make the Amiga disappear forever. An ASIC is feasible

Until anyone creates a new 680x0 ASIC its dreamland stuff, let peaple use what is available.

Quote:

and less fantasy than dropping 6 figures on hardware like the A1222 hoping to bring the AmigaNOne back with bastard PPC hardware after 20 years.

Perhaps it is, but you can buy A1222 now, you can't but a new 680x0 asic.

Quote:

A 100MHz FPGA CPU is not a solution. Emulation is EOL as the Amiga fades away.

Well 100Mhz is too much for anyone who only are playing 7Mhz games.
They do not care about new games, or anything like that, they are not developers.

Last edited by NutsAboutAmiga on 12-May-2024 at 09:05 PM.
Last edited by NutsAboutAmiga on 12-May-2024 at 09:04 PM.
Last edited by NutsAboutAmiga on 12-May-2024 at 09:03 PM.
Last edited by NutsAboutAmiga on 12-May-2024 at 09:02 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@matthey

Quote:

matthey wrote: An ASIC is feasible

An ASIC is 7 figure adventure with most of it spend at least 1 year before any returns.

So unless you got a proper business plan and the financing...


... wishful thinking,

I mean there is a reason why the rPI foundation started with a readily available low end chip (doing the PCB is the easy part) and focused on SW and utility for several generations until they have reached the size where doing semi custom made sense (those chips are still based on existing and fully validated cores).

As for the Amiga disappearing forever, that ship has sailed decades ago and once biology takes its turn it will just as common as someone fiddling with steam tractors today.

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

@matthey

Quote:

Emulation is EOL as the Amiga fades away.

Oh man. I just can't do it. It's like shooting a puppy. Can someone else break the news to him?

_________________
Doing stupid things for fun...

NutsAboutAmiga Quote:

Perhaps it is, but you can buy A1222 now, you can't but a new 680x0 asic.

The A1222 hardware is only available now because of poor planning.

Kronos Quote:

An ASIC is 7 figure adventure with most of it spend at least 1 year before any returns. So unless you got a proper business plan and the financing... ... wishful thinking,

A-Eon signed a $1.2 million USD contract for Ultra Varisys to design, develop and manufacture PPC hardware way back in 2013.

A-EON Technology & Ultra Varisys sign $1.2M agreement for new PowerPC hardware
https://web.archive.org/web/20140328052600/http://www.a-eon.com/18-10-2013-3.pdf

The U.S. inflation calculators says that $1.2 million USD in 2013 adjusted for inflation in 2024 is $1,608,873.74 USD. I expect much more than this has been flushed down the drain by Trevor. More recent money flushing in the last 10 years obviously has zero chance of financial success. Everyone knows it which is why Ben stopped AmigaOS 4 development. Even Trevor knows it as he has talked about the Amiga money pit at shows. These kinds of figures are nothing special for small businesses today. What is special is continuing to lose money with no hope anyone sane would invest money in it which is what PPC AmigaNOne was a decade ago and the situation is much worse today. There are likely investors who would invest in the hot retro gaming market though.

Kronos Quote:

I mean there is a reason why the rPI foundation started with a readily available low end chip (doing the PCB is the easy part) and focused on SW and utility for several generations until they have reached the size where doing semi custom made sense (those chips are still based on existing and fully validated cores).

The RPi Foundation had good leadership and planning but they never expected the resulting sales volume. It was convenient to be able to use commodity ARM SoCs but they haven't been able to get exactly the hardware they want. Their hardware standards and consistency across RPi models is less than desirable. Their GPU standard is low end although a more open and powerful GPU standard is a challenging hurdle.

Last edited by matthey on 13-May-2024 at 05:48 AM.
Last edited by matthey on 12-May-2024 at 10:21 PM.