@matthey

Quote:
Maintaining a standard API between PPC and 68k AmigaOS would have prevented the split between 2 factions and made 3rd party development easier. Instead of 68k AmigaOS, PPC AmigaOS, MorphOS and AROS, there would have been only AmigaOS, MorphOS and AROS with AmigaOS being the largest user base for developers to target.


You wrote a BIG TRUE, it was a big mistake but they can still make a patch.
I think an OS4 porting for Vampire could be really a great thing and they could make a little of money with it.

I think also a correct commercial strategy today is to sell hardware at "cost price" and make money selling software.
A-Eon and others too always had a bad strategy to gain market share and involve new users.
I'll never spent more than 300€ to get a "nothing useful" hardware, today I use Amiga just for fun and I can go with emulators or used cheap hardware if they don't offer me something I can buy.
But when I (and surely others too) will bought a new hardware at right price, I'm sure I'll look forward to buy useful software for it and sellers will make money on my shopping.

In few words you have to invent nothig, just look at well consolidated strategies of sony, microsoft and nintendo for console world or all developer machines like RPi for pc embedded world.

Last edited by Fl@sh on 29-Jan-2019 at 07:28 PM.

_________________
Pegasos II G4@1GHz 2GB Radeon 9250 256MB
AmigaOS4.1 fe - MorphOS - Debian 9 Jessie

Quote:

cdimauro wrote: Currently I decided to don't vote, because some clarification is required IMO. First, what we intend for "AmigaOS"? The (original) Amiga o.s.? AmigaOS4? Some re-implementations (MorphOS, AROS)? The aboves, but running on an hosted environment? Or an emulated one? Or a "virtualized" one (e.g. transparently running "AmigaOS" applications on another o.s.)? Etc.

AmigaOS is based on code from the original C= AmigaOS. MorphOS and AROS are re-implementations of the API but usually not referred to as AmigaOS.

Quote:

Second, which kind of "future" do we expect? Running legacy 68K software faster / with more features (e.g.: improved RTG/RTA. Something else)? Enhancing 68K software? Enhancing the o.s. (e.g.: new APIs), maybe to put it on par with other modern o.ses? Rewriting 68K software and or/o.s.? Making it easier to port software from other platform? Etc. The same questions apply to AmigaOS4, MorphOS, AROS, with 68K eventually changed to the wanted CPU architecture.

Good questions. Different answers are possible for different architecture choices yet many people already have an idea of what architecture they would like for the AmigaOS. Amiga users and developers have historically been very opinionated about the architecture choice so I was curious what they like today.

Quote:

Third, what benefits should the CPU architecture bring to the "AmigaOS"?

Amiga users are more technically literate than average and I believe the results of the poll are answering the question. The general view is that the architectures have the following "benefits".

68k - Amiga compatibility, code density, embedded market, retro market
ARM - cheap price, common, low energy use, code density, embedded market
POWER - Amiga compatibility, server energy efficiency, server market
PowerPC - Amiga compatibility, embedded market
RISC-V - open hardware, customizable, code density, embedded market
x86_64 - performance, performance/price, common, PC and laptop market, server market

While many of these "benefits" are true, there are misconceptions, some of which I have tried to dispel already. Many people choose based on performance, price or overall viability. Few, I believe, look at the potential and strengths of the AmigaOS and architectures we use as the grass is always greener on the other side of the fence. Few grasp the importance of compatibility and standardized hardware. Also, the easiest choice is not necessarily the best choice. The route of easy choices leads to entrenched competition.

Quote:

Amigo1 wrote: I will not vote because my vote doesn't change anything in this Amiga-democracy. Because my knowledge of the market is not good enough and it would not be wise to pick a choice that will potentially (but then again very unlikely) influence a decision of the higher ups and lead to a situation which leads to a cul-de-sac.

Fair enough, even though the Amiga is far from a democracy. Mob rule isn't always best but mob appeal and opinion is still important.

Quote:

Anyway, I like the idea of a Power Amiga, more than Risc-V or ARM. But >68x80 would be nice too, any chance it will leave the Virtual-FPGA-reality and become real?

POWER Amiga is one of the easier options but, by itself, limited to high end high cost hardware. A much lower priced option is needed to expand the Amiga user base.

I believe a 68k ASIC is within reach financially for the Amiga. Old underutilized fab die sizes are cheap, embedded partners could help share the cost and a few million dollars isn't difficult to raise anyway. It's funny that Mr. "Pie in the Sky" Gunnar didn't believe an ASIC was possible after I told him how it could be done. Unfortunately, he continued to uber optimize both the core and ISA for performance in an FPGA. I think a 500MHz-1000MHz 68k ASIC SoC which could allow Amiga hardware for about double the price of a Raspberry Pi would sell well. Retro hardware remakes have been setting records and embedded markets are increasing in size exponentially but Amiga businesses keep the blinders on.

Last edited by matthey on 29-Jan-2019 at 08:47 PM.

Quote:

pavlor wrote: x86_64. What else for desktop/laptop today?

The x86_64 architecture is in desktops and laptops today primarily for 2 reasons, single core performance and economies of scale which allowed them to outperform the competition by moving to smaller die sizes quicker. There was an architecture which had better single core performance than x86 through 4 generations of CPUs called 68k (the 68k outperformed the x86 by a larger gap in energy efficiency). Moore's Law is coming to an end and driving up the cost of die shrinks which reduces the cost advantage they give. Intel recently took a hit because of problems moving to a smaller die size. The ISA is ugly, has grown fat and is full of security holes. Still think it is a good idea to move to x86_64?

@matthey

didnt vote anything, frankly i don't care as long amigaos dosen't die i'm good with whatever

_________________
nox's in the house!

@matthey
Power PC of course.
The x86-64 have arrived to end, only add cores and cotes to the cpus.
The arm well is like PPC, but with more good luck.

@matthey

If you’d given me multiple choice, I’d go x86, 68k, PPC and ARM. Since I can’t, I have to go x86.

Might not be the 'nicest', but it’s everywhere, and you can make your own power consumption/performance trade offs.

_________________
"Knowledge is power. Power corrupts. Study hard, be evil."

I want a laptop. I just don't get to go spend time in a computer cave where a desktop or power is tethered to a desk, so yet another ATX or MiniITX board is not very interesting to me at all, regardless of architecture. I just don't get to use those things anymore.
So only one reasonable choice at all is x86/amd64.
I'd love to say Arm, but I don't want an ipad or a phone. I use both ipads and smartphones, but I cannot be productive on them, they are really only for entertainment or email, but not reasonable for productivity apps.
Nor do I want a Chromebook. Give me performance.

Last edited by billt on 30-Jan-2019 at 03:12 AM.

_________________
All glory to the Hypnotoad!

@matthey

Quote:

...What did you have in mind?

The 68k extensions you and I discussed.

_________________
All the way, with 68k

@matthey

Quote:

matthey wrote: Fair enough, even though the Amiga is far from a democracy. Mob rule isn't always best but mob appeal and opinion is still important.

to quote Daniel Muessener: "I really can't say anything bad about that!"
Quote:

matthey wrote: Quote: Amigo1 wrote: Anyway, I like the idea of a Power Amiga, more than Risc-V or ARM. But >68x80 would be nice too, any chance it will leave the Virtual-FPGA-reality and become real? POWER Amiga is one of the easier options but, by itself, limited to high end high cost hardware. A much lower priced option is needed to expand the Amiga user base.

The way I understand it, PowerPC will not be developed further and is a "dead" platform. This leaves only the other options. So what you wrote sounds plausible but...
Quote:

I believe a 68k ASIC is within reach financially for the Amiga. Old underutilized fab die sizes are cheap, embedded partners could help share the cost and a few million dollars isn't difficult to raise anyway. It's funny that Mr. "Pie in the Sky" Gunnar didn't believe an ASIC was possible after I told him how it could be done. Unfortunately, he continued to uber optimize both the core and ISA for performance in an FPGA. I think a 500MHz-1000MHz 68k ASIC SoC which could allow Amiga hardware for about double the price of a Raspberry Pi would sell well. Retro hardware remakes have been setting records and embedded markets are increasing in size exponentially but Amiga businesses keep the blinders on.

...if the 68k ASIC way is chosen, will it still be cheaper than POWER? Can the work of Gunnar be "recycled" for the design of an 68K ASIC?

And what I'm not so really grasping: will this 68k ASIC be designed to include most of the classic Amiga hardware or just the 68k part?
If the latter, and a 68K ASIC "68080" or higher is produced, how would it differ to a "real" 68080.
I mean eventually the ASIC68k80 will be developed to an 68k90 or whatever, following the (at the moment) usual CPU evolution.

@matthey

Quote:

Also, out of development PPC hardware is quickly falling behind so no longer NG while having little retro appeal.

What do you mean by "out of development?" FreeScale/NXP still manufacture the PPC chips. The PPC chips for the X5000 and A1222 are indeed NG. NG meaning after 68k. If you mean Post-NG (beyond current NG systems), then that's another matter.

Quote:

PPC CPUs will be available for awhile but prices of CPUs are more likely to increase than decrease over time.

Why would PPC CPUs go away anytime soon? They're still used in automotive, industrial, and AI systems. I don't think the AmigaOne line will stay with a particular PPC chip beyond the CPU's EOL cycle (e.g. for the e500), so such a cost increase shouldn't matter. As a side note, the Nintendo WII U uses the Freescape PPC processor, so the PPC isn't exactly dead.

Quote:

Hyperion will likely lose money on every copy of AmigaOS 4 sold because the PPC market is not large enough to pay for their development costs.

This is part of investment risk. And it is why growing the user base is important for future returns.
Quote:

Absolutely true. This is one of the reasons why I would like to see a return to the 68k. AmigaOS 4 was working on the 68k until Hyperion made the arrogant mistake of discontinuing it. If AmigaOS 3.1.4 and Trevor bailout rumors are true, this mistake should have resulted in their bankruptcy.

Yes, that would be nice for retro Amiga, but 68k is dead and slow compared to PPC. 68k is definitely not NG. I'd pick ARM over 68k because at least ARM has a future. And Trevor did recently mention releasing an Enhancer Software Package for Amiga Classic, so that will be good for Classic users.

Quote:

AmigaOS 4 was working on the 68k until Hyperion made the arrogant mistake of discontinuing it.

I'm okay with AOS4 on 68k, but (1) can 68k handle it and be responsive, and (2) what future is there in 68k? The only benefit would be using UAE to emulate it on other platforms. Hyperion probably discontinued development on the 68k due to lack of resources and focused on PPC instead. Back then, at least PPC had a future while 68k was dead.

Quote:

The AmigaOS PPC user base is too small to be profitable for Hyperion and Tabor is not likely to change that. If Tabor does sell well enough to ramp up production, it would be good to upgrade the CPU to use the more compatible e500mc core but the damage will have already been done with the less compatible e500v2 core of lowering the standard and making purchasers of the 1st run of Tabors unhappy.

Indeed the AmigaOS PPC user base is small. It is growing slowly. I don't think Tabor users will be unhappy. The Tabor is faster than previous AmigaOne's except the X5000 and maybe the X1000, so most Amiga users should find it a good entry-level NG system. It should handle retro gaming adequately, offer web surfing, and document creation (once LibreOffice is available).

Quote:

Maintaining a standard API between PPC and 68k AmigaOS would have prevented the split between 2 factions and made 3rd party development easier. Instead of 68k AmigaOS, PPC AmigaOS, MorphOS and AROS, there would have been only AmigaOS, MorphOS and AROS with AmigaOS being the largest user base for developers to target.

Yes, indeed. There is good compatibility between AmigaOS 3, Aros, MorphOS. The challenge with updating OS versions is maintaining backward API compatibility while introducing new classes, methods, and other features. I'm still learning the AmigaOS 4 API, so I'm not too familiar with how it differs from 3.1-3.9, but I do see OS 4 API has many things in common with OS 3.1. Some Amiga Company should offer a "AmigaOS Compatibility Certification" feature (like Microsoft's) for software and other Amiga-like systems; this would help, I think.

Quote:

SMP, memory protection and security are extremely important for the server market. It would be sad to sell Amiga branded servers running Linux/BSD (Rebranded Raptor servers?).

Yes, SMP and such are important. Linux could run on X5000 as a server while AmigaOS (and others) could run on clients accessing the X5000. Although it could be considered as sad, yet it can provide additional sales for the X5000 PPC platform. IBM does it. They run AIX and Linux on their own hardware. The X5000 can do the same thing.

Quote:

New Amiga slogan: "We started the multimedia revolution but now we are here to make a cheap buck".

That's funny. Maybe also fitting: "We started the multimeide revolution, but now we're trying to catch back up."

@redfox

Quote:

AmigaOS4 is locked into PPC hardware. As long as inventory exits, this will not change. They are supporting X5000, X1000, SAM460, SAM440, AmigaOne XE, MicroA1, Pegasos II, and classic Amiga systems with compatible PPC accelerator cards. They are working on Tabor. 68K and PPC are the most likely candidates for present and future projects.

Yes. PPC was the natural path for AmigaOS 4 because the PPC had been introduced to AmigaOS 3 as accelerator cards. With X5000 and the upcoming A1222 using PPC, PPC is on the roadmap for the next 3-5 years.

@Fl@sh

Quote:

I think an OS4 porting for Vampire could be really a great thing and they could make a little of money with it.

Yes, and it should be considered. I would buy an A1222 since it would be faster solution for AmigaOS 4.


Anyhow, those are my thoughts for now.

Obviously Power would be the easiest and more logical for the higher end, specially now with lower priced systems available.
Trevor probably didn't anticipate this, otherwise the money spent on the X5000 could have been used to port AmigaOS to a Raptor like Power system.

Funny, I asked a question about this in one of the Amiwest Q&A about two years ago and I was practically joked about (there were systems anounced already by the time).

Last edited by Jose on 30-Jan-2019 at 01:34 PM.

_________________

José

@Topic

If I would not be restricted to reality, then either as powerful 68K as todays computers are, or completely new processor that is backwards compatible. Therefore, the other vote.

@matthey

Quote:

matthey wrote: Quote: pavlor wrote: x86_64. What else for desktop/laptop today? The x86_64 architecture is in desktops and laptops today primarily for 2 reasons, single core performance and economies of scale which allowed them to outperform the competition by moving to smaller die sizes quicker. There was an architecture which had better single core performance than x86 through 4 generations of CPUs called 68k (the 68k outperformed the x86 by a larger gap in energy efficiency). Moore's Law is coming to an end and driving up the cost of die shrinks which reduces the cost advantage they give. Intel recently took a hit because of problems moving to a smaller die size. The ISA is ugly, has grown fat and is full of security holes. Still think it is a good idea to move to x86_64?

Yes because the history isn't relevant now. The full of security holes thing I don't know anything about, what architectural features are you referring to?

In the future ARM or RISC-V will probably be the best choice but for now x86 is obviously better, price performance is very good and availability is superb. Running big-endian code isn't a problem since several years (just reduces fusion opportunities) with the MOVBE instruction.

The problem isn't choosing a processor architecture, it's that Amiga OS and clones are using a dead-end design and have to be rewritten from scratch. That was obvious over 20 years ago...

I voted x86 due to abundance but I'm not particularly arsed.

_________________
Robert
--
A1XE G4, OS4.1. Peg1 G3, MOS 1.4.
Abel Soul - Check out our tunes on Spotify

Quote:

Templario wrote: Power PC of course. The x86-64 have arrived to end, only add cores and cotes to the cpus.

Most recent PowerPC processors had been tweaks of G3/G4 designs with more cores, more caches and die shrinks as well. It is PPC which has arrived at an end with no new off the shelf designs. There may be a few custom designs in the future but they may not be for sale to everyone.

The x86_64 designs are approaching their limits but there is nothing to compete with them in single core performance. When Intel had trouble moving to a smaller die size in their fabs, ARM temporarily closed the performance gap using a smaller die size but it was AMD which saw the biggest boost in the high end CPU market.

Quote:

The arm well is like PPC, but with more good luck.

AArch64 is much like PPC on the surface (32 gp registers and fixed size 32 bit instructions) but it is not just luck or ARM market leverage which puts it ahead. The AArch64 ISA is more powerful and modern than the PowerPC ISA. AArch64 code uses fewer instructions, has better code density and has fewer branches. These advantages are often significant and boost single core performance. AArch64 has more powerful addressing modes which act as a performance multiplier to increase single core performance. AArch64 has a more powerful SIMD unit and it is standard. The robust (fat) ISA does require more logic so trades cores for better single core performance (good trade-off in most cases). Just as PPC came in and improved on MIPS, SPARC and Alpha ISAs, AArch64 and RISC-V have improved on PPC (often by using CISC traits like more powerful instructions and addressing modes for AArch64 and variable length instructions for RISC-V compressed code density). Now PPC joins the bone yard full of uncompetitive and too riscy of RISC ISAs.

@matthey

could not vote for:

68k because, it is outdated and underpowered
ARM because, it is restricted to obscure hardware
POWER because, there are no cheap PC like boards available
PowerPC because, there are no cheap PC like boards available
RISC-V because, it is restricted to obscure hardware

There are not much alternatives left, x386 would be a waste of time, so best solution might be AMD_64 (x86_64).

Quote:

Amigo1 wrote: The way I understand it, PowerPC will not be developed further and is a "dead" platform. This leaves only the other options. So what you wrote sounds plausible but if the 68k ASIC way is chosen, will it still be cheaper than POWER?

Yes, but comparing POWER to a first 68k ASIC would be like comparing apples and oranges. POWER CPUs are high end server CPUs costing $375 U.S. for the cheapest 4 core POWER9 CPU from Raptor. A 68k ASIC would initially be a low end CPU. There are substantial up front costs but high quantity production could lower the per unit price to a fraction of the cost of off the shelf CPUs from NXP. The FleaFPGA board was available for $45 U.S (on Indiegogo) and an ASIC could replace the more expensive FPGA while providing probably around 100 times the performance.

https://fleasystems.com/fleaFPGA_Ohm.html

Of course more and faster memory and Gigabit ethernet would be wanted at the very least and which is already in his redesigned board.

http://www.fleasystems.com/forums/showthread.php?tid=67

AGA+RTG and maybe SATA would be nice as well. In any case, it should be possible to create a cheap Raspberry Pi like Amiga that is more powerful than any existing 68k hardware for cheap enough to buy as Christmas stocking stuffers.

Quote:

Can the work of Gunnar be "recycled" for the design of an 68K ASIC?

Some of the optimization Gunnar does is simplifying the logic which is good for an ASIC while other optimizations are to use features available only in an FPGA (bad) and join CPU units (very bad). It is likely that major parts of the current design could be used for an ASIC if it is in good enough shape. Gunnar is a sloppy but fast coder while Jens was the original Natami CPU designer and he is very neat and organized. It is difficult to say what shape the code is in. I tried to have a professional CPU designer take a a look at the code but Gunnar was too suspicious. Jens has been more open but was already less active at the time I left.

Dealing with Gunnar is something else too. His antics were already a problem in the Natami project even as he made valuable contributions to the CPU core. He can be overbearing and unprofessional at times, to put it nicely.

Quote:

And what I'm not so really grasping: will this 68k ASIC be designed to include most of the classic Amiga hardware or just the 68k part?

Eventually, the 68k ASIC could become an SoC including the Amiga custom chips and more but not necessarily at first. How much to include initially in an ASIC would likely be up to the team of developers creating it (technical decisions) and the people paying to have it created (marketing decisions). A smaller FPGA could be included on the board for custom chip sets and embedded customization.

Quote:

If the latter, and a 68K ASIC "68080" or higher is produced, how would it differ to a "real" 68080. I mean eventually the ASIC68k80 will be developed to an 68k90 or whatever, following the (at the moment) usual CPU evolution.

In the case of an SoC with custom chips, the custom chips could be turned off (power gated or permanently cut the power to the logic) for customers wanting just a 68k CPU. Transistors are cheap now days and the Amiga custom chip logic is small and efficient because it was created at a time when every transistor counted and few were wasted.

Quote:

davidf215 wrote: What do you mean by "out of development?" FreeScale/NXP still manufacture the PPC chips. The PPC chips for the X5000 and A1222 are indeed NG. NG meaning after 68k. If you mean Post-NG (beyond current NG systems), then that's another matter.

NXP announced that there would be no new PPC designs and they were the last developer of off the shelf PPC CPUs. Some current PPC CPUs may still be in production but they will likely never move to new die sizes so they will fall further and further behind in performance (most of these are energy efficiency embedded designs rather than performance designs for PC use as used in the Amiga).

Quote:

Why would PPC CPUs go away anytime soon? They're still used in automotive, industrial, and AI systems. I don't think the AmigaOne line will stay with a particular PPC chip beyond the CPU's EOL cycle (e.g. for the e500), so such a cost increase shouldn't matter. As a side note, the Nintendo WII U uses the Freescape PPC processor, so the PPC isn't exactly dead.

Go to NXPs web site and see what they are marketing. ARM replaced PPC. I know the feeling as Motorola replaced the 68k with the PPC when the 68k was still the best selling 32 bit embedded CPU in the world.

Quote:

This is part of investment risk. And it is why growing the user base is important for future returns.

Losing money on a product for years with few prospects or a plan for change is no longer an investment risk but financial suicide.

Quote:

Yes, that would be nice for retro Amiga, but 68k is dead and slow compared to PPC. 68k is definitely not NG. I'd pick ARM over 68k because at least ARM has a future. And Trevor did recently mention releasing an Enhancer Software Package for Amiga Classic, so that will be good for Classic users.

The 68060 outperformed most of the early PPC CPUs clock for clock on the same sized die (601 was close but used more transistors, had more caches and ran hotter, 603 was trash, 603e was still slower per clock even with a die shrink). Each die shrink gives something like 50%-100% performance boost if higher achievable clock speeds are counted so please compare on the same sized die. The 68k has very good performance traits, better than PPC and often better than AArch64, RISC-V compressed and even x86_64. It has excellent single core performance, few instructions executed, excellent code density and low energy use per task (running cooler is better for performance). PPC was always a performance disappointment especially in being able to clock it up and single core performance. Even POWER can not compete with x86_64 in single core performance.

Quote:

I'm okay with AOS4 on 68k, but (1) can 68k handle it and be responsive, and (2) what future is there in 68k?

The beauty of the AmigaOS is its modularity and scalability. The AmigaOS is more marketable when it is small for low powered CPUs than it is big for high end CPUs. IMO, it was pointless to go big without SMP support, better memory protection and better security. There aren't very many OSs that can go as small as the AmigaOS while still being as functional. This is valuable for certain markets like embedded and the smallest and lowest cost hardware where lower performance hardware can be used. The 68k was part of the recipe which allowed a small 68k AmigaOS foot print and there is room to go smaller. Modern electronic devices keep going smaller while OS resource requirements keep going up but this is unsustainable with Moore's Law ending. Opportunity?

Quote:

Indeed the AmigaOS PPC user base is small. It is growing slowly. I don't think Tabor users will be unhappy. The Tabor is faster than previous AmigaOne's except the X5000 and maybe the X1000, so most Amiga users should find it a good entry-level NG system. It should handle retro gaming adequately, offer web surfing, and document creation (once LibreOffice is available).

Tabor owners who bought production run #1 boards using the e500v2 core would be "unhappy" if production run #2 boards used the more compatible e500mc core. Tabor is still not so cheap that the old board can be discarded when the new one comes out. It is kind of ironic to call it NG and then say "it should handle retro gaming adequately" (most retro Amiga games won't run on it either).

Quote:

megol wrote: Yes because the history isn't relevant now.

Those who fail to learn from history are doomed to repeat it.

Quote:

The full of security holes thing I don't know anything about, what architectural features are you referring to?

x86_64 gives too much control in user space. Is it just a coincidence that Spectre and Rowhammer attacks were first documented using the CLFLUSH instruction?

Quote:

In the future ARM or RISC-V will probably be the best choice but for now x86 is obviously better, price performance is very good and availability is superb. Running big-endian code isn't a problem since several years (just reduces fusion opportunities) with the MOVBE instruction.

The best single core performance architecture gets the games (profitable PC/laptop market) and stays king. Do you think AArch64 or RISC-V single core performance will eventually match x86_64? How?

Quote:

The problem isn't choosing a processor architecture, it's that Amiga OS and clones are using a dead-end design and have to be rewritten from scratch. That was obvious over 20 years ago...

Maintaining full compatibility while adding SMP, memory protection and security is a problem. I don't think Hyperion has been honest while stringing AmigaOS 4 users along for years. I disagree that AmigaOS needs to be rewritten completely from scratch though. These issues can be addressed with custom hardware where off the shelf commodity hardware would be more of a challenge.