Just thought I would begin a new thread to discuss what the pros and cons of each architecture for possible ports of a future 2nd generation MorphOS.

I know that the team has announced, or given hints that the choice has already been made to use x64 architecture, and I don't want to create a divide in the community, but the more I think about a future version of MorphOS, and how it could grow and succeed, the more I question the choice, and wonder if ARM might not be better choice for many reasons.

I'll suggest just a very few Pro's to support why I feel this way, and let more knowledgeable people add their thoughts and opinions, to support my thinking, or to explain the Con's, and why they outweigh the Pro's for porting to ARM.

Here are a few obvious Pro's

1. Low power consumption, making the creation of laptops and tablets easier.
2. Low cost and wide spread use around the world
3. Lack of any single OS currently dominating the market on ARM devices, gives MorphOS greater opportunity to appeal to new users, if/when we grow our developer base and port or create enough MorphOS for ARM applications/programs.
4. Powerful enough to run almost all currently available programs Natively ported to MorphOS, and if the next generation of MorphOS is going to include multi-core support, something as simple and cheap as the Raspberry Pi 3B should be able to match or surpass any of our current single core results running on a G5 Mac (I'm guessing on this, but think I have seen such statements made by others on this forum before).
5. Some members here have mentioned that ARM (or some) CPU's can run big endian and small endian code (if I understood what they were writing), so maybe that is some advantage when porting some of our existing MorphOS PPC programs to the new MorphOS NG?

Some obvious Con's

1. The MorphOS Dev. Team are mostly, or all "Power Users", who create MorphOS for themselves in their free time, so naturally, they want to use a powerful, fast system, which will compile their code in minutes, not hours, (or days, like the original Amiga 68000@7mhz).
2. The lack of currently available full laptop and desktop systems using high performance ARM CPU's, to easily port MorphOS NG to. As we know from past experience, the Team does not port to non-existent products/platforms.

I'm sure other members here will point out many other Con's, so I will stop with just the two obvious ones I know of above.

If you think this discussion is just a waste of time, then ignore this thread. There might be some members who have interesting ideas to add here, and I'm interested in reading them, even if it has no effect, or change of outcome in our direction.

> lack of currently available full [...] desktop systems using high performance ARM CPU's

https://www.avantek.co.uk/avantek-thunderx2-arm-workstation-thunderx2station/
https://store.avantek.co.uk/avantek-thunderx2-arm-workstation-thunderx2station.html

;-)

Oh, WOW, they are cheap . . . . .

I'm getting one of those . . . . never ! . . . . I have no interest in specialised hardware like that, if that's any kind of example of what its exhorbitant price would be.

The biggest restrictions in MorphOS are not the hardware. It is the design of any Amiga-like OS and the necessary legacy compatibility.

A shared memory address space means no way to move to 64 bit memory range to beat that 2GB barrier, even with bank switching. No way to use multiple cores. No way to use virtual memory properly. No way to use "memory protection".

The ideal CPU target, then, would have a single core but a high clock rate. They don't really make those any more.

> No way to use multiple cores.

...except in an ASMP type of use.

The advantages of ARM that you list are advantages for smartphone manufacturers. The morphOS team aren't going to build hardware.

Frankly the obsession with this or that CPU architecture is a waste of time. They're all fast enough. The correct choice is to port to the one that's standard and easily available in normal computers, not one that could hypothetically be good if it was used in them instead.

You can build a low volume ARM desktop motherboard if you want but it'll be just as overpriced, buggy and unsupported as any of the PPC attempts of the past 20 years.

> as [...] buggy [...] as any of the PPC attempts of the past 20 years

That's not the impression I got from the Apple, bplan/Genesi, ACube and Varisys/A-Eon PPC hardware.

Quote:

If you took the time to read my original post, you would realize that this is a discussion about CPU architecture choice for the future MorphOS NG, or MorphOS4.x, or what ever you want to call it, which will have SMP and Memory Protection, not our current single CPU/Core MorphOS3.11 without memory protection.

Quote:

Their other desktop ARM offering is about the same, or less than an X5000, and provides 32 cores, not 2 or 4 cores, that the X5000 provides. Remember, this discussion is for the future of MorphOS, after it has SMP and Memory Protection, so it can use all those cores.

Here is the link to the ARM desktop:

https://store.avantek.co.uk/avantek-32-core-cavium-thunderx-arm-desktop.html

The specs:

Super-Quiet ATX Tower Chassis
Cavium® ThunderX™ CN8890_CP 1.8GHz 32 core ARM SoC
(option 2.0GHz 48 core ARM Soc)
Full-height Full-length slot with PCIe x16 (Gen3 x8 bus).
Full-height Full-length slot with PCIe x8 (Gen3 x8 bus)
8 x DDR4 ECC DIMM slots
1 x 40GbE QSFP+ LAN port
4 x 10GbE SFP+ LAN ports
3 x 3.5” (5 optional), 2x 2.5" HDD/SSD bays, motherboard supports MAX 4x SATA.
Super-Quiet fans
500W PSU

The price: = £1,360.00, or $1.774.86, or 1,565.77 Euro

sure, you still need to purchase RAM and a video card, but I still think it compares very well to the X5000, if the Next Gen MorphOS (w/SMP support & 64bit, & Memory Protection) was ported to the Avantek Cavium-ThunderX ARM desktop.



[ Edited by amigadave 25.02.2019 - 11:40 ]

My point in suggesting ARM as something to consider and discuss, is that it can be scaled from something as small and cheap as the Raspberry Pi, to the monster servers and workstations that Andreas Wolf linked to, and NewSense complained about the price. X64 cannot be scaled down to something that can be sold for only $35, but still provide enough speed and power to run MorphOS.

I am sure that more companies will be building ARM desktop and laptop computers in the near future, and that in the not too distant future, we may even see ARM challenging Intel and AMD, as a true competitor in the consumer computer marketplace. Look at how popular the Maker community has become, and they are very fond of small ARM boards to control their projects. It would not be hard to convince many of those people to switch to an ARM based computer, either desktop, laptop, or tablet. Personally, I dislike tablet computers, so I hope that the laptop form factor will continue to be in demand enough to warrant the interest of several different manufacturers.

[ Edited by amigadave 25.02.2019 - 11:16 ]

Quote:

Exacly my mantra since years. x64 is still leading and will stay significantly in business for the next decades and hence it is the most logical choice. It'a a conservative approach indeed, but for MorphOS with its limited resources and user base a conservative approach may be the safest option.

Quote:
Quote:
So, this ARM Desktop System being ONLY £1,360 ... is ONLY £1,200 too MUCH, or probably about 80% percent of these amounts $1.774.86, or 1,565.77 Euro too much, IMHO !

Comparing it to the X5000 is all very well, but that hardware is too expensive as it is, for our "hobby" Amiga systems.

If we had a full office package that was compatible with other systems, and integrated graphics software, along with other more advanced features we still don't have then .... maybe .... it might be worth paying more for a system. However, as it stands, for MorphOS the Apple hardware does all that we need, apart from some limitations browsing, and slow-downs as a result of the PPC hardware getting over-worked with JavaScript and the like, but then the hardware is only costing £60 on average, plus the cost of the licence, which I feel is good value for money, and well worth the commitment to remain on our OS. I'd like to be able to donate more, but I'm regularly and all-too-soon out of spare cash for donating, but when I can I do.

> Their other desktop ARM offering [...]. Here is the link to the ARM desktop:
> https://store.avantek.co.uk/avantek-32-core-cavium-thunderx-arm-desktop.html

There's even a third offering in the desktop/workstation category, which lies between those based on ThunderX and ThunderX2, with regard to both price and performance:

https://www.avantek.co.uk/ampere-emag-arm-workstation/
https://store.avantek.co.uk/ampere-emag-64bit-arm-workstation.html

I think Apple's attempt to move their desktops and notebooks to ARM will fail, and I wouldn't want to be on ARM in case that happens. Not that it will affect the family on mobile (no chance of that), but the higher end ARM designs could end up PowerPC'd.

It's a feeling more than anything else. For one, Apple's move is very ambitious - probably too ambitious. I've seen a lot of danger signs too about Apple hardware over the last few years, how they now cut corners on components and build materials to shave money off, where once Apple was renowned for build quality. This, I think, has been responsible for their slow decline, that and a distinct lack of investment in MacOS which has now left them some years behind.

TL;DR: x64 still seems the brightest prospect. Stay away from anything Apple does, they have the smell of slow death on them.

[ Edited by KennyR 26.02.2019 - 14:36 ]

Quote:

All of that stuff is flawed. The apple stuff is the best working but a lot of models have serious problems, ibooks, certain g4 powerbooks fall to bits, powermac G4s are loud and blow up their PSUs.

Then you look at the MOS and OS4 bespoke hardware. Pegasos I was ruined by a bad chipset, the Efika has no ram. At least Peg II was comparable to a Powermac at the time it was made.

On the OS4 side we get the original AmigaOne motherboards that went out of production before they actually saw an OS release, the sam440 with it's soldered down ram, and the x1000 which shipped with redundant sound and Ethernet cards because they didn't bother making drivers for their own bespoke motherboard. After a decade their lauded "xorro" chip still doesn't do anything. I won't even mention the prices on those boondoggles!

All in all just a mess really.

>>> as [...] buggy [...] as any of the PPC attempts of the past 20 years

>> That's not the impression I got from the Apple, bplan/Genesi,
>> ACube and Varisys/A-Eon PPC hardware.

> All of that stuff is flawed.

Your claim that I quoted and thus referred to was about hardware bugs in particular.

> a lot of models have serious problems, ibooks, certain g4 powerbooks
> fall to bits, powermac G4s are loud and blow up their PSUs.

After how many years of usage? How does x86 hardware compare after same time of usage?

> Pegasos I was ruined by a bad chipset

The chipset bugs were remedied by the April2 chip, which, while slightly lowering performance, means that the bugs were not exposed.

> the Efika has no ram.

The low amount of RAM on the Efika 5200B is not a bug.

> the original AmigaOne motherboards

...is not part of "that stuff" that I mentioned, for a reason.

> the sam440 with it's soldered down ram

Not a bug either.

> the x1000 which shipped with redundant sound and Ethernet cards because
> they didn't bother making drivers for their own bespoke motherboard.
> After a decade their lauded "xorro" chip still doesn't do anything.

Again, no hardware bugs.

> I won't even mention the prices on those boondoggles!

Prices are not bugs.

> All in all just a mess really.

And virtually nothing of this mess has been caused by hardware bugs, at least when applying the generally accepted meaning of the term.

Always in these threads you forget in the MorphOS team because they develop the os for the architecture and they will have a 68k emulation plus a PPC to can runs the current software available.

Quote:

Wrong on two counts. One, April2 did not fix all the bugs - my Peg1 still had DMA hangs from time to time. Two, it wasn't slightly lowering performance - the April chip bottlenecked the whole system and made RAM speeds ridiculously poor, and AGP not even worth having.

Quote:

But also it made it incredibly unsuitable for desktop. The 5200B was aimed more at fixed kiosks.

Quote:

They were untested CPUs, ballasted by untested chipsets, mounted on untested board logic. Almost nobody else in the world used them. Anyone who tried would learn the bugs the hard way. Of course part of the mess was caused by hw bugs.

Why won't you ever accept that PPC is dead because of these problems? You continue to argue forever while the world moved on 20 years ago.

[ Edited by KennyR 26.02.2019 - 14:43 ]

@Andreas
In Kenny's point of view "bad design" is also some kind of bug

Well.. all of these 'porting MorphOS to another CPU architecture' discussions are nice, but a bit pointless - unless you are a Dev with MorphOS source code access.

If you want to run some AmigaOS on other archs you can already do this with AROS. Here you can already help/improve existing ARM or amd64 support. Which in turn could even be a benefit for MorphOS in the future if some Dev wants to have a look at already working AROS ARM code for instance.

Quote:

Are you proud of what you wrote here? It's like you have some sort of internet-caused brain disease resulting in pedantry so extreme you miss the weight not only of what others say, but what you say also.

You've gone down my post with a big checklist going "technically not a bug! technically not a bug! I win I win!"

you're right in that a soldered down, useless low amount of ram is not a "bug" itself... so my corrected statement: These computers are overpriced, buggy, unsupported, and also: badly designed.

> April2 did not fix all the bugs - my Peg1 still had DMA hangs from time to time.

Interesting. My April2-fixed Peg1 did not. Where can I read more about those remaining bugs, preferably from people who know what they're talking about (like bplan engineers, Linux developers, MorphOS developers)?

> it wasn't slightly lowering performance - the April chip bottlenecked the whole system

April2 lowered the FSB from 133 MHz to 100 MHz. Of course, it's debatable whether a 25% bus speed decrease is a slight decrease or a significant decrease and I can easily accept any opinion favouring the latter view.

> and made RAM speeds ridiculously poor, and AGP not even worth having.

You mean compared to the Eyetech AmigaOnes which had the FSB running at full 133 MHz? I mean, wasn't even the Peg2's fake AGP x1 faster than the Eyetech AmigaOnes' "real" AGP x2 that was not hampered by April2 in speed?

>> The low amount of RAM on the Efika 5200B is not a bug.

> it made it incredibly unsuitable for desktop. The 5200B was aimed
> more at fixed kiosks.

Absolutely, and thanks for confirming that it is not a bug.

>> virtually nothing of this mess has been caused by hardware bugs

> They were untested CPUs

No, they were not. Not one single type of CPU in any Apple, bplan/Genesi, ACube or Varisys/A-Eon PPC hardware was untested. And I'm confident enough to even extend this list to all the CPU types used in MAI/Eyetech hardware. Most of these CPU types are/were used in high-reliability systems in various fields of application for many years, which was for a reason.

> ballasted by untested chipsets

...in case of the Peg1. Can you name the other "untested chipsets" used in Apple, bplan/Genesi, ACube or Varisys/A-Eon PPC hardware?

> mounted on untested board logic.

Untested Apple boards? You must be joking. The lists of bugs of "untested" Apple, bplan/Genesi, ACube or Varisys/A-Eon PPC boards must be huge compared to the list of bugs of tested x86 boards. Where can I find these lists?

> Almost nobody else in the world used them.

Not true. For instance, Apple purchased less than 5% of the PPC CPUs produced by IBM, so the other customers together used about 20 times as many IBM PPC CPUs as Apple did. That's certainly not "almost nobody". And then there's Motorola/Freescale who also sold the major quantities of their "desktop class" PPC CPUs to customers other than Apple. These non-Apple customers of IBM and Motorola/Freescale also used the PPC-specific, non-Apple support chipsets. The Apple chipsets and boards were used by all Apple users of the PPC era, which is a significant number. And most chipsets on bplan/Genesi, ACube and Varisys/A-Eon boards have been deployed on x86 boards as well.

> Anyone who tried would learn the bugs the hard way.

That's true for all hardware developments exceeding a certain complexity and involving complex circuits, regardless of ISA.

> Of course part of the mess was caused by hw bugs.

A part that amounts to virtually nothing compared to the parts of the mess caused by other things, and a part that's not significantly bigger than the counterpart in the x86 world.

> Why won't you ever accept that PPC is dead because of these problems?

Because bugs is simply is not why PPC is dead. It's dead because of various reasons, but none of them is bugs in CPUs, chipsets or boards.

> You continue to argue forever while the world moved on 20 years ago.

That's because you continue your attempts at rewriting history. PPC may be dead, but facts shouldn't be changed after the fact (if you pardon the pun), no matter the time passed.

> Are you proud of what you wrote here?

Are you?

> It's like you have some sort of internet-caused brain disease [...]

What disease is it that causes someone to resort to insults when he lacks arguments?

> you miss the weight not only of what others say, but what you say also.
> You've gone down my post with a big checklist going "technically not a bug! [...]"

You claimed that all PPC hardware "of the past 20 years" has been buggy. I challenged this claim and what you came up with in response made clear that you didn't even grasp which claim in particular I was challenging, although it was in the quote.

> you're right in that a soldered down, useless low amount of ram is not a "bug" itself

And neither is any of your other examples.

> so my corrected statement: These computers are [...] buggy [...]

I see what you did there, and it's still untrue :-)

> the MorphOS team [...] will have a 68k emulation plus a PPC to can runs
> the current software available.

The question is whether this will be a transparent emulation like Trance is now, or just an encapsulated emulation like UAE, especially considering that MorphOS NG will most likely be a little-endian OS for reasons of software portability.

Your argument style is just to scrutinize posts on a line by line basis, looking for imagined grammatical flaws in casual speech to attack, rather than actually engaging with the meaning of that speech.

This causes you a big problem, you're obsessed with the Efika's ram "not being a bug". Nobody said it was, it was brought up as an example of how PPC Amigaish systems have generally sucked. What does this weird grandstanding over if it's a bug or not get you? It's still a shit and useless amount of ram, although now you'll probably dance on the edgecase of a pin to dispute the word "useless".

I told you I'm not interested in your turbo-pedantry already, so I dunno why you think doubling down on it will get me to take you seriously now. My main reaction is contempt for your socially maladjusted nature, reinforced by how adenoidal you sound complaining about me making fun of you for this.

[ Edited by khorse 27.02.2019 - 00:14 ]