@BigD

Quote:

The only sane answer is PPC...

Oh I see, so the rest of us that chose another option are insane?

Quote:

... when that ISA isn't available any more ... embrace emulation on x86

What @matthey said.

Quote:

There is not money for ...

There is money. But "money" is generally used to invest. The projects you've outlined can only be funded philanthropically.

Quote:

Vampire for retro compatibility with knobs on or a Tabor to push the platform forward there is nothing else IMHO and rightly so.

Sounds too righteous. It's my personal mission to make sure that those two are not the only options.

Quote:

... why screw it up when hardware and software is available ...

I guess you and I have different definitions of "available".

But overall, you have a very defeatist attitude. Sure things look a bit bleak, Tabor is still a no-show, Vampire development is not moving fast enough, the market for the Power ISA is shrinking, not much movement on the OS front, not much movement on porting some of the good ol' apps, prices on X5000 hardware are prohibitively high, we're looking at yet another protracted legal battle.

Um, where was I going with this? Oh yeah, but there are people in the community that are not giving up. They're not sitting idly by and watching the entire situation play out in despondent resignation. Despite things appearing hopeless, there is still hope. You may not like what these people have to say, you may not agree with their go-to-market plan, but they are thinking about moving beyond the current quagmire we're in. No matter how insane or wrongly so, I'm with these people.

_________________
All the way, with 68k

@OlafS25

Quote:

... why should any serious investor invest the money in something like AmigaOS?

Correct. No serious investor would find anything worth investing in, when it comes to Amiga OS.

The key is to create a market opportunity around a product that has greater and broader appeal, and have that product based on something that the Amiga OS and community benefit from by proxy. Like the halo effect.

Imagine you're working in the Toshiba HDD group in the late '90s and you propose to manufacture and market a 1.8" drive. Everyone else from the 3.5" classic and 2.5" portable HDD divisions thinks the idea has no merit. Then along comes Apple and use the drive in their new portable music player. The colleagues from the other divisions still think the 1.8" team is delusional. Who cares if Apple makes a portable music player using the new diminutive drive?

As we know, lots of people cared, and this sourcing agreement worked out very well. Toshiba didn't need to invest in making a portable music player, or make sure the iTunes software works with it, or work to port the iTunes software to Windows. Nevertheless, they made a bundle. Less than Apple did, but enough so that if they had some other far less profitable passion projects, they could fund them.

_________________
All the way, with 68k

@agami
Quote:

The key is to create a market opportunity around a product that has greater and broader appeal, and have that product based on something that the Amiga OS and community benefit from by proxy.

I want to thank everyone for their comments and views in this poll and for settling once and for all the direction of AmigaOS hardware.

So that's it then, IBM POWER for the future.


The End.

_________________
Tinkering with computers.

Quote:

Indeed. The ending of PPC support in Linux is certainly a bad sign for PPC. Regarding the X5000 and A1222, maybe an ARM accelerator could be developed in the future so that those who purchase the X5000 and A1222 now could eventually upgrade to ARM (or whatever CPU is chosen).

The x5000 would be an expensive dongle for cheaper ARM hardware.

Quote:

Very nice. So new hardware for Classic Amiga. Running AmigaOS 3.1 natively? I like this option. Trevor mentioned at AmiWest 2018 that they are considering updated software packs for 3.x. This would be a good fit, I think.

I would hope at least AmigaOS 3.1.4 would be used. I would expect the single core performance of an ASIC could be 10x that of the 68060 and custom hardware may allow for SMP giving better performance. With the improved performance, it should be possible to bring the 68k AmigaOS mostly on par with AmigaOS 4 (at least most of the API for simplified development).

Quote:

Amiga now has an online store like Steam. It's AMIStore. Good idea.

Yes. Good idea but it would work better if there were more Amiga users online.

Quote:

Here's an idea for another poll... Q: "Which AmigaOS platform would you prefer? Options: AmigaOS 3.x 68k Classic, AmigaOS 4.x 68k, or AmigaOS 4.x PPC.

There are too many hypothetical question marks with AmigaOS 4.x 68k. How good is compatibility? How good is performance on 68k hardware? Is new and faster 68k hardware included?

Quote:

I think this is a key part of the success of Amiga in the mass market. A bundle of software that people want on a capable computer at a competitive price. I think this is also why the Amiga attracted many artists to the platform back in the 80s and 90s. The Amiga provided artist software on a system capable of running it at a cost less than the competition. I think a combination like this may help to sell your standalone idea to more than simply Amiga fans.

The Amiga brought multimedia to the masses because it could do more with less. I think that philosophy still works today. Bundling can be an effective marketing technique.

Quote:

Hopefully. My guess would be either A-Eon or Cloanto. Or maybe Amiga, Inc.

I think Amiga Inc. is knocked out of the Amiga battle royal. Hyperion is bleeding and up against the ropes but still swinging. Cloanto countered strongly after being picked on and A-Eon has been keeping a low profile despite possible deals with other contestants.

Quote:

Not that the X5000 and future A1222 don't increase the number of users and developers, but your affordable hardware would probably help to increase it faster. A newly produced standalone Amiga 68k device would, I think, bring back previous AmigaOS users for a retro experience, and maybe some would stay to develop software for the device. These options are all viable.

The Tabor CPU would probably be considered mid performance despite missing units commonly found on mid performance hardware. Low performance PPC hardware has had disappointing performance. It makes sense to use 68k or ARM hardware for cheaper low performance hardware. Using 68k hardware would keep from fragmenting the user base more.

Quote:

Hmm. This would be interesting to implement and to see the result. Any way to simulate this for analysis?

I wrote the following in another post.

'The Apollo core did have 3 in-order superscalar integer pipes at one time but I am not aware of it ever having a dual ported DCache (I suggested a 3rd pipe after looking at average instruction length statistics for 68k code). I told Gunnar a dual ported DCache looked really good and may be necessary to make the 3rd pipe "worthwhile". There would still be only one DCache write per cycle allowed but loads are much more common than stores and would help instruction scheduling of true dependencies instead of making them more difficult to schedule. I could write an instruction scheduler for the 68040+ which includes a hypothetical 3 pipe in-order superscalar core with dual ported DCache. I have been tempted to write a 68060 instruction scheduler for vbcc before but the backend code generation is bad enough that I would just be rescheduling many instructions which should be removed.'

https://amigaworld.net/modules/newbb/viewtopic.php?mode=viewtopic&topic_id=42709&forum=2&start=320&viewmode=flat&order=0#816710

Gunnar probably doesn't want to dual port the DCache because he has one big L1 DCache. More levels of caches are needed when the CPU is much faster clocked than the memory (smaller caches are lower latency than larger caches). A CPU in an FPGA has a low clock speed so the high latency of large caches is not a problem. Dual porting a large cache would be expensive though.

Writing an instruction scheduler for vbcc that simulates a hypothetical CPU with 3 integer pipes and dual ported DCache could give an idea of how well it would perform. Perhaps vbcc is doing better than I expected based on some recent Quake results from Samuel Devulder.

quake.sasc.040 1.01.build#3477 (02/10/2019)
969 frames 35.46 seconds 27.33 fps

quake.sasc.030 1.01.build#3476 (02/10/2019)
969 frames 35.42 seconds 27.35 fps

quake.sasc.060 1.01.build#3478 (02/10/2019)
969 frames 35.28 seconds 27.47 fps

quake.gcc-6.4.1b.040 1.01.build#3471 (02/10/2019)
969 frames 33.67 seconds 28.78 fps

quake.gcc-6.4.1b.060 1.01.build#3472 (02/10/2019)
969 frames 33.18 seconds 29.21 fps

quake.gcc-2.95.3.040 1.01.build#3480 (02/10/2019)
969 frames 33.11 seconds 29.27 fps

quake.gcc-6.4.1b.030 1.01.build#3470 (02/10/2019)
969 frames 33.08 seconds 29.30 fps

quake.vbcc.040 1.01.build#3474 (02/10/2019)
969 frames 33.06 seconds 29.31 fps

quake.vbcc.030 1.01.build#3473 (02/10/2019)
969 frames 33.00 seconds 29.37 fps

quake.gcc-3.2.2.040 1.01.build#3483 (02/10/2019)
969 frames 32.89 seconds 29.46 fps

quake.gcc-3.2.2.060 1.01.build#3484 (02/10/2019)
969 frames 32.87 seconds 29.48 fps

quake.vbcc.060 1.01.build#3475 (02/10/2019)
969 frames 32.83 seconds 29.51 fps

quake.gcc-3.2.2.030 1.01.build#3482 (02/10/2019)
969 frames 32.78 seconds 29.56 fps

quake.gcc-2.95.3.060 1.01.build#3481 (02/10/2019)
969 frames 32.75 seconds 29.58 fps

quake.gcc-2.95.3.030 1.01.build#3479 (02/10/2019)
969 frames 32.68 seconds 29.65 fps

Vbcc is outperforming all but GCC 2.95.3 which I knew had one of the best compiler backends for integer and early versions of GCC 3.x some of which were buggy (Geek Gadgets support collapsed after GCC 3.4.0). I expect the improved vbcc performance is due to the improved assembly inlines, support code and FPU optimizations I worked on rather than an improved 68k backend (Volker hasn't had much motivation to improve it when there is no new 68k hardware). Perhaps a 68k instruction scheduler would be enough for vbcc to generate the best performance code of any Amiga C compiler, at least for mixed integer and FPU code. Bebbo may have something to say about that but improving the code generation of GCC 6.x looks like more of a challenge considering the results above.

Quote:

Better marketing may help. That costs money, though.

Online advertising can be relatively cheap (smart use of Google Adwords for example). Cheap products don't need as much advertising as word gets around. How many Raspberry Pi adds have you seen?

@agami

Quote:

The key is to create a market opportunity around a product that has greater and broader appeal, and have that product based on something that the Amiga OS and community benefit from by proxy. Like the halo effect.

Similar to what I've thought, too. Two ideas I've had: (1) an art studio using Amiga software for art and animation while contributing to software code during production; (2) a company that runs Linux on an X5000 for business processes while contributing to Amiga coding efforts that are useful to the business.

@Signal

Quote:

I want to thank everyone for their comments and views in this poll and for settling once and for all the direction of AmigaOS hardware. So that's it then, IBM POWER for the future

I vote for going to the top: POWER8/9. AmigaOS could tap into the AI market that IBM is promoting for the POWER9 processor. Blender rendering on AOS4 on a POWER8/9 would make a good 3D workstation.

@matthey

Quote:

I would hope at least AmigaOS 3.1.4 would be used. I would expect the single core performance of an ASIC could be 10x that of the 68060 and custom hardware may allow for SMP giving better performance. With the improved performance, it should be possible to bring the 68k AmigaOS mostly on par with AmigaOS 4 (at least most of the API for simplified development).

AmigaOS 3.1.4 with RTG on new and faster hardware at $50-$100 would be a good addition. Although, are you sure the price could even be below $100 as most recent 68k accelerators are above $100?

Quote:

I think Amiga Inc. is knocked out of the Amiga battle royal. Hyperion is bleeding and up against the ropes but still swinging. Cloanto countered strongly after being picked on and A-Eon has been keeping a low profile despite possible deals with other contestants.

Sounds like a chess game.

Quote:

davidf215 wrote: I vote for going to the top: POWER8/9. AmigaOS could tap into the AI market that IBM is promoting for the POWER9 processor. Blender rendering on AOS4 on a POWER8/9 would make a good 3D workstation.

AmigaOS has a long road ahead before it would be competitive on POWER hardware for any market. Porting AmigaOS 4 to the POWER hardware would likely be relatively easy though. The POWER9 hardware does appear to have competitive performance and energy efficiency for the server market but pricing is just as important.

Qualcomm's ARM AArch64 attempt at the server market is practically dead. This may help POWER as the small RISC portion of the market is usually shared. I would not expect any more ARM attempts after all the big name failures but investors seem to like RISCy investments.

Quote:

AmigaOS 3.1.4 with RTG on new and faster hardware at $50-$100 would be a good addition. Although, are you sure the price could even be below $100 as most recent 68k accelerators are above $100?

The FleaFPGA was $45 U.S. without mass production.

https://www.indiegogo.com/projects/fleafpga-ohm-fpga-experimenter-board-arduino

FPGAs are turned into ASICs to improve performance, lower power consumption and reduce per unit cost in quantity. Parts on the board, the board itself and assembly would likely be significantly cheaper in quantity as well. The only thing expensive is the AmigaOS but the economics on that need adjusting if serious about expanding the Amiga user base.

Quote:

Sounds like a chess game.

It would have to be multiple chess games as there are more than 2 Amiga players. Maybe a battle royal of chess.

@matthey

Quote:

AmigaOS has a long road ahead before it would be competitive on POWER hardware for any market.

Competitive?

It's Amiga, not Linux or windows or apple something or other. Amiga!

_________________
Tinkering with computers.

Quote:

Signal wrote: Competitive? It's Amiga, not Linux or windows or apple something or other. Amiga!

Ok, POWER may be competitive in the Amiga PPC/POWER market which is maybe 5000 users. Of those, most would not be interested and those who would pay that much likely already have an X5000.

Let's say the AmigaOS developers pull off the miracle of adding SMP, adding 64 bit support, improving memory protection and adding security without breaking compatibility (AmigaOS and MorphOS developers have been unsuccessful after many years and AROS x86_64 breaks compatibility). The AmigaOS would *not* be attractive in the server and workstation markets due to lack of software. IBM has the same problem with Linux and BSD for POWER and that is why they are supporting ppc64le to ease endian issues with ports (most servers are x86_64 hardware and can use the same software as desktops). The AmigaOS would want big endian for compatibility and it is less POSIX compatible making it that much more difficult to port software. It would take an epic effort to bring the AmigaOS up to par and the best place to start would probably be breaking compatibility.

Last edited by matthey on 22-Feb-2019 at 01:58 AM.

@matthey

After I posted my POWER9 comment, I did a search for POWER9 performance charts. I found that the Intel Xeon Gold product line is keeping up with POWER9. In some cases the Xeon Gold performs better than POWER9. So it would be neat to see Aros running on Xeon, if it doesn't already. If the rumor is true of MorphOS being ported to x86, then maybe it could run on the Xeon processor.

*edit add*

Quote:

The FleaFPGA was $45 U.S. without mass production.

Very cool. So how much would a prototype cost?

Last edited by davidf215 on 22-Feb-2019 at 04:51 AM.

While comments in thread keep in their castling, talking about power9 and other useless hypothesis, support for X86_64 keeps growing, and growing..

@vision

Quote:

While comments in thread keep in their castling, talking about power9 and other useless hypothesis, support for X86_64 keeps growing, and growing..

I'm sure if there was a poll option for Hyperion open sourcing OS4.x that would keep growing and growing as would the poll option for Trevor to personally fund the porting of the OS to x89-64 / Arm / Power9 etc. It is very easy to demand these radical changes to an OS that has struggled to implement SMP and update classic apps to PPC despite that being a small challenge compared to porting the whole OS and apps and adding backwards compatibility to an entirely new CPU line

I vote Hyperion and Trevor each give every last active Amiga owner $20 each and the AmigaOS 4.1 FE source code and a free X5000 and a pre-release Tabor board

x86-64 didn't save BeOS and neither will a port of AmigaOS to x86-64 with no software. Just because it's cheap will not automatically bring new users. BootCamp is cheap but I only use it on my Mac for specific software. What specific software would tempt a new user to boot AmigaOS rather than Windows or Linux on their x-86-64 PC? If you want an Amiga(One) buy the whole package. The only machine that WOULD grow market share outside of PPC would have been the RPi but we would now be late to market and RiscOS has taken the glory there IMHO.

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

@davidf215

Eehh, the price of Xeon Gold is 1000USD+, and that is only the cpu.

@BigD

Exactly, cheap Mac’s did not bring thousands of new Morphos users either.

Quote:

davidf215 wrote: After I posted my POWER9 comment, I did a search for POWER9 performance charts. I found that the Intel Xeon Gold product line is keeping up with POWER9. In some cases the Xeon Gold performs better than POWER9. So it would be neat to see Aros running on Xeon, if it doesn't already. If the rumor is true of MorphOS being ported to x86, then maybe it could run on the Xeon processor.

The x86_64 server CPUs typically perform better than POWER9 in benchmarks as they have better single core single thread performance. In other words, most programs and games would be faster and finish more quickly on the x86_64 CPUs. Although there is a higher power draw, this is good for energy efficiency as work is done faster. RISC CPUs can't compete with x86_64 here so the POWER designers worked on more efficient multi-threading to improve the total amount of work done per core by multiple threads. POWER9 needs to have multiple threads on each core (as typical for a server) at which point it offers an energy savings. It would gain more server market share except that it is generally more expensive (higher up front costs) and has less software support than competing x86_64 server CPUs.

Quote:

Very cool. So how much would a prototype cost?

Most prototypes are done in an FPGA. FPGA costs vary by size and features. Ideally, an FPGA which was planning for an ASIC would hold all the logic for an SoC design so would be expensive. This is in contrast to smaller more affordable FPGAs used in low quantity production.

There are different kinds of ASICs (or ASSPs). The simplest solution would be to take synthesizable HDL source and give it to an ASIC service to make a structured or cell based ASIC. This would give an approximation of the logic in an affordable ASIC although the performance may not be much better than in an FPGA. FPGAs actually use a small die size. The Cyclone V uses a 28nm process which is smaller than that of the Raspberry Pi 3 SoC at 40nm. The FPGA is inefficient because of poor placement and routing of elements and paths (wires) in the FPGA. Smaller fab processes allow greater transistor density and faster transistor switching but have smaller paths which increase resistance. Paths from metal layers can be placed above the chip in increasing sizes as distances increase but this raises per unit production costs. Structured ASICs use standard metal routing paths instead of completely customized paths. Rather than a cheap ASIC conversion, I would want an experienced professional to help with the design and ASIC creation.

@vision

Quote:

While comments in thread keep in their castling, talking about power9 and other useless hypothesis, support for X86_64 keeps growing, and growing..

To my knowledge x64 was always growing since its inception and never faltered.

But, apparently, Acer own the Commodore/Amiga assets and have a clause about no x86 ports. Which I take would extend to x86/64 as well.

So, by the looks of things, talk of anything x86 woud remain equally as a hypothesis if Acer put it to an instant halt.

@matthey

Quote:

matthey wrote: ... Rather than a cheap ASIC conversion, I would want an experienced professional to help with the design and ASIC creation.

That would be expensive. Many M$ for wages, software licenses, masks, manufacturing, packaging, testing...
There are many good reasons companies that created fully custom ASICs in the past now buy extremely expensive FPGAs instead.

Quote:

megol wrote: That would be expensive. Many M$ for wages, software licenses, masks, manufacturing, packaging, testing... There are many good reasons companies that created fully custom ASICs in the past now buy extremely expensive FPGAs instead.

Even professional CPU designers often avoid full custom ASIC designs. I would want at least an experienced project consultant for advice well before making ASIC desicions rather than not planning and turning over synthesizable HDL code to a cheap ASIC service.

https://www.theregister.co.uk/2019/02/23/linus_torvalds_arm_x86_servers/

@Hypex

Quote:

But, apparently, Acer own the Commodore/Amiga assets and have a clause about no x86 ports. Which I take would extend to x86/64 as well.

It still leaves the door open for an Arm version if a port could be justified for use on the RPi.

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

@matthey
I meant that companies that in the past did full custom now simply buys very expensive FPGAs instead to save money. Even cell based designs are costly.

@hardwaretech
Don't see what's interesting about Linus opinions in some area he don't know much about.