@matthey

Quote:

The poll was a joke based on a post about PPC by ppcamiga1.

redfox

BigD Quote:

It's sad it's legacy wasn't bigger but at least it started with the 68k and faded before people remembered it as a sad PPC Mac wannabe. The AmigaOne project technically peaked with the X1000 which used PA-Semi chips that had already been snapped up by Apple! There was no second coming through PPC and only through ARM and new 68k FPGA designs like the 68080 Apollo core have we forged ahead with anything of lasting significance.

The buyout of PA-Semi was unlucky. Apple did more to kill PPC than anyone else and they were part of the AIM alliance.

1. Exponential Technology
https://en.wikipedia.org/wiki/Exponential_Technology
Exponential Technology had a X704 PPC CPU that ran at close to 500MHz using a 500nm chip fab process while the 68060@50MHz is by far the most common clock rating using a 500nm process also (68060@66MHz production quantities were promised for late 4Q94 but didn't materialize). Apple stopped the PPC clone market and strung Exponential along with contract negotiations until the startup ran out of cash in 1997. Apple then purchased the successor business Intrinsity.

2. PA Semi
https://en.wikipedia.org/wiki/P.A._Semi
PA Semi was another startup business that designed the PPC PWRficient PA6T-1682M CPU and planned other CPUs based on the modular design. Apple bought them out in 2008, discontinued PPC production as fast as they could and switched the developers to ARM development.

3. Mac transition to Intel processors
https://en.wikipedia.org/wiki/Mac_transition_to_Intel_processors
Steve Jobs announced at the 2005 Worldwide Developers Conference that Apple would gradually stop using PPC microprocessors.

I believe the loss of the PPC desktop market was a blow to the PPC embedded market as well. The following embedded market study shows PPC for embedded use in decline from 2008 to 2009.

https://www.embedded.com/wp-content/uploads/2019/12/2009_embeddedmarketstudy_ubm.pdf

See the question "Which of the following 32-bit chip families would you consider for your next embedded project?" on page 62. "Freescale 68K, ColdFire" was still considered by 23% of respondents in 2008 and 19% in 2009 despite minimal support and new development. PPC results are divided among PPC producers with IBM and AMCC retaining PPC support better than Freescale.

BigD Quote:

To demonstrate, I play the PS1 version of WipEout 2097 on THEA500 Mini using the Pandory500 softmod and not the Amiga PPC version on an overpriced AmigaOne or CyberstormPPC/BlizzardPPC board! Why would anyone do that! MorphOS continues to be a shining beacon of sensible compromise and may survive on Intel chips but the PPC period was pretty much a disaster for anyone other than Apple!

PPC was bad for the Mac too as Apple lost desktop market share during the mid to late 1990s. PPC was disappointing, especially for low end hardware.

https://en.wikipedia.org/wiki/PowerPC_600#PowerPC_603 Quote:

The 603 was intended to be used for portable Apple Macintosh computers but could not run 68K emulation software with performance Apple considered adequate, due to the smaller processor caches. As a result, Apple chose to only use the 603 in its low-cost desktop Performa line. This caused the delay of the Apple PowerBook 5300 and PowerBook Duo 2300, as Apple chose to wait for a processor revision. Apple's use of the 603 in the Performa 5200 line led to the processor getting a poor reputation. Aside from the issue of 68K emulation performance, the Performa machines shipped with a variety of design flaws, some of them severe, related to other aspects of the computers' design, including networking performance and stability, bus problems (width, speed, contention, and complexity), ROM bugs, and hard disk performance. None of the problems of the 5200 line, aside from 68K emulation performance, were inherently due to the 603. Rather, the processor was retrofitted to be used with 68K motherboards and other obsolete parts. The site Low End Mac rates the Performa 5200 as the worst Mac of all-time. The 603 found widespread use in different embedded appliances.

The PPC 603e with double the caches of the PPC 603 and 68060 was rushed into production. The PPC 604 also had its caches doubled to improve performance. The 68k was a cache and memory miser and the PPC was a resource hog. Apple Almost went bankrupt in 1997 soon after switching to PPC and sales picked up after switching to x86.

AmiRich Quote:

Not quite dead, POWER9 is still competitive with current Intel. But the AmigaOne systems of the past 20+ years have pretty much ruined the reputation of PPC as a popular CPU product. Just look at the A1222+: a SoC with no AltiVec or even MMU - makes PPC look like a joke to anyone unfortunate enough to stumble across it, and the other AmigaOne systems aren't really any better better.

I have serious doubts about POWER CPU competitiveness with x86-64 CPUs. Single thread performance/MHz is important for the desktop and games where I don't think POWER can compete. I believe x86-64 CPUs are significantly cheaper too but it could be because of economies of scale.

There have been many disappointing PPC CPU core designs like the PPC 603, PPC G5 and PPC Cell but this may have something to do with the PPC architecture. The A1222+ CPU was a bad design decision but it was intended for very low end embedded use where the lack of standard FPU wasn't supposed to matter even though development tools do matter. PPC was not designed to be programmed in assembly and requires mature higher level development tools. The 68k in contrast is one of the easiest architectures to program in assembly ever. Many 68k game programmers chose to program their games completely in assembly. Embedded programmers need to get low level sometimes for optimizations, features and debugging as well which is one of the reasons why the 68k was much preferred for embedded use along with the much smaller footprint.

AmiRich Quote:

So not extinct, I'm sure PPC is still competitive in some niche products, but the most likely consumer path into a PPC system, AmigaOne, is littered with low-performance systems nobody outside of a handful of Hyperion fanatics wants. It goes without saying this isn't a great look for the Amiga world in general either.

The X1000 and X5000 were fine for what they were. High end PPC CPUs had better performance than most RISC architecture CPUs like ARM and MIPS offerings which allowed PPC to survive for high end embedded use for many years. AArch64 is similar to PPC and supersedes it in almost every way which led to the swift demise of PPC. ARM looked at SuperH because of RISC code density and improved on it with Thumb and Thumb-2 which gave them the embedded market. ARM looked at PPC for RISC performance and improved on it with AArch64. Now ARM thinks they can use their new AArch64 ISA to take on the CISC x86-64 ISA with all its compatibility baggage. CISC architectures can be very high performance and very good code density even at the same time but x86-64 is lacking in areas, even compared to the 68k.

Last edited by matthey on 04-Nov-2024 at 09:36 PM.

@AmiRich

your confused, A1222 has MMU, its the 68080 that does NOT have a MMU

A1222 has a SPU insted of a FPU (Its incompatible),
so it was fixed it in software after many months (years)..

Quote:

But the AmigaOne systems of the past 20+ years have pretty much ruined the reputation of PPC as a popular CPU product.

I will blame the embedded market for that, this embedded chip should never have come close to a desktop system. They are designed for automation.

one thing you can be sure about when using AMD or Intel chip, is that no instructions are missing.

You also be a bit confused if you read a book, with some of the pages ripped out.

Last edited by NutsAboutAmiga on 04-Nov-2024 at 09:32 PM.
Last edited by NutsAboutAmiga on 04-Nov-2024 at 09:31 PM.
Last edited by NutsAboutAmiga on 04-Nov-2024 at 09:28 PM.
Last edited by NutsAboutAmiga on 04-Nov-2024 at 09:19 PM.

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

@agami

Hehe,
I have to admit I would buy a Talos / Raptor running AOS /
morphOS with Power 7,8,9 or 10

These systems have been around for ages, and stupidly fast.

Wol.

_________________
It is my conviction that killing under the cloak of war is nothing but an act of murder.~Albert Einstein

@matthey

Quote:

matthey wrote: The buyout of PA-Semi was unlucky. Apple did more to kill PPC than anyone else ...

The first-generation Intel-based Macs were released in January 2006 with Mac OS X 10.4.4 Tiger.
On 23 April 2008, Apple announced that they had acquired P. A. Semi for $278 million.

Officially Apple announced they bought P.A. Semi for iPhone chips, seems still weird to me because many talents left P.A.Semi after it was acquired by Apple.

The AIM Alliance became a failure very soon after it was founded:

1. IBM did not release OS/2 for PPC, they had a port but didn't release it.
2. Apple used to much time with the development for an MacOS successor (I'n not talking about OSX)
3. Motorola didn't provide the right CPUs for Apple, the advantage of PPC was lost soon.
4. Apple/Jobs destroyed the market of Mac compatibles.

So PPC ended with all major OS providers gone, no hardware market with free competing manufacturers, expensive or over complex CPUs not suitable for desktop systems, and MCUs slower and more expensive than ARMs.

Amiga (AOS4/MOS) with its tiny user base doesn't play a role on any of these events. The PPC went down, without any influence from us.

@Wol

They are noisy as hell, as some people found out, serves typical has SAS, and registerd memory, instead of unresisted memory that’s normal for PC’s, formfactor is also not always ATX, but blade format.

There are people who have created desktops out of servers, and second hand servers are not always that expensive.

Last edited by NutsAboutAmiga on 04-Nov-2024 at 10:00 PM.
Last edited by NutsAboutAmiga on 04-Nov-2024 at 09:57 PM.

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

Wol Quote:

Hehe, I have to admit I would buy a Talos / Raptor running AOS / morphOS with Power 7,8,9 or 10 These systems have been around for ages, and stupidly fast.

I hope so. A CPU with 3,162 68060 cores may have some parallel performance too, especially on more modern silicon like the POWER 9.

68060 - 2,530,000 transistors
Power 9 - 8,000,000,000 transistors (3,162 times as many transistors as the 68060)

https://en.wikichip.org/wiki/ibm/microarchitectures/power9

Power 9 is an option for Trevor's Amiga collection and the Amiga classes. The Amiga masses can always eat emulation cake, according to Trevor.

Last edited by matthey on 05-Nov-2024 at 01:54 AM.
Last edited by matthey on 04-Nov-2024 at 10:04 PM.

@matthey

Quote:

matthey wrote: I hope so. A CPU with 3,162 68060 cores may have some parallel performance too, especially on more modern silicon like the POWER 9. 68060 - 2,530,000 transistors Power 9 - 8,000,000,000 transistors (3,162 times as many transistors as the 68060) https://en.wikichip.org/wiki/ibm/microarchitectures/power9 Power 9 is an option for Trevor's Amiga collection and the Amiga classes. The Amiga masses can always eat emulation cake, according to Trevor.

I can purchase relatively low-cost POWER9 CPUs, but it's missing a matching low-cost motherboard.

There's no guarantee Trevor's "System 54" will run on a randomly selected 3rd party POWER9 motherboard.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@matthey

Quote:

https://spectrum.ieee.org/the-truth-about-benders-brain

It proves nothing. 6502 CPU is not a creditable modern "AI" chip. Hint: Tesla Optimus

https://news.accelerationrobotics.com/tesla-optimus-robot-brain-computer-architecture-hardware-software/

Tesla's Bot Brain featuring 1x Tesla SOC (left) and the Robotic Processing Unit (right), a robot-specific processing unit that contains CPUs, FPGAs and GPUs and maps the Robot Operating System (ROS) efficiently to them for best performance

Optimus 2022's Tesla SoC has 12 ARM-72 CPU cores with 37 TOPS INT8 AI NPU.


Quote:

The 6502 is about as simple as a CPU gets which means extremely low power for battery use. The Z80 is a small step up in complexity and still alive in MCU form because it has a better ISA and better code density which may be better with more code. These CPUs are not disappearing anytime soon although they may be hidden in deeply embedded hardware because they are tiny cores.

There's ARC RISC CPU, https://en.wikipedia.org/wiki/ARC_(processor)


Quote:

The original ARM ISA is on the way out. ARM is starting to develop AArch64 only cores except for their Cortex-M cores. .

Where's 32-bit development for 6502 i.e. 65832? Apply the same standard for ARM with 6502.

Your argument is filled with double standards and inconsistencies.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@matthey


Nemo board is fine for what it is, again software isnt
I was one of rare users to publish results.

https://openbenchmarking.org/result/1303300-FO-AMIGAONEX47

Results very its CPU wise weak per Mhz less then G4
FPU is OK
Memory controller is great

_________________
OS 3.x AROS and MOS supporter, fi di good, nothing fi di unprofessionalism. Learn it harder way!
SinclairQL and WII U lover :D
YT http://www.youtube.com/user/rasvoja

Hammer Quote:

It proves nothing. 6502 CPU is not a creditable modern "AI" chip. Hint: Tesla Optimus https://news.accelerationrobotics.com/tesla-optimus-robot-brain-computer-architecture-hardware-software/ Tesla's Bot Brain featuring 1x Tesla SOC (left) and the Robotic Processing Unit (right), a robot-specific processing unit that contains CPUs, FPGAs and GPUs and maps the Robot Operating System (ROS) efficiently to them for best performance Optimus 2022's Tesla SoC has 12 ARM-72 CPU cores with 37 TOPS INT8 AI NPU.

AI units are huge and Bender has a primitive pea brain in Futurama.

Hammer Quote:

There's ARC RISC CPU, https://en.wikipedia.org/wiki/ARC_(processor)

The register file of even the 16 GP register RISC ARC ISA likely uses more transistors than the whole 6502 CPU. The original ARM CPUs were very simple and had 16 GP registers too.

CPU | bits | GP registers | transistors
4004 4 17 2,300
6502 8 3 3,510
Z80 8 4 8,500
6809 8 5 9,000
ARM1 32 16 24,800
ARM2 32 16 27,000
8086 16 8 29,000
68000 16/32 16 68,000 (microcoding uses maybe 1/3 of transistors)

Smaller than the 6502 is possible as the Intel 4004 demonstrates with a 4-bit CPU. The 8-bit 6502 is very simple and has good compiler support with a large library of code. The Z80 with a better more orthogonal ISA, more GP registers and better code density is a better option for larger code. With an OS, the 6809 may be preferable with 2 stack pointers and good position independent code (PC relative) support which were carried over to the 68000. With more than 64kiB of memory needed, the CPUs with 32-bit ISAs supporting large flat address spaces become preferable.

Hammer Quote:

Where's 32-bit development for 6502 i.e. 65832? Apply the same standard for ARM with 6502. Your argument is filled with double standards and inconsistencies.

The 6502 ISA is poor and has limited upgrade potential but that doesn't make it any less useful for what it already is. It is useful where the smallest area and lowest power 8-bit CPU with simple code is needed but that is hundreds of million of CPUs per year.

vox Quote:

Nemo board is fine for what it is, again software isnt I was one of rare users to publish results. https://openbenchmarking.org/result/1303300-FO-AMIGAONEX47 Results very its CPU wise weak per Mhz less then G4 FPU is OK Memory controller is great

The PA Semi developed PWRficient PA6T-1682M CPU is targeting power efficiency (performance/Watt) for the embedded market. There are Power, Performance, Area (PPA) tradeoffs that can be made in CPU core design. A Desktop core design may target performance and performance efficiency (performance/MHz) more. PA Semi was smart to target the embedded market first where they already had design wins unlike the Exponential Technology startup which targeted the desktop market with a very high power design as their first offering only to have Apple practically deny them the desktop market. The extremely high clocked Exponential X704 CPU may have looked more impressive and been better for marketing but it was less practical than high clocked Alpha CPU designs as it was less power efficient and I expect it was severely bottlenecked by the tiny 2kiB direct mapped L1 instruction cache (a 68020/68030 256B direct mapped I-cache should have the performance of a PPC 1kiB direct mapped I-cache but it is clocked to 50MHz instead of 500MHz). I believe the PWRficient PA6T-1682M CPU is a much better design although it may be more comparable to a mobile CPU core design like would go into a laptop.

Last edited by matthey on 05-Nov-2024 at 08:42 PM.
Last edited by matthey on 05-Nov-2024 at 08:37 PM.

@PPC

PPC is dead.

The ARM workstations are coming: System76 Thelio Astra

Last edited by agami on 06-Nov-2024 at 01:40 AM.

_________________
All the way, with 68k

agami Quote:

PPC is dead. The ARM Workstations are coming: System76 Thelio Astra

Purpose built ARM desktop/workstation hardware? No.

It looks like embedded hardware for an autonomous vehicle was put in a tower case. Probably lots of AI performance if you don't mind AI units driving up your price. It has more networking capabilities than a desktop is likely to use and even a "1 x DB15 (VGA)" for retro use which probably attached to some kind of display within the vehicle. Maybe it is a better choice than Ampere Computing ARM server chips though.

Ampere Computing (fabless semiconductor developer)
business investors: Arm Holdings, Oracle Corporation
business partners: Microsoft, Oracle Corporation

Raspberry Pi Holdings (fabless semiconductor developer)
business investors: Sony, Arm Holdings
business partners: ?

SiFive (fabless semiconductor developer)
business investors: Intel, Qualcomm, Western Digital, SK Hynix, Saudi Aramco
business partners: Renesas Electronics

Fabless semiconductor developers are getting a lot interest and investments. Well, that's not anything Amiga related is it? Amiga forever has AmigaNOne and AmigaNOwhere. Only Amiga makes it possible.

Last edited by matthey on 06-Nov-2024 at 01:31 AM.
Last edited by matthey on 06-Nov-2024 at 01:29 AM.

@matthey


Quote:

matthey wrote: Purpose built ARM desktop/workstation hardware? No.

Just posting here as a sign of things to come.

Ego-driven individuals such as Trevor are likely to get more interested in high-end ARM64 workstations with Nvidia PCIe graphics, than ARM64 SoC solutions.
You know he prefers big-box Amigas and doesn't want the Amiga to be a PC.

In about a year, there should be a handful of sub $2,000 ARM64 modular desktop solutions. And then a year after that we should see sub $1,000 ARM64 modular desktop solutions and laptops competing with the mainstream x64-based open computing platform.

With the right level of software investment in AROS ABI v1 over the next couple of years, it should be able to yield a better UX on ARM64 than AmigaOS 4 UX on PPC.

Then it becomes easier to support Raspberry Pi hardware for those who wish to keep things under $500.

Because PPC is dead.

_________________
All the way, with 68k

@OneTimer1

but POWER9 are not available for that price and will never

agami Quote:

Just posting here as a sign of things to come. Ego-driven individuals such as Trevor are likely to get more interested in high-end ARM64 workstations with Nvidia PCIe graphics, than ARM64 SoC solutions. You know he prefers big-box Amigas and doesn't want the Amiga to be a PC.

Trevor invested in AMD GPU drivers not Nvidia. Sure, he has outdated tunnel vision so high performance PCIe on a large expensive board is what he wants. It was not what was popular with the Amiga but he doesn't really understand the philosophy of the Amiga with its integration and elegance or the 68k and chipset that made it possible and good. He doesn't even seem to understand computers in general or he would have rejected the A1222+ CPU. He probably thinks PPC is better than AArch64 even though it improved on PPC in almost every way leaving nothing for PPC. At least the 68k is different and still has advantages over AArch64 (code density, assembly readability, smaller ISA/cores), especially for low end computers. I wonder if he has ever done any programming because Trevor seems to like the 6502 family C64 and his PPC AmigaNOne abominations better than the 68k Amiga.

agami Quote:

In about a year, there should be a handful of sub $2,000 ARM64 modular desktop solutions. And then a year after that we should see sub $1,000 ARM64 modular desktop solutions and laptops competing with the mainstream x64-based open computing platform.

From Ampere Computing? There has been ARM SoCs with PCIe support although gen 3 x4 lanes in a x16 slot is about the best I have seen. Some embedded hardware needs the high speed but expensive SerDes and a SoC can support some flexibility with the allocation to I/O but a desktop board is likely to be different and there hasn't been enough economies of scale for them. SiFive has RISC-V semi-desktop hardware with the advantage that they have the whole RISC-V market compared to the divided ARM market but with the disadvantages that a competitive high end OoO RISC-V CPU core is very expensive to develop and the RISC-V ISA is weak which doesn't help. Their attempts are still available at least.

HiFive Unmatched Rev B (https://www.sifive.com/boards/hifive-unmatched-revb)
SiFive Freedom U740 SoC (4xU74 1xS71 in-order CPU cores with roughly Cortex-A55 performance)
16GiB DDR4 memory
32MB Quad SPI Flash
MicroSD Card
Gigabit Ethernet Port
4x USB 3.2 Gen 1 Type A Ports (1 Charging Port)
1x MicroUSB Console Port
x16 PCIe Gen 3 Expansion Slot (8-lanes Usable)
M.2 M-Key Slot (PCIe Gen 3 x4) for NVME 2280 SSD Module
M.2 E-Key Slot (PCIe Gen 3 x1) for Wi-Fi / Bluetooth Module
Industry Standard Mini-ITX
$299 USD at Mouser (https://www.mouser.com/ProductDetail/SiFive/HF105-001?qs=Imq1NPwxi75JBw6ulD0quQ%3D%3D)

This first attempt at desktop hardware is an adequate entry level but not high end setup. It compares nicely to Amiga1 boards but with a much cheaper price. The price above was dropped considerably likely due to lack of demand but I don't see anywhere that it is discontinued. The U74 in-order CPU cores are very small and cheap with an extra S71 embedded monitor core with a similar design thrown in. The low power from in-order CPU cores and the low price give it a dual purpose for low end desktop and high end embedded markets.

HiFive Premier P550 (https://www.sifive.com/boards/hifive-premier-p550)
ESWIN EIC7700X SoC (4x P550 OoO CPU cores, Imagination AXM-8-256 onboard GPU, AI NPU)
16GB LPDDR5 + 128GB eMMC
2x 10/100/1000 Ethernet
1x M.2 Key E connector for Wi-Fi / Bluetooth module (Interface via SDIO / UART)
1x PCI Express Gen 3 x4 via a PCIe x16 slot
2x Stacked USB 3.2 Gen 1 Type-A Connectors
1x USB 19-pin male connector to support 2 USB 3.2 Gen 1 connectors on front panel
1x USB type-E connector to support 1 USB 3.2 Gen Type-C connector on front panel
1x JTAG Header
1x SATA 3 connector (6Gb/s)
1x microSD card connector
1x USB Type-C (USB2 only) connector for debug UART/JTAG support through FT4232H
1x HDMI 2.0 connector
1x Mini-ITX case compliant Front Panel Connector
1x CR1220 battery holder for Real Time Clock
3x fan headers
1x header for audio interface (front panel stereo line-out and line-in / microphone signals)
1x rear panel stereo jack with microphone input
1x 40-pin Peripheral I/O Header (1 I2C, 1x QSPI, 1x UART, 16x GPIO)
Mini-ITX form factor requiring standard ATX power supply
$599 USD (https://www.arrow.com/en/products/hf106/sifive-inc)

This newer SiFive desktop offering includes upgraded higher performance OoO CPU cores, an integrated GPU in the SoC, LPDDR5 and SATA 3 but the PCIe dropped from 8 to 4 lanes in the x16 connector. It wouldn't be difficult to support 16 SerDes lanes in a x16 PCIe connector but it is expensive where a GPU is integrated now and easier to support. ARM hardware designers often choosing to integrate the GPU into the SoC and save on PCIe SerDes but this is not true for x86-64 desktop hardware, the reasons being the following.

1. x86-64 has Windows with software
2. x86-64 has plug and play PCIe GPU gaming

ARM and RISC-V can't match these advantages which is why x86-64 will not be dislodged from the desktop market anytime soon. However, Windows adds a further cost to x86-64 systems which adds up with x86-64 baggage and bloat. An expensive ARM desktop board is like a frontal attack on a nearly impenetrable fortress. The price has to be aggressively reduced to compete and integrating the GPU is the place to start. The RPi has successfully created a micro-desktop standard and Eben Upton has been very careful not to scale up the RPi 5 too far into x86-64 territory. However, the OoO CPU cores may have been scaled up too far using the power and transistor budget and leaving the standard low power GPU behind, especially for gaming. So far, he has not been interested in creating a RPi micro-console standard but then the RPi lacks games like the Archimedes and customers are left to create their own micro-consoles playing mostly retro games using emulation. The RISC-V StarFive VisionFive 2 has a better integrated GPU than the RPi and is under $100 USD using SiFive IP which is good but even less support and software. SiFive licenses their IP with I believe no royalties so fabless semiconductor businesses could license it and add non-RISC-V cores to the micro-desktop SoCs above or reconfigure it as needed since it is modular. ARM may already have adequate IP to create similar hardware although they try to lock in customers to their IP including GPUs with royalties. It would be better to choose a better gaming ISA than ARM anyway.

6502 (little endian but doesn't matter much for an 8-bit CPU)
computers: Apple I-III, Pet, C64, Atari 400/800, Acorn various
consoles: Atari 2600/5200/7800, NES, SNES, TurboGrafx-16, Atari Lynx
+ largest game library?
+ easy core IP licensing from Western Design Center
- very limited upgradeability
- poor performance metrics

68k (big endian)
computers: Amiga, Atari ST(E)/TT/Falcon, Apple Mac/Lisa, X68000, NeXTcube/NeXTstation, Sinclair QL, workstations
consoles: Sega Genesis/Mega Drive/CD, Neo Geo, CD32
+ highest quality 2D game library
+ good upgradeability
+ good performance metrics
+ very good code density for a small footprint
+ good selection of open and licensable cores available

SuperH (big endian)
computers: ?
consoles: Sega 32X, Saturn, Dreamcast
+ J Core project revived with open cores
+ small cores
+ good code density
- limited games
- no computers?

MIPS (big endian)
computers: SGI and other workstations
consoles: PS1, PS2, N64
+ large game library
- poor code density

PPC (big endian except XBox 360)
computers: Apple PPC Mac, Amiga1
consoles: Nintendo GameCube/Wii/Wii U, PS3, XBox 360, Apple Pippin
+ semi-modern
+ wide selection of cores from different businesses increases licensing chances
- advanced 3D chipsets difficult to support
- PS3 Cell CPU difficult to support
- poor code density

The 6502 CPU and chipsets can be simulated in an affordable FPGA and the CPU doesn't have much upgrade potential anyway. The 68k has a nice 2D game library for retro fans, cores and chipsets are relatively small and code density is very good for a cost savings, the 68060 has proven semi-modern performance and the Amiga, Atari ST(E)/TT/Falcon, Apple Mac, X68000 and CD32 have software compatible with it. SuperH and MIPS games are primarily for consoles. Early 3D games don't hold up well and have to compete with later, similar and better 3D games. The PPC generation has a respectable number of games but the hardware is varied and more difficult to add compatibility. The 68k Amiga has emulators for the Atari ST, 68k Mac, Sega Genesis and Neo Geo where the CPU is not emulated. PPC Mac and Amiga1 hardware have difficulty emulating the PPC consoles which could give them many games. I've seen a newer x86-64 gaming laptop emulating a Nintendo PPC Wii console well though. It is likely possible to beef up PPC hardware to play the console games but it would be competing with x86-64 hardware in their citadel of power. Opportunities are in micro-desktop and micro-consoles with standardization and leveraging large game libraries with less hardware. Emulation needs more hardware. Most of the retro gaming architectures were or defaulted to big endian despite big endian architectures and support becoming more and more difficult to find. There is existing embedded hardware using big endian that is expensive to change as the 68k/ColdFire, SuperH, MIPS and PPC were popular in the embedded market. SuperH deserved to die because Thumb-2 was similar but better, MIPS deserved to die because RISC-V is similar but better, PPC deserved to die because AArch64 is similar but PPC was different from the 68k which was even recognized with the ColdFire by scaling the 68k down further than PPC could go. ColdFire still had competitive performance and efficiency but all the planning was to scale it down below PPC so there was no plan for 64-bit support or anything advanced. The 68060 was locked in the basement with no clock rating increases because the in-order CPU was competing with shallow pipeline PPC OoO CPUs like the PPC 601 and PPC 603 that were difficult to clock up. Instead of taking advantage of the 68060 deep pipeline to clock it up and the code density to save caches, the PPC 601+ and PPC 603e were given expensive die shrinks and doubled caches. Motorola/Freescale marketing did everything they could to promote the PPC over the 68k yet the 68k remained popular for embedded use. The competition didn't kill the 68k, Motorola/Freescale did. The competition didn't kill the 68k Amiga either, CBM did. It was an inside job of financial suicide.

agami Quote:

With the right level of software investment in AROS ABI v1 over the next couple of years, it should be able to yield a better UX on ARM64 than AmigaOS 4 UX on PPC. Then it becomes easier to support Raspberry Pi hardware for those who wish to keep things under $500. Because PPC is dead.

AROS x86-64 has a better chance at the desktop than AROS ARM will anytime soon. At least it is possible to boot Windows and play games with x86-64 hardware. AROS would be more interesting for embedded use on ARM but with ARM moving to mostly AArch64, I'd rather have 68k AROS with smaller footprint 68k hardware. All we need is affordable competitive 68k hardware and it solves so many problems.

Last edited by matthey on 06-Nov-2024 at 07:13 PM.

Quote:

agami wrote: The ARM workstations are coming:

Some years ago you could only get desktop ARM motherboards, below the performance of a RasPi 1 but above the price of an AmigeOne.

This has changed a lot, I can get RasPis between 50-110 €, desktop systems or Laptops in the price range of an WintelPC.

So if someone ask me today about an Amiga successor that shouldn't be a standard PC, I would suggest an ARM system, there are some around.

Last edited by OneTimer1 on 06-Nov-2024 at 09:01 PM.

@matthey
Where there is Nvidia, AMD will soon follow.

Quote:

matthey wrote: From Ampere Computing?

From multiple OEMs.

Quote:

AROS x86-64 has a better chance at the desktop than AROS ARM will anytime soon.

Oh absolutely.

I was just saying that for those who see x64 as PC, and PC as bad, then with the right (appropriate) amount of investment in AROS ABI v1 on ARM64 those people could have an actual NG Amiga, which is not a PC.

Personally, I'd welcome either.
I wonder which way MorphOS is leaning. They did share a while back that they're moving toward x64, but since then I've heard whispers about a pivot to ARM64.

¿Por qué no los dos?

_________________
All the way, with 68k

agami Quote:

Oh absolutely. I was just saying that for those who see x64 as PC, and PC as bad, then with the right (appropriate) amount of investment in AROS ABI v1 on ARM64 those people could have an actual NG Amiga, which is not a PC.

The x86-64 ISA has more baggage and is more bloated than ever before but I don't think there are as many "Intel Outside" people anymore. The OS is important too and x86-64 pretty well has the desktop market locked up with Windows and the best support for Linux. A new OS has to match features, bring libraries of software and somehow come up with some special advantage or feature. Haiku is one of the better non-Unix heritage alternative OSs for x86-64 and more competitive than AROS x86-64 yet is in no danger of cracking 1% of the desktop market. There may be more Amiga users than Haiku users but certainly not more AROS x86-64 users. It would be easier to compete on ARM64 desktop hardware but ARM desktop hardware is not doing a good job of competing with x86-64 desktop hardware. The way ARM64 has survived so far is to scale down below where x86-64 will scale like RPi hardware. SiFive is doing the same thing. They call it hobby and embedded but some of it is what I am calling micro-desktop at the high end. There is still a lack of software and games. Most games are ports or emulated. A 68k Amiga could bring many native retro 68k games and some native retro productivity software as well. Retro nostalgia is a pretty good marketing tool. Retro games are a good draw. The AmigaOS is weak on desktop features compared to Linux so either it needs to be scaled down further than Linux is comfortable to scale or leverage the advantages like retro games. The 68k AmigaOS scaled down a lot further than the PPC AmigaOS and can scale lower than ARM64 AmigaOS or x86-64 AmigaOS. There still are some ISA advantages and strengths.

68k & Thumb(-2): watch how fast the tiny programs load on the cheap and tiny footprint hardware
ARM64: we make it simple, easy and standard for an affordable price
RISC-V: we have open and transparent hardware and software so we don't depend on anyone else
x86-64: standard, compatible and high performance is our game
POWER: security, efficient threading, high end features, compatibility and service for the win

The 68k with very good code density and the lightweight AmigaOS has synergies that the PPC AmigaOS does not. I don't even know what the ISA advantages of PPC was as it was mediocre in more things than it was good at.

agami Quote:

Personally, I'd welcome either. I wonder which way MorphOS is leaning. They did share a while back that they're moving toward x64, but since then I've heard whispers about a pivot to ARM64. ¿Por qué no los dos?

I expect MorphOS would be more competitive on ARM64. They need to add at least 64-bit support and SMP support to have a chance on either architecture. Why not both? Less standard, more to support, executables not compatible, etc. The standard 68k Amiga ISA and chipsets has been an advantage over Linux. The standard x86-64 ISA and BIOS is one of the reasons why Linux is best on x86-64. ARM being at a disadvantage with their a la carte cores is one of the reasons why they standardized on ARM64.