@Hypex

Quote:

Hypex wrote: @Kronos Quote: "Amiga people" 2024 != "Amiga people" 1994 Except for some newbies, they'd be the same people, just older.

Nope they area small non representative subset 1994 Amiga people.

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

@matthey

Quote:

Certainly on lower end hardware, virtual addresses give an overall performance decrease due to TLB misses which also commonly have longer penalties. Jitter (worst case latency) is increased on all virtual address using hardware using TLBs. A MMU can likely increase performance in some cases but in most cases I expect overall performance to be better without virtual addressing. PPC CPUs were good at providing MMUs with virtual address support but this may have reduced their ability to scale down to compete with ARM/Thumb. The idea was to use higher level languages and more advanced OSs like Linux since PPC is difficult to hand code at a low level and has too fat of code to scale low anyway. Using pointers directly instead of handles is a performance vs security/stability trade off. Not catching bad accesses with the MMU has a good chance of leading to more bad accesses and likely a crash. I think most Amiga users would like to have the stability increase of a MMU but there are times when the performance and compatibility hit may be too much and it would be nice to be able to turn it off. I believe optional MMU support is possible with a reboot required to turn the MMU on/off.

On 68K MMU tends to scale with CPU. The lower end CPUs usually lack it. And it's only on 68040 and above where most systems would have MMU software installed. The pro Amiga person. My DKB1240 (confusing name) has a 68030 without MMU. I chose the faster 40Mhz variant as I didn't see how an MMU would benefit, but a faster clock would. A trade off. I agree the MMU would help with stability. And on faster CPU would have least impact on performance.

Quote:

ThoR's mmu.library can provide the same support and programs should exit gracefully if MMU support is not available. As important as 68k emulation is to AmigaOS 4, another method would be needed when MMU support is not available.

The Thor MMU libraries were like some kind of de facto MMU API. But why wasn't it included with OS3.9 as standard? Other Thor software was IIRC.

Quote:

The most likely reason a MMU would not be available in AmigaOS 4 would be because PPC code is being emulated meaning there would be an emulator using emulation which is cringe worthy. Two decades after the 68k to PPC transition, Macs had not only dropped built-in emulation of 68k code but built-in emulation of PPC code after transitioning to x86(-64). If 68k support is still so important, isn't it time to transition back to the 68k to at least solve the emulation in emulation issue? How much more absurd can Amiga Neverland get?

There just hasn't been enough CPUs in the Amiga lineage. Right now ARM should be an established co-processor but ARM still lacks the support PPC has. There is still no PowerARM or WarpARM kernesls allowing programs to execute native ARM subroutines. AmigaOS hangs on 68K compatibility because we never got over it. On Mac, it helped that Apple survived and redefine what Mac was to be, or I think Mac would have ended up in limbo like Amiga did. Mac transitioned from 68K to PPC, then Mac OS to OSX, then PPC to x86, to now macOS on x64 and ARM64. So whatever apps or games people liked on 68K, they could move on to the next generation, and dropped support in OSX for classic software meant users were forced to adapt. Or if not, keep running old machines, or give up totally and convert to Windows!

Around 1995, Mac was going PPC. Amiga was still going on 68K. By 2000 Mac was on PPC. Amiga had limited support for PPC and OS still on 68K. By 2005, Mac was depreciating PPC. While Amiga had no new Amiga machines the last decade. Then this AmigaOS replacement is released for those with exclusive PPC cards. Around the same time this AmigaOne is released, and so much time had lapsed, that it was barely a shadow of what inspired it and as a bare board. AmigaOS4 comes out 20 years after AmigaOS1.x, a full two decades have passed. It needs that 20 year 68K compatibility, or it's dead in the water! 30 and 40 years later, some three and four decades later, AmigaOS still needs 68K compatibility. Biggest software and games library it has!

Quote:

There are newly released programs still using the 68k IEEE math libraries as it simplifies support if there is light floating point use. The 68k VBCC compiler is one of them although the source code is available to compile with direct FPU support and it is relatively easy due to few dependencies. I would not recommend compiling Quake with the IEEE libraries but it should have less overhead with a FPU than the A1222 executing standard PPC code with FPU instructions. I wouldn't be surprised if it is less buggy too. Of course both will be a slideshow.

I've read comments from time to time about 68k IEEE math libraries. Perhaps Commodore should have kept them up to date? I even read some comment years ago from someone who wanted to port some software to OS4 but wanted to keep using the math libraries. Well you can, but then you using an old 68K library from native code, and any modern compiler gives you FPU computed floating point math for free. They seemed upset that OS4 didn't use those 68k IEEE math libraries on PPC like they were used on 68K, and were totally put off by it. I thought this was rather strange, not even common 68K software back in the 90's used it any more. But this just demonstrates how entrenched 68K is in the Amiga. The fact is it can never let it go. Look at AROS, a portable PC compatible AmigaOS reimplementation, source compatible to move Amiga onto commodity hardware. What happens? They end up back porting it to 68K again!

Quote:

BookE(mbedded). I doubt it is possible to sue someone for bad advice unless there is a safety or qualification problem. The leadership should take the blame.

Probably. But there is also opportunity. For example, Friedens could look into it, then report back that it has some slight incompatibilities. But they can be worked around in software, in the OS. The project starts, everybody gets to work, time passes. Oh, here's our fee!

Quote:

The time frame was 1995 so I doubt it would have been a 68060 yet. The 68k was still in production and even number one for 32 bit embedded use long after it disappeared from the desktop.

So still in plentiful supply. But it makes me wonder, why did Apple move away from 68K? Why didn't Apple produce Mac on 68060? A logical step up.

Quote:

They still had a plan to create a single chip 68k Amiga SoC at 57MHz with the diagram targeting early 1995. This would have been through a license for 68k CPUs. It could have been a 68EC020 or 68EC030 core even though the clock speed is given as 57MHz because newer silicon could be used and better integration allows to increase the clock speed of the system due to shorter distances for the electricity to travel. It could have been a 68040V core but this was likely too new and expensive. It could have been a custom core as changing things like cache sizes could be done. Sony customized the caches in the PS1 MIPS CPU for example. Surprisingly, from the article above and thinking of the success of the PS1, it wasn't MIPS that finally surpassed the 68k for 32 bit embedded use but SuperH and then ARM. Hitachi's SuperH had good code density with tech borrowed from the 68k while they were a 2nd source 68k producer. They licensed the tech to ARM who created the good code density Thumb and Thumb2 ISAs allowing them to become number one in embedded in combination with their prolific licensing. Motorola was not much better than CBM at licensing, especially after their lawsuits with Hitachi.

There were talks of an Amiga SoC for years. Even a PCI Amiga chipset that could be used as a PC graphics card. Seems funny now. But this could have been a foot in the door. Of course, had they even got AGA on a PCI card, VGA after 1987 would have mounted pressure on it. But even EGA had planar optimisations the Amiga didn't. So if they made it happen around 1990 it would have needed AAA really. By comparison, the ARM in the Archimedes, was produced in a SoC format with less chips needed.

Quote:

The FPGA development is how CPU and chip development is done. The design could be turned into a low production cost ASIC (mass produced 1-2GHz 68k Amiga SoC for maybe $1 USD/chip). It is obvious that the value is not competitive in FPGA which is why I tried to convince Gunnar to plan for an ASIC including ISAs. I even tried to find the help to make it possible but Gunnar is oblivious, uncompromising and unprofessional which is not true of Thomas and Jens in my experience. I doubt anyone outside of the Amiga community would be interested enough in the SoC as is to help fund an ASIC and it is unlikely on his terms. It's too bad. SAGA looks pretty good and the 68k CPU has good performance and compatibility despite weirdness. Embedded market customers aren't going to want all the low utilization non-orthogonal registers and if someone wants a SIMD unit they are unlikely to want a 64 bit SIMD unit with no floating point support that can't easily be upgraded.

It was my understanding producing a 68080 ASIC was an intention of Gunnar. He always says he has worked on PPC and his design in FPGA form even has a faster memory controller than a PPC chip. It looked to me that the speed was relying on the host chip memory controller. Which I read was DDR3. Yes that would beat DRAM in an AmigaOne XE and beat DDR2 in an X1000. The Cyclone 5 of the 68080 also features an ARM core. At that point, you can further accelerate it, and even make a PiStorm killer.

Quote:

SAGA more closely resembles the AA+ chipset which was more practical than AAA. Some of the features are common between AAA and AA+. Chunky modes, multiple playfields and display layers, and audio upgrades are the biggest upgrades. There are minor upgrades like doubling the number of sprites and adding some other features that were sorely missing in the Amiga chipset. Some things are just natural extensions like increasing the amount of chip memory. There are at least 3 AGA compatible FPGA re-implementations so there are plenty of choices. It would be nice if other developers were able to provide input and create standards but Gunnar likes development closed so he can do what he wants. I miss the Natami days with more open development instead of the Cult of the Vampire with a single dictator.

Looks like it does. What is strange is the half chunky. No one thinks of VGA chunky as half chunky because it's only 8 bits. Bit it seems to parallel Atari as well where chunky is 16 bit. I don't know why but they add 16 but full colour, instead of expanding it to packed modes modes up to 8 bit, with or without 16 bit. I would have expected it to match AAA. After all that had 20 years to catch up with the AAA design. Now some things are not practical like 16 bitplanes so don't know where those crazy ideas come from. But 24 bit hybrid, or RGB planar (like YUV planar I expect) are practical but now days obsolete with enough bandwidth.

Given the history of Natami, which I liked the look of, I would say the Vampire is the spiritual successor of the Natami.

Hypex Quote:

On 68K MMU tends to scale with CPU. The lower end CPUs usually lack it. And it's only on 68040 and above where most systems would have MMU software installed. The pro Amiga person. My DKB1240 (confusing name) has a 68030 without MMU. I chose the faster 40Mhz variant as I didn't see how an MMU would benefit, but a faster clock would. A trade off. I agree the MMU would help with stability. And on faster CPU would have least impact on performance.

The 68k MMU changed often and wasn't really worthwhile until the 68030.

68000 - no MMU practical
68010 - 68451 MMU is slow and limited (barely usable)
68020 - 68851 MMU is usable but incompatible with 68451
68030 - MMU on chip but required changes again
68040 - MMU on chip but significant changes were made
68060 - MMU on chip and mostly compatible with 68040 MMU

There was 68k Mac OS software which used the 68851 MMU and later including virtual memory software which was incorporated into System 7. ThoR's mmu.library abstracts MMU specifics to support at least back to the 68030. The CBM A2620 card could use a 68851 but I don't know if it is supported and is probably not worthwhile.

Hypex Quote:

The Thor MMU libraries were like some kind of de facto MMU API. But why wasn't it included with OS3.9 as standard? Other Thor software was IIRC.

I don't think ThoR's MMU libraries are integrated into Hyperion's versions of the 68k AmigaOS either even though they are sometimes recommended in documentation. The libraries are much improved since the AmigaOS 3.9 days. I'm not even sure ThoR was actively developing back then. I sent him an e-mail bug report for the Mu 68060.library and he sent an e-mail back that his Amiga 2000 broke years ago and that he doubted such a bug could exist. A while later, he sent me an e-mail back saying he was able to revive his Amiga 2000, confirmed the bug report and provided a fix for me to test. He has been more active since with many updates and has helped with development of P96 and AmigaOS as well.

Hypex Quote:

There just hasn't been enough CPUs in the Amiga lineage. Right now ARM should be an established co-processor but ARM still lacks the support PPC has. There is still no PowerARM or WarpARM kernels allowing programs to execute native ARM subroutines. AmigaOS hangs on 68K compatibility because we never got over it. On Mac, it helped that Apple survived and redefine what Mac was to be, or I think Mac would have ended up in limbo like Amiga did. Mac transitioned from 68K to PPC, then Mac OS to OSX, then PPC to x86, to now macOS on x64 and ARM64. So whatever apps or games people liked on 68K, they could move on to the next generation, and dropped support in OSX for classic software meant users were forced to adapt. Or if not, keep running old machines, or give up totally and convert to Windows! Around 1995, Mac was going PPC. Amiga was still going on 68K. By 2000 Mac was on PPC. Amiga had limited support for PPC and OS still on 68K. By 2005, Mac was depreciating PPC. While Amiga had no new Amiga machines the last decade. Then this AmigaOS replacement is released for those with exclusive PPC cards. Around the same time this AmigaOne is released, and so much time had lapsed, that it was barely a shadow of what inspired it and as a bare board. AmigaOS4 comes out 20 years after AmigaOS1.x, a full two decades have passed. It needs that 20 year 68K compatibility, or it's dead in the water! 30 and 40 years later, some three and four decades later, AmigaOS still needs 68K compatibility. Biggest software and games library it has!

The Apple Mac leaving PPC for x86 was devastating for PPC and the recent arrival of Amiga1 targeting the desktop. PPC desktop CPUs were disappointing enough before the embedded CPU and castrations began. All the bloated RISC architectures are gone but Trevor still tries to put a square peg through a round hole. The surviving ISAs have good code density which is what the 68k Amiga started with and one of the friendliest and most liked ISAs ever. Drop "PPC" and "desktop" to go "68k" and "retro/hobby" would give the Amiga a chance at least.

Hypex Quote:

I've read comments from time to time about 68k IEEE math libraries. Perhaps Commodore should have kept them up to date? I even read some comment years ago from someone who wanted to port some software to OS4 but wanted to keep using the math libraries. Well you can, but then you using an old 68K library from native code, and any modern compiler gives you FPU computed floating point math for free. They seemed upset that OS4 didn't use those 68k IEEE math libraries on PPC like they were used on 68K, and were totally put off by it. I thought this was rather strange, not even common 68K software back in the 90's used it any more. But this just demonstrates how entrenched 68K is in the Amiga. The fact is it can never let it go. Look at AROS, a portable PC compatible AmigaOS reimplementation, source compatible to move Amiga onto commodity hardware. What happens? They end up back porting it to 68K again!

There is nothing wrong with the decision to drop the IEEE libraries for PPC and use a standard FPU instead (the 68k needs compatibility and there isn't a standard enough FPU on the ancient hardware). There is something wrong with the decision to use hardware that removes the standard FPU though.

Hypex Quote:

So still in plentiful supply. But it makes me wonder, why did Apple move away from 68K? Why didn't Apple produce Mac on 68060? A logical step up.

I expect Apple approached IBM for the following reasons.

1. Apple wanted a 2nd source high end CPU producer and competitor which was lacking for the 68k
2. RISC was hyped and quickly gaining in performance in the late 1980s and early 1990s
3. the 68k seemed to be falling behind with the late and too hot to clock up 68040

Apple and IBM approached Motorola about creating the AIM Alliance. Motorola could have said no but they would have been in danger of losing the Apple CPU market. They could have tried to negotiate with Apple other ways of improving 68k development and sources but the situation had already escalated with Apple approaching IBM first. The AIM Alliance was formed in 1991 before much would have been known about the 68060. Ironically, the shallow pipelines PPC designs and large PPC caches limited PPC performance and disappointed Apple while the 68060 was one of the deepest pipelined CPUs of the time and had good performance with small caches meaning it had more potential to clock up. The 68060 was also a low power and cool running chip that could be used for a laptop. Not only did Motorola sabotage the 68060 by not clocking it up but Apple 68k Mac OS upgrades became less compatible with the 68060 than earlier ones. Motorola and Apple wanted higher performance and more competitive 68k CPUs but once the 68060 arrived they wanted it downgraded (to embedded) and to disappear from the desktop.

Hypex Quote:

There were talks of an Amiga SoC for years. Even a PCI Amiga chipset that could be used as a PC graphics card. Seems funny now. But this could have been a foot in the door. Of course, had they even got AGA on a PCI card, VGA after 1987 would have mounted pressure on it. But even EGA had planar optimisations the Amiga didn't. So if they made it happen around 1990 it would have needed AAA really. By comparison, the ARM in the Archimedes, was produced in a SoC format with less chips needed.

There would have been more business opportunities if CBM had upgraded the Amiga chipset faster, integrated the Amiga chipset into one chip and converted it to CMOS sooner. It could have been the 2D chipset for 3dfx 3D Voodoo cards, set top boxes, kiosks and NTSC/PAL genlock video boxes. CBM did a poor job of proliferating through licensing, as did Motorola. CBM was still semi-successful in the embedded market by providing competitive and standard computers with the CD32 SBC nearly becoming like standard embedded RPi SBC hardware. The ARM based Acorn Archimedes used the opposite philosophy of the original Amiga as the chipset was simplified and the CPU used for everything. This allowed for fewer chips sooner but the Amiga chipset is small and could have easily been integrated into one chip in the 1980s, likely even with AGA/AA+ upgrades. Even the 32 bit version of AAA would have been double the size of AGA/AA+ or more and required more chips. AGA/AA+ was more practical. The CBM bankruptcy docs still shows AAA before Hombre but then AAA disappears while the 68k Amiga SoC using AA+ remains. I suspect AAA would have disappeared even if CBM was on track with their development although certain features were borrowed for AGA/AA+ so development was not a total loss.

Hypex Quote:

It was my understanding producing a 68080 ASIC was an intention of Gunnar. He always says he has worked on PPC and his design in FPGA form even has a faster memory controller than a PPC chip. It looked to me that the speed was relying on the host chip memory controller. Which I read was DDR3. Yes that would beat DRAM in an AmigaOne XE and beat DDR2 in an X1000. The Cyclone 5 of the 68080 also features an ARM core. At that point, you can further accelerate it, and even make a PiStorm killer.

As I recall, Gunnar has said the plan is to eventually make an ASIC but it doesn't look to me like he is planning toward it. Maybe he thinks a cheap ASIC with the decisions he has made would be adequate. The memory controller is the FPGA memory controller and even affordable FPGAs use fairly modern silicon provided by economies of scale. An ASIC may not be able to use the same memory controller or other hard constructs like multipliers and SRAM blocks if the HDL code for them is not available. The process size would likely change and some rework would likely be necessary. I'm not so sure the Cyclone V used has an accessible ARM CPU even though an earlier dev board that was considered did. The FPGA could always be upgraded to a FPGA with an ARM CPU to stay competitive with the PiStorm but this is counter to the logic of moving toward an ASIC. A larger more expensive FPGA would make more sense to both improve performance and move toward an ASIC but this would further reduce sales for Amiga use, at least until the ASIC is ready.

Hypex Quote:

Looks like it does. What is strange is the half chunky. No one thinks of VGA chunky as half chunky because it's only 8 bits. Bit it seems to parallel Atari as well where chunky is 16 bit. I don't know why but they add 16 bit full colour, instead of expanding it to packed modes modes up to 8 bit, with or without 16 bit. I would have expected it to match AAA. After all that had 20 years to catch up with the AAA design. Now some things are not practical like 16 bitplanes so don't know where those crazy ideas come from. But 24 bit hybrid, or RGB planar (like YUV planar I expect) are practical but now days obsolete with enough bandwidth.

True chunky has to contain all the palette data and there isn't much room with 8 bits. The result is a course color selection that is poor for shading.

https://en.wikipedia.org/wiki/List_of_monochrome_and_RGB_color_formats#8-bit_RGB

R3G3B2 (8 bit chunky)
R5G6B5 (16 bit chunky)
R5G5B5 (16 bit chunky, potentially with alpha/transparency bit)
R8G8B8 (24 bit chunky)
A8R8G8B8 (32 bit chunky with alpha)

The Amiga OCS/ECS uses a 12 bit palette (R4G4B4) and AGA 24 bit palette (R8G8B8).

https://en.wikipedia.org/wiki/List_of_16-bit_computer_color_palettes#Amiga_OCS

Chunky/packed with CLUT gives a palette selection not limited by the number of chunky bits per pixel. Once the resources are available for 16 bit chunky, true 8 bit chunky looks bad in comparison and 8 bit chunky/packed with CLUT is limited to 256 colors or the number of color registers if it is less. The only reason to use 8 bit chunky/packed is to save bandwidth and memory but there is a large loss in quality. There is not nearly as much loss in downgrading from 32 or 24 bit chunky to 16 bit chunky to save bandwidth and memory which can look good with dithering and may even look better at a higher resolution.

Hypex Quote:

Given the history of Natami, which I liked the look of, I would say the Vampire is the spiritual successor of the Natami.

Successor does not always mean spiritual successor even if some of the same people remain. Was CBM the spiritual successor of Amiga Corporation?

@Kronos

x86 and arm followers failde to provide something like mac os x
amiga gui and graphics on unix base

no reason to use aros x86 or emu68
when you may just use windows or android on exact the some hardware
so yes intel outside amiga is interesting as long as it is not a pc
as long as it is not a computer with x86 and arm

@ppcamiga1

Multi-core 64-bit PowerPC followers failed to provide something like Mac OS X or Linux, a modern Amiga OS that can run the latest development frameworks, modern secure and threaded apps, and latest games.

No reason to use AmigaOS 4 or MorphOS, when you can just use a modern web browser on Linux on the exact same hardware, or even cheaper x64 or ARM64 hardware.
The special thing about Amiga was the OS and chipset. All other operating systems have caught up and surpassed Amiga OS, and the chipsets have been superseded by modern GPUs, and APUs.
So "NG" Amiga is now just another PC, but not a very good one.

_________________
All the way, with 68k

@matthey

Quote:

Apple and IBM approached Motorola about creating the AIM Alliance. Motorola could have said no but they would have been in danger of losing the Apple CPU market. They could have tried to negotiate with Apple other ways of improving 68k development and sources but the situation had already escalated with Apple approaching IBM first. The AIM Alliance was formed in 1991 before much would have been known about the 68060. Ironically, the shallow pipelines PPC designs and large PPC caches limited PPC performance and disappointed Apple while the 68060 was one of the deepest pipelined CPUs of the time and had good performance with small caches meaning it had more potential to clock up. The 68060 was also a low power and cool running chip that could be used for a laptop. Not only did Motorola sabotage the 68060 by not clocking it up but Apple 68k Mac OS upgrades became less compatible with the 68060 than earlier ones. Motorola and Apple wanted higher performance and more competitive 68k CPUs but once the 68060 arrived they wanted it downgraded (to embedded) and to disappear from the desktop.

68060's FPU is not pipelined. Clock speed attainment is based on the slowest component.

68060's 4 byte fetch per cycle from L1 cache is a step backward from 68040's 8 byte fetch per cycle from L1 cache.

PowerPC 604 has a six-stage pipeline. PowerPC did NOT remain static with their pipeline depth.

P5 Pentium's FPU pipeline consists of eight stages. The "5-stage" pipeline refers to U and V integer pipelines.

Pentium MMX's integer pipelines have increased to six stages.

DEC Alpha 21064 pipeline depth,
integer pipeline = 7 stages
FP pipeline = 10 stages.

Pentium Pro has 14 stage pipeline in 1995.

During the tipple digit Mhz race, the PowerPC camp was largely keeping pace until X86's Ghz race. AMD and Intel increased the Mhz race pace during AMD K6/K7 vs Intel Pentium II/III. AMD had an engineering skills injection from ex-DEC Alpha engineers when Compaq-backed NexGen was purchased.

Last edited by Hammer on 05-Jul-2024 at 03:23 AM.
Last edited by Hammer on 05-Jul-2024 at 03:19 AM.
Last edited by Hammer on 05-Jul-2024 at 03:18 AM.
Last edited by Hammer on 05-Jul-2024 at 02:44 AM.
Last edited by Hammer on 05-Jul-2024 at 02:39 AM.
Last edited by Hammer on 05-Jul-2024 at 02:39 AM.

_________________
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 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@Hypex

Quote:

There just hasn't been enough CPUs in the Amiga lineage. Right now ARM should be an established co-processor but ARM still lacks the support PPC has. There is still no PowerARM or WarpARM kernesls allowing programs to execute native ARM subroutines.

Refer to the A314 project.

Emu68 accelerates the entire 68K CPU workload.

Quote:

AmigaOS hangs on 68K compatibility because we never got over it. On Mac, it helped that Apple survived and redefine what Mac was to be, or I think Mac would have ended up in limbo like Amiga did. Mac transitioned from 68K to PPC, then Mac OS to OSX, then PPC to x86, to now macOS on x64 and ARM64.

Amiga is not Mac.

3rd party Mac software is not allowed to "kick the OS". Userland 68K emulator is suitable for Mac's 68K legacy.

Amiga's hit-the-metal and "kick the OS" 68K games resemble PC's i386-protected mode/hit-the-metal VGA DOS games.

For the healthy PC platform, it took about 6 years for the Windows games library to be built up to displace its DOS counterparts. PC's Ghz CPU power can brute force DOSBox during the Windows 2000/XP era.

The very thin hypervisor Emu68 fits Amiga's hit-the-metal and "kick the OS" 68K games.

AmigaNG PowerPC camp thinks Amiga is like a Mac. This is doubled down by Amithlon's Bernd "who cares about games" Meyer and the "AmigaOS X86" announcement by Bill McEwen.

Last edited by Hammer on 05-Jul-2024 at 03:10 AM.

_________________
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 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@kolla

Quote:

Windows NT ran on just about all relevant CPU archs, the list of contemporary archs for which NT 4.0 didn’t exist or for which there never was an effort for a port, is rather short. The major obstacle for any non-x86 WinNT was updates and native applications, and in case of any big-endian WinNT, exchange of binary data/files with little-endian systems.

Windows NT on non-X86 uses little-endian mode. Xbox 360 is big-endian.

Windows NT 4.0 PowerPC is a little-endian.
Windows NT 4.0 Alpha is a little-endian. Has FX32 to run IA-32 Windows apps.
Windows NT 4.0 MIPS is a little-endian.
Windows NT 5.0 Beta 3 Alpha is a little-endian. Has FX32 translator to run IA-32 Windows apps.
Windows NT 5.2 Itanium is a little-endian. Has JIT IA-32 translator to run IA-32 Windows apps.

Windows NT 10.0 ARM is a little-endian. Has JIT IA-32 translator to run IA-32/X86-64v2 Windows apps. Missing X86-64v3 AVX support. Intel will sue corporations without a valid AVX license. Intel specifically targeted Qualcomm in a potential large-scale legal battle.

During the Windows NT 4.0 era, the PC's hit-the-metal DOS games library is a major factor, hence the Windows 95/98 transition phase.

AMD sued Qualcomm on GpGPU patent breaches and Qualcomm's Xenos patent license is not enough.

Last edited by Hammer on 05-Jul-2024 at 03:43 AM.

_________________
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 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@matthey

Quote:

https://websrv.cecs.uci.edu/~papers/mpr/MPR/19980126/120102.pdf Motorola Stays on Top; Other Players Double Worldwide volume in 32-bit embedded microprocessors surpassed 180 million units, as Figure 1 shows. Of that total, three architectures—68K, MIPS, and SuperH—accounted for 80% of shipments in 1997. All the top vendors maintained their relative positions, although in some cases the gaps between players narrowed. ARM, MIPS, and PowerPC won the biggest advances in terms of multiples. The first two more than doubled from 1996 to 1997, growing by 129% and 138%, respectively. MIPS also enjoyed the biggest unit increase, shipping 24.8 million more chips and CPU cores than it did the previous year. PowerPC is still in startup mode, multiplying from half a million in 1996 to about 3.9 million in 1997. Motorola’s 79.3 million units put it on top, as usual. Its 68K line has been the embedded 32-bit volume leader since it created the category. As the figure shows, sales of 68K chips were about equal to worldwide sales of PCs. Taken together, that’s one new 32-bit microprocessor for every man, woman, and child living in the United States.

Your source publication year is 1998 and it's meaningless without a 68K model breakdown.


From your source
Quote:

Motorola’s 79.3 million units put it on top, as usual. Its 68K line has been the embedded 32-bit volume leader since it created the category. As the figure shows, sales of 68K chips were about equal to worldwide sales of PCs. Taken together, that’s one new 32-bit microprocessor for every man, woman, and child living in the United States. The company declined to break out its sales among 68K families (680x0, 683xx, and ColdFire) but suggested that ColdFire played a much more important role than in previous years. Given that ColdFire accounted for only about 1 million units in 1996, that trend is unsurprising.

Against your 1998 source's PC numbers.

http://edition.cnn.com/TECH/computing/9902/01/terrific.idg/index.html
Quote:

Title: '98 was 'terrific' year for PCs, Dataquest says Total global shipments for PCs hit 92.9 million last year.

Last edited by Hammer on 05-Jul-2024 at 04:14 AM.
Last edited by Hammer on 05-Jul-2024 at 04:13 AM.

_________________
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 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@agami

Quote:

Because, value. Putting the SA V4 aside, most of the V2 and V4 SKUs are packaged as accelerator upgrades for existing Amiga's. In a market where such upgrades are rare and expensive, be they 68k or PPC, the Apollo boards represent excellent value for enhancing the abilities of Amigas and extending their usefulness.

In that case the argument should be a CPU replacement for a PC is cheaper.

It does, but it the price is more towards the high end, so it is competing with some high end cards. However, some mid range cards are approaching a similar price, and it would offer more value over all.

The SA isn't far off and is a cleaner approach. But the SA, despite the hype of being the next Amiga, does suffer the same flaws as an AmigaOne. You can't plug an Amiga floppy into it or any floppy, it doesn't come with an Amiga keyboard or mouse, and it has no Zorro slots to plug a Toaster into! I recall that last one cost the AmigaOne XE.

Quote:

When I think about what I spent on BlizzardPPC + Bvision, the Apollo IceDrake V4 is an absolute bargain. Plus I don't have to put the A1200 in a tower.

Strangely, I wish I had purchased a BlizarrdPPC. I'll never forget when they were selling them locally for $800, that's why. Don't know where but I recall some Amiga shop in Geelong I never visited. I'd already spent $450 on a DKB1240 with 8MB RAM or something then $200+ something in Ferret SCSI. But, the Blizzard, especially with a BVision, really needed a proper tower. It wasn't suitable for a desktop hack. Suppose I'm like a Mac guy that likes it all neatly integrated. But I just had my SCSI tower, so I would have needed to buy a proper A1200 tower, which would amount to $300 or something for a plastic case to house my A1200 mainboard. Setting up CDROM already cost me $800 the first time for the other case.

By the end of retiring my A1200 from full time use, I had spent $2,500 on the full system. All I was left with was a 40Mhz 030 with 32MB RAM, 1GB SCSI HDD, CDROM, external CDRW and an old monitor with only AGA and no practical RTG. Funny enough, I spent less on my AmigaOne XE, only $2,000. And it was way better to use from the get go. Of course, I didn't build it up over time, so initial outlay was more.

Quote:

Whereas with A-EON's AmigaOS 4 (PPC) hardware, it's not the sticker price itself that gives me pause. It's the overall lack of value. Their products never made sense to me, which is also why I never thought that any attack on PPC hardware since 2010 has ever been senseless.

The sticker price scares me. It's closer to POWER9 servers, well the X5000 is, and even though they aren't cheap it's embarrassingly way lower power by comparison. Around 2010 was when the X1000 was being conceived. Although I bought mine at a beta board price, I think the X1000 was the last board to offer reasonable enough power for the price. Before, a few years prior, after seeing a new family PC a dual core 1.8Ghz CPU was exciting. When the X1000 arrived I thought a 64 bit dual core at 1.8 Ghz was a bit average. Ubuntu thinks my CPU is 3 core, teaser. I forget what it was called, but the X1000 hardware was almost exactly like an earlier board announced down to the specs. So someone must have take that design and made something of it. Unless the earlier design was based on the Electra board which the X1000 is.

I was talking to a friend, who also bought into the X1000, but later lost interest and now considers OS4 to be too unstable and buggy like a lot of people do. His issue isn't the lack of power or what price it is for the offered power, he has no issue with a lower powered board, but his issue is how useful it is and without standard working apps like a browser it becomes useless.

@Hypex

Quote:

Hypex wrote: @agami Quote: Because, value. ... Apollo boards represent excellent value for enhancing the abilities of Amigas and extending their usefulness. In that case the argument should be a CPU replacement for a PC is cheaper.

Cheaper? Yes. But less value because a drop in CPU upgrade for the x64 PC has a small window of opportunity. Most PCs which are 3 years old, are incompatible with the current offerings in the intel or AMD ecosystem.

There are some higher value scenarios where a person first purchased a 4-core or 6-core CPU 3 years ago, and now can drop in a CPU with 2x, 3x, or even 4x number of cores. Assuming they purchased the right kind of motherboard back then.

While gaming performance is only marginally increased by additional cores, there are non-insignificant productivity gains. Also, most drop in CPU upgrades only extend the usefulness of a PC by 2-3 years. I believe the average number of years between a complete gutting or whole PC replacement is around 5-6 years.

Whereas the Apollo expansions transform Amiga's without AGA into AGA-compatible ones, and bring RTG to wedge computers (A500/A600) for which previously it was never an option. Ensuring one doesn't need a "can die any day now" CRT screen. High quality audio, networking, vast amounts of fast RAM: These lowly gaming computers can now do everything only big box Amigas could previously, or even presently do as there is no Apollo CPU card for A3000/A4000 Amigas (yet).


Quote:

I was talking to a friend, who also bought into the X1000, but later lost interest and now considers OS4 to be too unstable and buggy like a lot of people do. His issue isn't the lack of power or what price it is for the offered power, he has no issue with a lower powered board, but his issue is how useful it is and without standard working apps like a browser it becomes useless.

Yep. Your friend's value measuring stick is based on usefulness. I'm sure that just like your friend, most people who bought PPC hardware for the privilege of running AmigaOS 4 went into it with open eyes. They knew what they were getting into. Most bought into the dynamic of value increasing over time. Sure it's kind of a beta platform now, but the value will grow as software improves, and they chose to be part of that story.

But when the story ends up being what it is, when the spoken and unspoken promises of future value are broken, then what is it actually worth?

I said it before and I'll say it again: While going with PowerPC back in 2010 was justifiable, sticking with PowerPC past 2015 is highly questionable.
And peddling low-value PowerPC-based AmigaOS 4 hardware in 2024 is borderline criminal.

Last edited by agami on 13-Jul-2024 at 02:49 AM.

_________________
All the way, with 68k

@agami

Quote:

Cheaper? Yes. But less value because a drop in CPU upgrade for the x64 PC has a small window of opportunity. Most PCs which are 3 years old, are incompatible with the current offerings in the intel or AMD ecosystem.

AM4 socket was launched in 2016 and still has CPU model releases in 2024.

AM4 X370 was launched in Feb 2017.

By September 2024, the AM5 socket would be 2 years old.

Quote:

While gaming performance is only marginally increased by additional cores, there are non-insignificant productivity gains.

Wrong. Ryzen 7 5800X3D significantly boosts gaming performance.

https://youtu.be/uo4WxpwHs3s
Does the Ryzen 7 5800X3D Work With B350 & X370 Motherboards?

https://www.techpowerup.com/review/amd-ryzen-7-5800x3d/16.html
For gaming performance, Ryzen 7 5800X3D boosted the AM4 platform into Core i9 12900K levels.

Quote:

Also, most drop in CPU upgrades only extend the usefulness of a PC by 2-3 years. I believe the average number of years between a complete gutting or whole PC replacement is around 5-6 years.

AM4 's Zen 1 was limited to 8 cores, expanding into 16 cores with Ryzen 9 3950X. SIMD width doubled from Zen 1's four 128-bit units into Zen 2's four 256-bit units.

In terms of productivity, the Ryzen 9 3950X is roughly equivalent to the Ryzen 9 5900X.

Quote:

Whereas the Apollo expansions transform Amiga's without AGA into AGA-compatible ones, and bring RTG to wedge computers (A500/A600) for which previously it was never an option.

PCs have partitioned graphics design e.g. AM4 X370 supports the latest RTX 4090.

Quote:

Ensuring one doesn't need a "can die any day now" CRT screen. High quality audio, networking, vast amounts of fast RAM: These lowly gaming computers can now do everything only big box Amigas could previously, or even presently do as there is no Apollo CPU card for A3000/A4000 Amigas (yet).

Big box A3000/A4000 has a Z3660 option. https://youtu.be/MqiEQtzGk-Q?t=1

Last edited by Hammer on 13-Jul-2024 at 08:14 AM.
Last edited by Hammer on 13-Jul-2024 at 08:07 AM.
Last edited by Hammer on 13-Jul-2024 at 08:06 AM.

_________________
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 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB