@NutsAboutAmiga

Quote:

Simply put Apple was not IBM's priority, this why they switched to Intel.

This was one of the problems with PowerPC. Apple was the only major buyer. It's not heallthy if a chips survival relies on only one company using it. There should have been Amiga and Atari as second buyers. Yes I know.

Apple also killed off CHRP so any PowerPC machine that wasn't a Mac got scrapped from the market. They made it so that the only PPC machines would be a Mac. All Linux distros for years on end target Macs. It's the only known PPC machine. Apple even delayed the AmigaOne by snapping up the CPUs so the XE G4 was delayed. The G3 was an IBM. Had a Mac not been the only known PowerPC desktop machine in existence then PowerPC may have lasted longer on the desktop.

There may have been the AIM alliance but Apple called the shots like how Commodore was brought down. Apple prevented other PowerPC machines flourising. And when they left PowewrPC they put effort into killing it off on the desktop. Did Apple own the rights to PowerPC? I mean, why didn't IBM and Motorola produce other PowerPC based computers? Why let Apple control it? Yes I know the A stood for Apple and they had to come first.

Last edited by Hypex on 18-Feb-2022 at 05:46 PM.

@Hypex
Well don’t forget PowerPC also had the console market for a while as pointed out Ps3, Xbox, GameCube, wii & wii u where all powered by PPC.

Plus most supercomputer / science computer / military computer run PPC in 2000’s, It’s nice to know there a PPC rovers on mars right now.

So apple weren’t the only ones pushing PPC, I think it was just the fact that x86 had Intel / Amd fighting did make power, performance and more importantly price to hard to ignore. Plus they where clever focusing on mhz/ghz number in the battle, as naturally consumers are just going to to think the higher the mhz the faster the cpu even though that was not always the case.

arm has seen the same thing happen thanks to the mobile phone market pouring investments into and you got competition with in the market, snapdragon vs Apple A / M chips vs Exynos and just the faster turn round time of everyone expecting some kind of improvement each year to their smart phone has made the chips catch up very quickly to x86 levels.

Last edited by amigang on 18-Feb-2022 at 10:12 PM.

_________________
AmigaNG, YouTube, LeaveReality Studio

The tri-core Wii U APU and board is the closest PPC platform ever got to a RPi...

https://www.extremetech.com/gaming/137746-nintendo-tears-down-wii-u-to-show-off-single-chip-ibmamd-cpu-gpu

I always wondered how much you could clock up that APU with some active cooling in a custom case.

Sadly no one's every bothered.... It could produce better graphics than the Xbox 360 and PS3 but was limited by clock speed... :/

Since PPC's programmable clock speeds, custom firmware could have clocked it up...

NutsAboutAmiga Quote:

PowerPC was used in printer, routers, in air craft Entrainment systems, blueray / dvd players, the military, it also used MRI’s and lots more that you won’t think of as PC, the low watt usage, high efficient, and reliability made it a nitche product.

PPC had good performance/W designs at the mid to upper end of the embedded market where ARM designs were too weak in performance. This included communications and engine management up until recently. The poor code density limited market penetration into lower end embedded markets where ARM Thumb dominated, with licensed technology from SuperH and resulted in Hitachi being sued by Motorola for copying the 68k into the H8 family which they changed more to become SuperH. The 68k, then SuperH, then ARM Thumb2 architectures became the embedded market leader and they all have significantly better code density than PPC. Motorola's 68k dominance of the embedded market was lost when they dropped development of the 68k and switched to PPC as their embedded architecture flagship.

NutsAboutAmiga Quote:

PowerPC was never really focus for IBM, there interest was in Server market, the Power CPU. Freescale / Motroola as into embedded market, and Apple was in the Desktop market. They formed AIM alliance, that meant they were not competing they stayed out of etch others way, this Is what I expect drives the price up. I believe the CPUs were made in the USA, that drives price up as well. Compared to other chips made in Asia.

IBM made PPC processors for the top 3 competing consoles. They also did a significant amount of the design and production of desktop PPC processors from the 601 to the 970 (G5). They do like to have firm orders from customers for custom designs which reduces risk as they have varied from this conservative philosophy and been burnt by products like the PPC 970.

NutsAboutAmiga Quote:

The ARM chip made into mobile market, and expanded quickly there, more and more multimedia was needed and like video, music, this made more and more useable in the embedded market, ARM was produced with less expensive license, In low-cost countries. With slave labor.

ARM was rather slow to be adopted and the original ARM 32 bit ISA did not gain them the embedded market. It was the good code density Thumb encodings which allowed them to scale further into the low end low power embedded markets. Low cost products are very important for embedded markets too. ARM designs were cheap and simple. It has nothing to do with slave labor. Foundry prices are competitive. Sure, there is more and cheaper foundry and assembly infrastructure in Asia but skilled labor is required and these workers make relatively good money. China doesn't even have the top foundry technology. Taiwan and Singapore have some of the highest tech foundries and have large semiconductor industries. TSMC stands for "Taiwan" Semiconductor Manufacturing Company.

NutsAboutAmiga Quote:

When IBM made PowerPC chips they designed the Power chip first, they made cut down version of it for desktop market, they did not design the chip to be the best desktop chip. That was a poor decision. Simply put Apple was not IBM's priority, this why they switched to Intel.

The PPC601 was Time magazine's 1994 product of the year. The IBM PPC designs were often based on bigger POWER cousins which were "cut down" but the big problem was the architecture didn't scale down well. PPC designs have always been resource hungry and the low end designs have been especially poor because of poor code density. PPC earned a poor reputation in low end Macs with the 603 which was difficult to recover from. Then the IBM designed PPC 970 was a major disappointment showing PPC wasn't competitive for high end performance either. PPC was just a mediocre architecture not particularly good at anything or anywhere so it was waiting to be replaced by something better. It was in IBM's best interest for PPC to succeed as it had the most to gain from it of the AIM companies because it is similar to POWER. Motorola was supposed to be the company supporting Apple with desktop designs not IBM but Apple was the first of the three to abandon the agreement and then Motorola designs went from partial desktop/embedded to full embedded designs.

NutsAboutAmiga Quote:

The success of ARM is result of big investment from Google in Android for Smart phones, smart tv’s and Chromecast. Google is like Microsoft for Mobile phones. Between ARM and PowerPC there is big cultural difference.

There is a big cultural difference between ARM which is alive and PPC which is dead alright. If Google and Android didn't exist I expect ARM would still be alive and PPC would still be dead.

NutsAboutAmiga Quote:

ARM was unless for desktop market until the M1 from Apple, that’s pretty recent event. While Pi Pico, Pi Zero, offer nothing for the desktop market. The CPU in PI 4, is a Broadcom BCM2711 64-bit Quad-Core Cortex-A72 1.5 GHz SoC, its only slightly higher clocked then the AMCC460. of couse we don't care about cores, not support by AmigaOS4.1, nor is ARM but, kind useless in this context anywhy. https://en.wikipedia.org/wiki/ARM_Cortex-A72 As Desktop chip is pretty useless with not a lot of connectivity.

The ARM Cortex-A72 core is a powerful design. It outperforms every PPC design ever made including the PPC 970 (G5). Unlike the PPC 970, it has good power efficiency for an aggressively OoO micro-oped core. The Broadcom BCM2711 SoC is an inexpensive mobile device SoC with high CPU performance but lackluster GPU performance. It is not suited for desktop use and is a compromise in the Raspberry Pi not unlike the embedded PPC SoCs in the Sam460LE and A1222. The difference is that fat PPC had to do more chopping to compete in power efficiency and the designs are older. The Raspberry Pi Foundation is starting to design their own SoCs which they can customize exactly for their target markets, starting small which is cheaper and likely working their way up.

I wouldn't make fun of the ARM reference designs anymore. Some 3rd party designers with ARM architectural licenses who used to make their own designs now use the reference designs with a few tweaks. Even the lower power and cheaper in-order ARM Cortex-A53 cores in the Raspberry Pi 3 are competitive in performance with the best OoO PPC cores.

NutsAboutAmiga Quote:

https://www.youtube.com/watch?v=x0wl3ZaOq_g I go for M1 if I wonted desktop ARM system, BCM2711 is not design for it.

Apple is a top fabless semiconductor designer now but the Apple SoCs don't have miraculous performance. They would like everyone to believe that and have been jumping the competition in chip fab process improvements but this is nothing new. From the last comparisons I have seen using a similar chip fab process, the Apple AArch64 cores are still behind x86-64 cores in performance. It doesn't surprise me though as I know RISC means reduced performance as well. The Apple SoCs have custom hardware acceleration that makes them better performance in some cases. Custom hardware acceleration and the advantages of better integration are likely what Apple was after anyway. Well, x86-64 doesn't scale down to as low of power and AArch64 is unlikely to scale up to x86-64 performance. There isn't much to choose from in the middle although the 68k had as good or better performance with reduced power to be able to scale lower but it isn't on the menu.

NutsAboutAmiga Quote:

https://www.mobileshop.eu/apple/desktops/mac-mini-2020-256gb-8gb-ram-mgnr3/ so 2 years old mac cost about same as Sam460, only problem no PCIe slots I guess ARM is perfect for Apple as they want there customer to buy new computer, instead of upgrading the hardware.

At least a 2 year old Mac has 2 year old Apple ARM technology instead of 20 year old PPC technology. Apple's reputation is top notch and Amiga's reputation is somewhere between poor and nonexistent.

Hypex Quote:

This was one of the problems with PowerPC. Apple was the only major buyer. It's not healthy if a chips survival relies on only one company using it. There should have been Amiga and Atari as second buyers. Yes I know.

The importance of compatibility was underestimated. Apple made the transition to PPC with limited success and the Amiga transition failed miserably with only a few thousand computers sold. Motorola would have been better off staying with the 68k. The 68060 was better for resource limited computers and low enough power to be used in a laptop which were both problems for fat PPC in the early days of transition. Then there is better compatibility (did I mention compatibility already?) and all the time wasted transitioning.

Hypex Quote:

Apple also killed off CHRP so any PowerPC machine that wasn't a Mac got scrapped from the market. They made it so that the only PPC machines would be a Mac. All Linux distros for years on end target Macs. It's the only known PPC machine. Apple even delayed the AmigaOne by snapping up the CPUs so the XE G4 was delayed. The G3 was an IBM. Had a Mac not been the only known PowerPC desktop machine in existence then PowerPC may have lasted longer on the desktop.

PPC was one of the best supported targets of operating systems early. It was better supported than any other RISC architecture. The PPC systems tended to be high priced workstations more than inexpensive desktops though. Yes, it was pretty much Apple alone driving down the cost of desktop PPC processors. In a more perfect world, one of either Amiga or Atari would have survived to compete but CBM upper management was too incompetent.

Hypex Quote:

There may have been the AIM alliance but Apple called the shots like how Commodore was brought down. Apple prevented other PowerPC machines flourishing. And when they left PowewrPC they put effort into killing it off on the desktop. Did Apple own the rights to PowerPC? I mean, why didn't IBM and Motorola produce other PowerPC based computers? Why let Apple control it? Yes I know the A stood for Apple and they had to come first.

Apple was big enough that they had negotiating power but that led to lower profit margins.

Mitch Alsup Quote:

When I was at Moto, we had a saying "Apple paid for the FAB, but you made no profit on them"

https://groups.google.com/g/comp.arch/c/wzZW4Jo5tbM

IBM did produce PowerPC based computers.

https://en.wikipedia.org/wiki/IBM_ThinkPad_Power_Series

Once again high end (modestly configured 850 cost upwards of $12,000) but a good OS selection with Windows NT 3.51 and 4.0, AIX 4.1.x, and Solaris Desktop 2.5.1 PowerPC Edition. IBM, like Apple, usually prefers the higher profit margins of high end markets rather than taking market share in lower margin markets like ARM Ltd. and the Raspberry Pi Foundation. Motorola upper management at least should have done more contemplating of the PPC situation as committing so completely to a lemon architecture lost them much of the embedded market and led to their downfall.

@matthey

Quote:

The poor code density

You keeps saying the same things over and over.

If look at modern C/C++ programs large chunk of that is libc or glibc, or newlib, or stllib, on amiga (680x0) you have tiny amigalib (that only includes few std functions), this are almost always linked in by default.

On Amiga (680x0) it pretty common to compress binary files, with power packer, stone cracker or something like that, because need to get it onto a tiny floppy disk, 2 mb is nothing today, so no effort goes into making anything smaller, on system with more ram and disk space.

plus a lot debug stuff is also included by default, like function names, as required for stack trace, in case your program crash, AmigaOS3.x programs are compiled without debug symbols.

If your compiling something complicated like Web Browser, you need a up to date SDK that includes all that stuff, on classic Amiga or even the Vampire you love talking about, there not lot of memory, so you worry about running out of memory all the time. That’s not problem when you have 2GB or more memory on a Sam460 or AmigaOne computer.

The largest part of memory usage is actual graphics, fonts and all wallpaper everything else you like to have like large true color icons. A modern game needs a lot of memory for textures. When look at something like the Vampire its not at all scaled for modern games. Vampire has only 512mb of RAM and 12mb of chipram, thats tiny.

Last edited by NutsAboutAmiga on 19-Feb-2022 at 10:50 AM.
Last edited by NutsAboutAmiga on 19-Feb-2022 at 10:35 AM.
Last edited by NutsAboutAmiga on 19-Feb-2022 at 10:33 AM.
Last edited by NutsAboutAmiga on 19-Feb-2022 at 02:32 AM.
Last edited by NutsAboutAmiga on 19-Feb-2022 at 02:26 AM.
Last edited by NutsAboutAmiga on 19-Feb-2022 at 02:25 AM.

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

@Hypex

Quote:

Yes I think the A1222 CPU is more trouble than it's worth. It's not like the 68K days were an FPU was optional. It's more like a ColdFire, as it is incompatible with standard CPU codes, but can run most of the common ones. So it needs software patching to work. I think this is the result of trying to keep it cheap. It's just too cheap and not effective in the long run. There wasn't any stocks of more standard PPC chips avail for a similar price I wonder.

Hindsight is a wonderful thing but I think Trevor would have been better off putting the A1222 money towards mass producing the Sam460 to get his low cost OS4 hardware. If it was £250-£300 for a Sam460 board with an OS4 license they may well have been able to sell 10,000 between 2015 and now.

Disclaimer.

I don't know if 10,000 units would have been enough to get the price breaks needed to achive that kind of price point in 2015.

@amigang

Yes PowerPC did come into fashion for consoles as short lived as it were. It was also strange as the XBox was pretty much a PC in a box but then the 360 comes out with the custom PowerPC making it more like a Mac! As well as the PS3 with the Cell. And along with what Nintendo was doing. For one generation anyway. Then life hit back in the face. The PowerPC had come and gone out of fashion. And then they moved on to x86 as expected.

So let me ask a dumb question here (naturally due to my long absence) but is the 460EX related to say the G4/PowerPC 7450 at all? Is there any native compatibility for WarpUp? How do these SAM boards fit into the AmigaNG picture.

Basically I found this interesting and realized how out of touch I am.

The landscape seems to have shifted.....

_________________

@matthey

Quote:

The importance of compatibility was underestimated. Apple made the transition to PPC with limited success and the Amiga transition failed miserably with only a few thousand computers sold. Motorola would have been better off staying with the 68k. The 68060 was better for resource limited computers and low enough power to be used in a laptop which were both problems for fat PPC in the early days of transition. Then there is better compatibility (did I mention compatibility already?) and all the time wasted transitioning.

The PPC version of MacOS was closer to 68K MacOS than the PPC version of AmigaOS is to 68K AmigaOS. Once solid on PPC they dumped it for OSX which provided a gateway for portability. But, Apple still existed, where as the Amiga and Commodore didn't by the time PPC cards came out. And ten years later when the AmigaOne and OS4 came out there was too much of a gap. The Amiga community wasn't ready to jump from 68K and Amiga chipset to PPC and cheap PC boards. Even though it had been using them for ten years at least. Heck, around 2003 or so, I was still using my A1200/030 and surviving on AGA while normal people had 24 bit graphics for years.

Quote:

PPC was one of the best supported targets of operating systems early. It was better supported than any other RISC architecture. The PPC systems tended to be high priced workstations more than inexpensive desktops though. Yes, it was pretty much Apple alone driving down the cost of desktop PPC processors. In a more perfect world, one of either Amiga or Atari would have survived to compete but CBM upper management was too incompetent.

Yes Apple drove down the prices to be used in their own systems. Well mostly. Unfortunately the Amiga wasn't powerful enough to keep the 68K range going. And Motorola even gave up on the 88K. I wonder how things would have been. Had they kept going with their follow up to the 68K.

Quote:

Apple was big enough that they had negotiating power but that led to lower profit margins.

Hard to to tell from the Mac prices. Though they were the cheapest PPC systems around. Compared to what you get today.

Quote:

Mitch Alsup Quote:

Ha. Are you the Matt in the thread? It has your signature all over it.

Quote:

IBM did produce PowerPC based computers.

Well I would have said that would have made a great OS4 laptop, even though it was ten years too early, but at $12,000 that's more expensive than an X5000!

Quote:

Once again high end (modestly configured 850 cost upwards of $12,000) but a good OS selection with Windows NT 3.51 and 4.0, AIX 4.1.x, and Solaris Desktop 2.5.1 PowerPC Edition. IBM, like Apple, usually prefers the higher profit margins of high end markets rather than taking market share in lower margin markets like ARM Ltd. and the Raspberry Pi Foundation. Motorola upper management at least should have done more contemplating of the PPC situation as committing so completely to a lemon architecture lost them much of the embedded market and led to their downfall.

Hello Moto. Goodbye Motorola. I bet even Motorola mobiles are not even using a big endian CPU with an M on top.

@Rob

Quote:

Hindsight is a wonderful thing but I think Trevor would have been better off putting the A1222 money towards mass producing the Sam460 to get his low cost OS4 hardware. If it was £250-£300 for a Sam460 board with an OS4 license they may well have been able to sell 10,000 between 2015 and now.

You're probably right there. But, the A1222 does have a dual core with 36 bit addressing, so it's better for Linux!

10,000 would be a lot. It's estimated the XE barely sold in the early thousands. But by then (circa 2005) it was being compared to a PC, not an Amiga.

10,000 would be a good number for such a machine.

@matthey

Quote:

The Broadcom BCM2711 SoC is an inexpensive mobile device SoC with high CPU performance but lackluster GPU performance.

I was surprised when the Pi4 was introduced with A72 cores; I thought for sure they would move from A53 to A55.

VideoCore VI is OK for 2D graphics, so for a lot of uses it's fine. I guess they probably choose it because it's more open than the alternatives.

It would be nice to get an ARM SoC with a Radeon GPU, if it wasn't too expensive.

Last edited by bison on 19-Feb-2022 at 03:08 PM.

_________________
"Unix is supposed to fix that." -- Jay Miner

@Rob

Quote:

OS4 license they may well have been able to sell 10,000 between 2015 and now.

I have my doubts we would ever got to them number, but I think the situation would be a lot better than it is.

I do wonder what the number of AmigaOS4 licenses have been sold by Hyperion in total is? My gut feeling is it might of reached 10,000 if you also include each version, OS4/OS4.1/OS4.1FE and all the platforms. Plus I think the classic version is likely their biggest seller thanks to emulation support. it be interesting to know.

I think there only roughly 400 or less active users of Os4 at the moment, how do I know this, well the main app I guess all OS4 owners would download is the browser, the new version of Odyssey on OS4depot only has 325 downloads since Jan of this year. True some may of downloaded else where and also some might not be bothered in the browser. But then you also have to think about people who maybe downloading twice for different machines etc. So the numbers dont look good for OS4.

I wish their was something more that could be done, but like many here point out here I doubt a Sam460 at 700euros is going to be the answer. A1222 has more chance if it can be done for 400 euros, that might make some new people check it out, but my gut feeling is with the cost of parts going up, its going to be hard for a small low volume product hit that price point, even Pi foundation is struggling to keep there prices down.

Last edited by amigang on 19-Feb-2022 at 04:18 PM.

_________________
AmigaNG, YouTube, LeaveReality Studio

@Panthro

ACube-Systems came in as white knight at the right time, after Eyetech fiasco, and Eyetech exited in shame, there where a period of time where there was no hardware to buy. Primarly they were designed as a low-cost solution from the begining.

First came the Sam440, and then the Sam460, essentially they are embed PowerPC chips, and motherboards were designed not only for Amiga market, but also for automation market.

Sam460 is the first motherboard with PCIe that can run AmigaOS4.1 there was lack of PCIe drivers in the beginning, but gradually number of drivers covered most of stuff you needed like SATA, Sound cards, and Graphic cards, what kicked new life in this relatively slow computer is excellent drivers from AEON, enabling drivers with GPU acceleration. For video and 3d, graphics. Making this little thing into small multimedia computer.

Sam460 is what think of as lowest end formfactor to support the newest AEON graphic drivers. And think will be minimum requirement for many games and demos.

WarpOS support has been a rocky road, basically there two different implementation, one worked well on G3/G4 cpu’s and one that support every other PPC. There are some issues,

lots of Amiga/WaprOS demos need 15bit (or 24bit) screen modes, we have only had 16bit and 32bit modes. Some of old demos might use AHI or might use Paula, and use hard coded screen mode id’s and some do not. Basically, it’s a bit hit and miss. There some demos work in a workbench window, but as there is no scaling the window is tiny.

Only small changes needed, but a lot people are not around anymore, and source code is not anywhere to be found, it be great if demo groups open sourced there stuff.

Last edited by NutsAboutAmiga on 19-Feb-2022 at 09:21 PM.
Last edited by NutsAboutAmiga on 19-Feb-2022 at 09:19 PM.
Last edited by NutsAboutAmiga on 19-Feb-2022 at 09:10 PM.

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

NutsAboutAmiga Quote:

You keeps saying the same things over and over. If look at modern C/C++ programs large chunk of that is libc or glibc, or newlib, or stllib, on amiga (680x0) you have tiny amigalib (that only includes few std functions), this are almost always linked in by default.

Extensive function inlining and loop unrolling has caused bloat even on the Amiga. It would be good if the core designs encouraged code sharing by reducing function overhead and the benefits of loop unrolling as much as possible. Even with the 68060 it is sometimes not worthwhile to unroll loops but the lack of a hardware return/link stack and the default 68k ABI passing arguments on the stack raises function overheads. Keeping code size down is not just about the ISA and ABI but also the CPU design and compiler. However, even the poorest 68k code is almost always significantly more dense than PPC code. There are more PPC programs which are 50% bigger than 68k programs than there are PPC programs which are the same size.

NutsAboutAmiga Quote:

On Amiga (680x0) it pretty common to compress binary files, with power packer, stone cracker or something like that, because need to get it onto a tiny floppy disk, 2 mb is nothing today, so no effort goes into making anything smaller, on system with more ram and disk space.

Most 68k Amiga users have hard drives today no thanks to CBM. I haven't used a floppy in years. We usually have several MiB of memory too. We are not immune to bloat on higher performance 68k systems but we are still much lower footprint than PPC Amiga systems. There is a huge difference in code density between the 68k and PPC. Most compiled 68k code isn't even very optimal. The 68k AmigaOS is still compiled for the 68000 using the old SAS/C compiler which hasn't been updated in nearly 25 years (the last update was 1998) and had average at best code generation quality.

Welcome to the Unofficial Official SAS/C Compiler Support Page
https://www.warped.com/amiga/

PPC is using semi-modern compilers with many times more man years of development and can't come close to touching the footprint of the 68k.

NutsAboutAmiga Quote:

plus a lot debug stuff is also included by default, like function names, as required for stack trace, in case your program crash, AmigaOS3.x programs are compiled without debug symbols.

Debug and symbol hunks may make programs a little slower to load but practically do not use more memory.

NutsAboutAmiga Quote:

If your compiling something complicated like Web Browser, you need a up to date SDK that includes all that stuff, on classic Amiga or even the Vampire you love talking about, there not lot of memory, so you worry about running out of memory all the time. That’s not problem when you have 2GB or more memory on a Sam460 or AmigaOne computer.

The new Vampires have 512MiB of memory and I can't find anyone on forums complaining about running out of memory. This is not true of forum threads about PPC Amigas and Raspberry Pis.

NutsAboutAmiga Quote:

The largest part of memory usage is actual graphics, fonts and all wallpaper everything else you like to have like large true color icons. A modern game needs a lot of memory for textures. When look at something like the Vampire its not at all scaled for modern games. Vampire has only 512mb of RAM and 12mb of chipram, thats tiny.

PPC Amigas aren't "scaled for modern games" either. They are at least a decade behind modern game hardware specs yet are more expensive than modern hardware. The goal should not be modern hardware but affordable hardware to build an Amiga base which can be incrementally improved on. No where does the Vampire claim to be modern. It is retro and hobbyist even though the affordability could be better. Maybe it pushes the 68k Amiga hardware forward 10 years from the demise of CBM using some of their plans. The 12MiB of chip ram is just a choice to save gates in an affordable FPGA. The address lines can be extended to as many bits as desired and even the idea of chip memory could disappear completely as modern memory bandwidths are adequate for both chip memory and fast memory. I recall ThoR suggesting unifying memory and making an argument for it but one person is qualified to make all the decisions for the so called team. There is a good reason why ThoR quickly lost interest in the project even though he should have at least been involved in a real MMU for the Apollo core.

@matthey

If you believe Gunnar, it has a real MMU.

Quote:

No where does the Vampire claim to be modern.

Vampire? Apollo core? Google says…

“Apollo Core 68080 is the latest, natural and modern evolution of 68k processors”

“68080 is designed in a modern way”

“Apollo Core 68080 has many modern features which only latest processors have“

“designed like a modern processor“

“APOLLO FPU is a modern design“

etc

Last edited by kolla on 20-Feb-2022 at 01:31 AM.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

Hypex Quote:

The PPC version of MacOS was closer to 68K MacOS than the PPC version of AmigaOS is to 68K AmigaOS. Once solid on PPC they dumped it for OSX which provided a gateway for portability. But, Apple still existed, where as the Amiga and Commodore didn't by the time PPC cards came out. And ten years later when the AmigaOne and OS4 came out there was too much of a gap. The Amiga community wasn't ready to jump from 68K and Amiga chipset to PPC and cheap PC boards. Even though it had been using them for ten years at least. Heck, around 2003 or so, I was still using my A1200/030 and surviving on AGA while normal people had 24 bit graphics for years. Hypex [quote] The Amiga was more reliant on the 68k Amiga hardware for compatibility. There were more Amiga developers who liked the 68k Amiga hardware. The Mac was never about hardware although the 68k was well liked and 68040 Mac sales were some of the best relative to PC clone personal computer sales. The PPC architecture was a reasonable alternate choice albeit forced on Amiga users, much less human friendly and fat. It is easy to see why many Amiga users rejected it. Back then there were few good choices but today the PPC is a poor choice and the 68k is an option again. Hypex [quote] Yes Apple drove down the prices to be used in their own systems. Well mostly. Unfortunately the Amiga wasn't powerful enough to keep the 68K range going. And Motorola even gave up on the 88K. I wonder how things would have been. Had they kept going with their follow up to the 68K.

Embedded markets were enough to keep the 68k going but Motorola's 68k development trailed off while PPC embedded products were poorly received resulting in the loss of much of the embedded market. There were too many desktop RISC architectures competing to survive and the 88k was a victim of that. There were a few proponents of the 88k but likely not enough. It was a friendlier architecture than PPC and had pretty good performance. It showed that an extended precision FPU like the 68k FPU can have good performance at least. Mitch Alsup, one of the architects, I believe felt that it was superior to PPC and perhaps felt somewhat betrayed when management canned it.

Hypex Quote:

Hard to to tell from the Mac prices. Though they were the cheapest PPC systems around. Compared to what you get today.

Desktop economies of scale gave Apple pricing power that PPC workstation and even IBM PPC ThinkPad markets lacked. The 68k had dominated the workstation market before too many RISC competitors didn't have enough volumes to survive.

Hypex Quote:

Ha. Are you the Matt in the thread? It has your signature all over it.

Yes, but it is more important to know who Mitch Alsup is. ;)

https://www.linkedin.com/in/mitch-alsup-8691537 Quote:

Employment history: Current: Have developed method and apparatus to calculate double precision IEEE 754-2008 elementary transcendental functions in less than 20 cycles including argument reduction and uses a function unit just barely larger than an FMUL unit that also performs FDIV and SQRT calculations. Results are faithfully rounded in any of the IEEE 754-2019 rounding modes. Accuracy proof available. USPTO 10,761,806. E-Mail or Text if interested. Samsung America: Oct 2012 to Dec 2015:: Shader core architect. Responsible for the Instruction set architecture of the shader core in the Samsung cellphone GPU, creation of the microarchitecture and sequencers for the shader core, coordination and advancement of instruction set architecture with compiler development; designed tessellator for the GPU. Advanced Micro Devices: Oct 1999 to May 2007. Title: Fellow. Chief Architect of K-9 designed to be a 5 GHz successor to the 3 GHz Opteron family. Member of Architecture Review Board, member of I/O MMU design team, member of Virtualization architecture team. Ross Technology: Aug 1992 to June 1998 Chief Architect (Vice Presidential position) of SPARC V-8 family of processors, including Colorado, Apache, and Viper (SPARC V9). Motorola: Aug 1983 to Aug 1992: E13 Member of Technical Ladder, Chief Architect of Mc88K family of processors. Provided the RICS vision within Motorola, architected the Mc88100 and Mc 88200 chips, wrote processor architecture, inventor of the Mc88100 instruction set. Later was codeveloper of the Mc88120 processor architecture with Mike Shebanow.

He didn't even mention his 68k experience but I guess he didn't have room. At one time I hoped to lure him out of retirement to head the Apollo core development. He was working on a new RISC ISA which could have been co-developed by a fabless semiconductor business to give options. His new RISC ISA uses variable length immediates and displacements in a variable length instruction encoding like the 68k uses. RISC-V and ARM Thumb2 stick to simpler variable length encodings requiring multiple instructions to execute instructions with large immediates and displacements instead of gaining the advantage of scaled immediates and displacements for instructions which can be executed in one cycle. It's easier to handle longer instructions and execute powerful instructions than fetch and decode many short weak instructions and try to execute them while trying to avoid the added dependencies and load-to-use penalties. I'm glad he could verify what I thought.

Hypex Quote:

Well I would have said that would have made a great OS4 laptop, even though it was ten years too early, but at $12,000 that's more expensive than an X5000!

They were introduced in 1994 some 6 years before the X1000. They were also the highest end ThinkPads but IBM equipment was pricey. A loaded Pentium system could probably cost half that. There was also the PC Power Series which offered a beefy PPC 604.

https://en.wikipedia.org/wiki/IBM_PC_Series#PC_Power_Series

No price mentioned.

Hypex Quote:

Hello Moto. Goodbye Motorola. I bet even Motorola mobiles are not even using a big endian CPU with an M on top.

Even IBM POWER computers have better little endian support than big endian yet the code is still in big endian order. In my opinion, a hybrid endian computer is worse than an all big or all little endian computer. Endian conversion features are good though.

kolla Quote:

If you believe Gunnar, it has a real MMU.

I do not know the specifics or the precise nomenclature, but I tend to call it a Memory Protection Unit (MPU) with address translation features. The 68EC060 with no paged MMU has four transparent translation registers (ITT0, ITT1, DTT0, and DTT1) as well. I believe all effective addresses (EAs) are physical addresses so no virtual addresses for a paged MMU. It may save one pipeline stage but the pipeline is likely already extended more from being in an FPGA. An Address Translation Cache (ATC) better known as Translation Lookaside Buffer (TLB) is not needed too. It does simplify things not that a paged MMU is much more complex.

kolla Quote:

Vampire? Apollo core? Google says… “Apollo Core 68080 is the latest, natural and modern evolution of 68k processors” “68080 is designed in a modern way” “Apollo Core 68080 has many modern features which only latest processors have“ “designed like a modern processor“ “APOLLO FPU is a modern design“ etc

These quotes are mostly saying that the Apollo core is designed like a modern core with many features of a modern core not that it is a modern core. It is good enough to be a modern core for low end purposes but it is certainly not a modern competitive desktop core.

@amigang

Quote:

I wish their was something more that could be done, but like many here point out here I doubt a Sam460 at 700euros is going to be the answer. A1222 has more chance if it can be done for 400 euros, that might make some new people check it out, but my gut feeling is with the cost of parts going up, its going to be hard for a small low volume product hit that price point, even Pi foundation is struggling to keep there prices down.

The way things are going, before too long, someone will create a PPC chip simulator using a RPi4 that can emulate code using PJIT at 1Ghz equivalence speed. I mean, it's RISC to RISC. ARM and PPC share similarities as pointed out by 68k PJIT sources like 32-bit codes for 16-bit loads. Of course, PPC is still relatively modern with large register file and vectors in the good ones, so it may be a tight fit. Needing local RAM for emulated registers wouldn't be a good but most emulators aren't as hard to the metal as I think they should be.

@NutsAboutAmiga

Quote:

WarpOS support has been a rocky road, basically there two different implementation, one worked well on G3/G4 cpu’s and one that support every other PPC. There are some issues,

A friend got stuck on this. He had a Sam 440 and found Heretic II would work fine for a while then crash. He wanted to know why a PPC game didn't fully work on OS4. I ran it on my XE and it didn't crash but a G3/G4 is more compatible with a 603 or 604. The simple answer is the game was made for an acutal Amiga with a different CPU. Like PPC code running on a A1222 it needs some kind of load time or run time patching.