@OlafS25

If you pay the import fee I can have a large treasure chest of gold, rare gems and man made diamonds shipped to you from Nigeria. It’s better than 420 since it’s 419. I’ll even throw in a rare Egyptian Pharaoh golden cat. It’s really rare and was owned by King Tut. You must pay the fee though before it can be sent out. Don’t worry if your emails for tracking info go unanswered after payment is made.

_________________
Sent from my Quantum Computer.

@matthey

Quote:

... Maybe the PPC Amiga emulation experience is poor but the 68k Amiga emulation experience is close enough to compete with PPC Amiga hardware and desktop PC hardware offers more value than PPC Amiga hardware.

You're kidding, right? 68K Amiga emulation can *NOT* compete with PowerPC Amiga hardware at all. 68K emulation only compares well to old classic Amiga hardware.

Regarding the ASIC, I simply don't see a big enough market. Yes, there are "plenty" of classic 68K Amiga Fans, but I doubt there's enough to create a fancy new ASIC. People are having more than enough fun with FPGA based implementations.

@NutsAboutAmiga
Quote:

I see no harm in sending LE textures / bitmaps to the GFX card, why convert in the driver? I can just as easily work with GBRA as can with ARGB format, older driver used offer PC/LE modes. And if you’re converting 8bit to 16bit LE or BE makes no difference, its only palette lookup table that’s needs to be different (a table of 256 is quick to convert.)

*Everything* needs to be converted. Textures, vertex coordinates and data, shader constants, command packets, etc., 32-bit RGBA textures can be passed to the GPU without conversion, but 16-bit textures cannot, and nor can 16/32-bit float textures. Likewise, vertex data also needs to be converted. Getting the GPU to the conversion isn't feasible, and would also come with a performance penalty.

Hans

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

@Hans

Quote:

You're kidding, right? 68K Amiga emulation can *NOT* compete with PowerPC Amiga hardware at all. 68K emulation only compares well to old classic Amiga hardware

In what sense? If you are talking performance, then JIT 68K emulation on a modern x64 processor is not to be dismissed lightly. I have a corpus of code written to compile
for 68K, WarpOS and OS4. Back when my G4/800 was working, UAE was already running the 68K version at speeds comparable to or better than the native PPC version on the G4. And that was on a Core 2 Quad.

Admittedly there are much faster PPC than that running OS4 and MorphOS, but equally there are also much faster x64 on which to run the 68K code. My older i7 laptop is much faster than the aforementioned Core2. I don't have a bang up to date x64 to test with but I'm assuming the trend continues.

_________________
Doing stupid things for fun...

@Hans

There were some benchmarking some time ago, I think it was a conversion of a music file. People tested it on a lot of configurations including all sort of PPC. I did it in WinUAE (with JIT activated) and that configuration beated many of the PPC configurations except G5 (cannot remember G4 here) and I think X1000 and certainly also X5000. Slower PPC hardware was outperformed. What is slow of course is all sort of 3D because you cannot directly access the GPU in UAE and all is rendered using the simulated processor.

But yes I also doubt that the market is big enough to finance a ASIC version of a new 68k processor. You must first persuade a investor to give millions of dollars in such a new venture, I guess a hard sell. And for retro existing FPGA based hardware is more than enough.

Last edited by OlafS25 on 23-Sep-2022 at 08:48 AM.
Last edited by OlafS25 on 23-Sep-2022 at 08:44 AM.
Last edited by OlafS25 on 23-Sep-2022 at 08:41 AM.

@Hans

Quote:

You're kidding, right? 68K Amiga emulation can *NOT* compete with PowerPC Amiga hardware at all. 68K emulation only compares well to old classic Amiga hardware.

My 100€ RYZEN 3300X gives some amazing results on JIT. To be fair, I don't think PPC can't compare with it from the tests I've made*. And I have 3 idle cores still...

*I got tests where Doom on Amiga was running faster then on native, so there's that

_________________
Indigo 3D Lounge, my second home.
The Illusion of Choice | Am*ga

@OlafS25

Quote:

What is slow of course is all sort of 3D because you cannot directly access the GPU in UAE and all is rendered using the simulated processor.

We have Wazp3D a 3d Wrapper for WinUAE.

_________________
Indigo 3D Lounge, my second home.
The Illusion of Choice | Am*ga

@Karlos

I have no benchmarks to compare brute CPU performance.

If you find the overall experience of 68K WinUAE emulation comparable to AmigaOS 4, then have fun with WinUAE. I uninstalled AmigaForever months ago after having not used it for years (needed the HD space).

Hans

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

@OlafS25

Quote:

But yes I also doubt that the market is big enough to finance a ASIC version of a new 68k processor. You must first persuade a investor to give millions of dollars in such a new venture, I guess a hard sell. And for retro existing FPGA based hardware is more than enough.

Millions just for the ASIC. The cost of drivers and software need to be factored in too, on top of the ASIC development, motherboard development, packaging, marketing, etc.

Hans

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

@Hans

Quote:

If you find the overall experience of 68K WinUAE emulation comparable to AmigaOS 4, then have fun with WinUAE. I uninstalled AmigaForever months ago after having not used it for years (needed the HD space).

That's a different metric all together, and one that is entirely subjective. We were talking about raw processing power, which is not subjective. Except for cases where highly tuned SIMD code is used, I've no doubt that 68K can deliver equal or perhaps better than native PPC performance. At least for any like-for-like scalar code.

Speaking subjectively, however, if all the UX related features you love from OS4 were recompiled for 68K and ran on the fastest 68K emulations available, I put it to you that you would not be able to tell the difference.

_________________
Doing stupid things for fun...

@Karlos

I think the main problem of the Amiga market has always been the excessive love of 68k and old chipsets. With a view to correct growth and development of the os, I believe it is more correct to pursue the path of researching new powerpc platforms for 'fast and simple' porting and, at the same time, the migration of the NG os on new platforms, fully supported and at low cost such as for example the Arm RPi or the new introduced Risc-V socs

Last edited by Fl@sh on 23-Sep-2022 at 02:18 PM.

_________________
Pegasos II G4@1GHz 2GB Radeon 9250 256MB
AmigaOS4.1 fe - MorphOS - Debian 9 Jessie

@Hans

I thought you always had 1 Petabyte free?

_________________
Sent from my Quantum Computer.

@Hans

Quote:

*Everything* needs to be converted. Textures, vertex coordinates and data, shader constants, command packets, etc., 32-bit RGBA textures can be passed to the GPU without conversion, but 16-bit textures cannot, and nor can 16/32-bit float textures. Likewise, vertex data also needs to be converted. Getting the GPU to the conversion isn't feasible, and would also come with a performance penalty. Hans

That’s true, but if you have automated conversion, doing every time something is uploaded, its kind wasteful, if you modify something, it won’t make sense to convert everything, only the part that was changed, I also argue that maybe it’s a good idea to only upload the changes, partial bitmaps, partial vertex data, it there is potential reducing overhead, by being smart about when and how, using preconverted data.

Kjetil

Last edited by NutsAboutAmiga on 23-Sep-2022 at 05:28 PM.
Last edited by NutsAboutAmiga on 23-Sep-2022 at 05:28 PM.

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

@Fl@sh

Quote:

Fl@sh wrote: @Karlos I think the main problem of the Amiga market has always been the excessive love of 68k and old chipsets. With a view to correct growth and development of the os, I believe it is more correct to pursue the path of researching new powerpc platforms for 'fast and simple' porting and, at the same time, the migration of the NG os on new platforms, fully supported and at low cost such as for example the Arm RPi or the new introduced Risc-V socs

Begging your pardon, but which "new powerpc platforms" ? POWER and by extension PowerPC is more obsolete and less relevant to computing in almost every sector (with the exception of IBM's own HPC systems) than the 68K was when PowerPC was introduced.

As for porting to new platforms, that's all well and good but the law of diminishing returns applies. Not all 68K software was ported to PowerPC, but that's OK because mature 68K emulation exists on PowerPC.

I've said it before and I'll say it again. 68K is the only future that makes any sense for the platform if you want to appeal to as many users as possible. Not necessarily as an OS, but as common bytecode format and a set of standard APIs for application code and binary distribution across the gamut of AmigaOS and work-alike platforms.

_________________
Doing stupid things for fun...

@Hans

On of the really cool drivers for shapeshifter is the Savage driver, I believe divides up screen in sectors, and uses MMU to notify when something needs to be converted, its where interesting trick that worked well. Not sure how this works, I guess part of screen that is not updated frequently is MMU trapped, while parts are updated is just converted, or something like that, it somehow need to know if something is updated or not. Anyhow it’s really smart about know what need updating, how often it needs to be updated. reducing overhead, by working smart, not hard.

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

@Karlos

That makes absolutely no sense.

68K was outdated when PowerPC came to market, PowerPC is outdated, so 68K is relevant again?

I beg your parted but that’s a logical fallacy. 68K has not changed or being updated to point where relevant, it can’t become relevant because something new has come un-relevant.

Yes, their good CPU manufacture out there make excellent chips and sure AMD make stunning PowerPC chip If there was demand for it. sadly, for the Amiga community there is no demand for 680x0, nor any demand for PowerPC, one chip is newer clocked at pretty high clock frequency’s, the other one is horrible slow, I know what chip I pick.

the instruction set is not important anymore it C/C++ code that’s the standard, the transition from 68000 assembler to C, make the PowerPC useful, as a bridge, between the sone age, and middle age.

it’s like middle age with modern weapons, it’s like a time traveling GPU going back to forgotten time in computer industry.

Last edited by NutsAboutAmiga on 23-Sep-2022 at 06:11 PM.

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

OlafS25 Quote:

There were some benchmarking some time ago, I think it was a conversion of a music file. People tested it on a lot of configurations including all sort of PPC. I did it in WinUAE (with JIT activated) and that configuration beat many of the PPC configurations except G5 (cannot remember G4 here) and I think X1000 and certainly also X5000. Slower PPC hardware was outperformed. What is slow of course is all sort of 3D because you cannot directly access the GPU in UAE and all is rendered using the simulated processor.

2D RTG chunky is fast because native chunky support is used directly and the 68k Amiga has RTG drivers for this. The 68k Amiga has no drivers for the native 3D graphics of most modern PCs and the 68k 3D APIs are old for these newer graphics cards using shaders. Newer AmigaOS 4 3D APIs exist that could be ported back to the 68k and used under 68k emulation which theoretically could give competitive performance and features as AmigaOS 4 hardware. This would proliferate these new 3D APIs on the Amiga which should increase adoption by Amiga developers resulting in more software supporting them. Not all GPUs would be supported and AmigaOS 4 hardware producers may choose to block access to these new APIs to try to leverage a technological advantage over 68k Amiga emulation and encourage Amiga users to buy AmigaOS 4 hardware. I expect more Amiga users are choosing the better compatibility and much cheaper cost of 68k Amiga emulation while foregoing some 3D performance for now. The 3D situation in emulation is still a little better than '90s Amiga hardware.

WinUAE Amiga OS 3.1: Quake3
https://www.youtube.com/watch?v=Va87x7ookw8

My 68060@75MHz with Voodoo 4 maybe could struggle with it too if I had enough fast memory and enough texture memory (Mediator bus performance is very slow but performance is descent if all textures can be loaded into GPU memory).

OlafS25 Quote:

But yes I also doubt that the market is big enough to finance a ASIC version of a new 68k processor. You must first persuade a investor to give millions of dollars in such a new venture, I guess a hard sell. And for retro existing FPGA based hardware is more than enough.

The size of the market depends on the price of the product. Jeff Porter commented about developing the low cost Amiga 500 where dropping the price by half could double demand. A SoC ASIC allows to drop the price to a fraction of niche market hardware Amiga products but it should also boost performance by several times. Performance increases and price decreases have synergies which increase demand. The Raspberry Pi Foundation understands the importance of the "price-performance ratio" and how much of a game changer a custom ASIC is for competitiveness.

https://www.arm.com/blogs/blueprint/raspberry-pi-rp2040 Quote:

Designing a sub-$5 computer involves several tough decisions. Decisions such as which instruction set architecture (ISA) to use and which class of chip IP to select from within that ecosystem in order to balance price, power and performance. Then there are design and manufacturing questions: who’s going to design the system on chip (SoC) and who’s going to fabricate it? All of these considerations dictate how useful and affordable your final product is. So when it came to selecting a chip to power our first microcontroller-class product, Raspberry Pi Pico, we knew that we needed to push that price-performance ratio harder than ever before.

Retro Games Ltd. understands the value of an ASIC even when primarily used to reduce cost. They didn't view the Amiga market as too small or they never would have approached Jeri Ellsworth. Their C64 direct to TV joystick sold ~186,000 units in one hour on QVC. The Amiga has more hardware to support but it has more retro gaming appeal, modern features can easily be added and the performance can be upgraded with little additional cost in an ASIC. THEA500 Mini has surely sold at least tens of thousands of units to make the Amazon UK top computer products chart and will likely sell hundreds of thousands of units when they lower the price.

You saw the Natami excitement. You were there. How do you explain the Natami "MX Bringup Thread" with 761,487 views (just what I observed one day and likely climbed after). No doubt there were people who were responsible for 10 views but even considering this as an average would give ~76k people with no advertising (only grass roots word of mouth retro hardware project excitement). If the number of Natami forum users was the same size as the PPC AmigaOS 4 community which is about 5k users, they would have each had to read that thread 152 times each. The Natami wasn't even planned to be much if any cheaper than PPC AmigaOS 4 hardware and CPU performance would have been a fraction of PPC AmigaOS 4 hardware. Thomas knew that 68k CPU options were the Achilles heel of the project and likely one of the major reasons why he pulled the plug. If there were 76k people that each donated $/€10 a cheap ASIC would have already been possible. The CAGA funding asks more from each user and may raise enough to produce an ASIC. Would a 68k Amiga ASIC or another Amiga web site do more for the Amiga?

Hans Quote:

Millions just for the ASIC. The cost of drivers and software need to be factored in too, on top of the ASIC development, motherboard development, packaging, marketing, etc.

You sound like Dr. Evil asking for one million dollars ransom. A few million dollars is cheap considering the alternatives are continued subsidizing of noncompetitive PPC AmigaOS 4 hardware that has only resulted in about 5k users after about 2 decades or becoming a software only business spending hundreds of thousands of dollars to port the AmigaOS to ARM and millions of dollars to make it competitive on the desktop where it will be competing against more mature free OSs with much larger software libraries. There is the other option of just calling it quits and open sourcing everything. There is a little a bit of money to be made selling the AmigaOS and ROMs for emulation.

@NutsAboutAmiga

If you read past the first few sentences or makes perfect sense. I'm not suggesting you use a 68K machine, but if matthey offered me his super N GHz ASIC 68K tomorrow it would be a more interesting proposition than any current PowerPC.

What I am saying is, use (an appropriate subset of) the 68K instruction set as a bytecode for implementing binaries that run on all Amiga and Amiga like systems. The obvious additional requirements are a set of standardised API for it to interact with. For non 68K host platforms, add native code support for drivers, libraries and datatypes and what are you missing?

It is the absolute zenith of denial to believe that PowerPC is anything other than an expensive total dead end as a platform. It has zero hope of gaining any further relevance, not even as an emulated target since PPC emulation is a pain in the arse to implement properly and everyone else that was using it has migrated away already.

By stark contrast 68K JIT is far more mature. Moreover 68K on x64 JIT can already outperform native PPC hardware and that's a gap that's only going to get wider as time passes.

Even without the JIT performance argument, 68K is more relevant to more users. PPC is a cloistered sub niche of what was already a niche, with expensive entry requirements. You only have to look at the interest in 68K projects. PPC was fun, but it's dead, Jim.

Quote:

the instruction set is not important anymore it C/C++ code that’s the standard, the transition from 68000 assembler to C, make the PowerPC useful, as a bridge, between the sone age, and middle age.

This is only true when the code is there to be shared and written to be appropriately portable. AROS is already there in many respects. However, someone may not want to share their code and may not want to have to compile and maintain it for multiple platforms. A well defined 68K binary target that works across the spectrum of Amiga like systems (including real 68K HW) is a far better proposition.

Last edited by Karlos on 23-Sep-2022 at 08:40 PM.
Last edited by Karlos on 23-Sep-2022 at 08:38 PM.

_________________
Doing stupid things for fun...

@matthey

Quote:

noncompetitive PPC AmigaOS 4 hardware that has only resulted in about 5k users after about 2 decades...

Keep spanking that dead horse AmigaOne Team!

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

@matthey

I thought MorphOS on PPC was cool when it allowed Amigans to play the Amiga version of WipEout2097! But now I can run the Playstation version on THEA500 Mini, why bother even with that? Blender, Candy Factory, Spencer, Tower57... I've run out...

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

Karlos Quote:

Begging your pardon, but which "new powerpc platforms" ? POWER and by extension PowerPC is more obsolete and less relevant to computing in almost every sector (with the exception of IBM's own HPC systems) than the 68K was when PowerPC was introduced.

POWER is an option for the high end of the 5k user PPC AmigaOS 4 market. It could be a niche of a niche market for a handful of elitists. Just jack up the price until there is a comfortable profit. Amiga for the classes instead of the masses.

Karlos Quote:

As for porting to new platforms, that's all well and good but the law of diminishing returns applies. Not all 68K software was ported to PowerPC, but that's OK because mature 68K emulation exists on PowerPC.

WinUAE with 68k+chipset compatibility and lower cost hardware > PPC "mature 68k emulation"

Karlos Quote:

I've said it before and I'll say it again. 68K is the only future that makes any sense for the platform if you want to appeal to as many users as possible. Not necessarily as an OS, but as common bytecode format and a set of standard APIs for application code and binary distribution across the gamut of AmigaOS and work-alike platforms.

+1

NutsAboutAmiga Quote:

That makes absolutely no sense. 68K was outdated when PowerPC came to market, PowerPC is outdated, so 68K is relevant again?

No. The 1st gen PPC 601 was a general purpose design that came out circa 1993 (68060 was released in 1994). Let's compare.

PPC601 OoO 4 stage superscalar, up to 3 instructions/cycle issued
68060 in-order 8 stage superscalar, up to 3 instructions/cycle issued

PPC601 32kiB L1 caches
68060 16kiB L1 caches

PPC601 64 bit data bus, external L2 cache support
68060 32 bit data bus, no external L2 cache support

PPC601 instruction fetch: 32 bytes/cycle
68060 instruction fetch: 4 bytes/cycle

You can probably see by now that the PPC 601 has more performance features while the 68060 has balanced, if not embedded, features. The 601 was clocked up more despite the 68060 having the deeper pipeline but high clock speeds are not as important for embedded use or for a CPU that could out clock shallow pipeline PPC CPUs making them look bad. The PPC 601 is likely a little higher performance at the same clock speed but the 68060 was competitive at a lower hardware cost and should have been able to out clock the 601 to help close the surprisingly narrow gap (when the Rev6 68060 finally received a die shrink, most of the chips were still marked 50MHz even though they commonly run at 100MHz today).

The closest 68060 competitor was the 2nd generation 1994 PPC 603 designed for embedded use but Apple was going to use it in Mac laptops.

PPC603 OoO 4 stage superscalar, up to 3 instructions/cycle issued
68060 in-order 8 stage superscalar, up to 3 instructions/cycle issued

PPC603 16kiB L1 caches
68060 16kiB L1 caches

PPC603 32/64 bit data bus, no external L2 cache support
68060 32 bit data bus, no external L2 cache support

PPC603 instruction fetch: 8 bytes/cycle
68060 instruction fetch: 4 bytes/cycle

The 68060 outperformed the PPC 603 at the same clock speed. Apple decided not to use the 603 in laptops due to low performance but stuck it in the non-performance Performa instead leading to a bad reputation for low end PPC CPUs and Motorola quickly coming out with the 603e which doubled the L1 caches. The 68060 was still competitive but the 603e also received a die shrink and was clocked up while the 68060 with double the pipeline depth was not.

The 68060 was competitive with similarly targeted 1st and 2nd generation PPC CPUs. It's uncommon to see in order CPUs competing with OoO CPUs today and even more unusual to see an in order CPU with fewer performance features competing with OoO CPUs. Pay no attention to the 68060 behind the curtain though. It's for embedded use only.

NutsAboutAmiga Quote:

I beg your parted but that’s a logical fallacy. 68K has not changed or being updated to point where relevant, it can’t become relevant because something new has come un-relevant.

Was PPC AmigaOS 4 hardware ever relevant with 5k users? Was PPC AmigaOS 4 hardware ever more relevant than 68k hardware and emulation with a much larger user base?

NutsAboutAmiga Quote:

Yes, their good CPU manufacture out there make excellent chips and sure AMD make stunning PowerPC chip If there was demand for it. sadly, for the Amiga community there is no demand for 680x0, nor any demand for PowerPC, one chip is newer clocked at pretty high clock frequency’s, the other one is horrible slow, I know what chip I pick.

Supercomputer CPUs from the '90s are slow today in comparison to desktop CPUs. Is it so difficult to figure out why? Why are your decade old PPC CPUs being outperformed by a $35 RPi?

NutsAboutAmiga Quote:

the instruction set is not important anymore it C/C++ code that’s the standard, the transition from 68000 assembler to C, make the PowerPC useful, as a bridge, between the sone age, and middle age. it’s like middle age with modern weapons, it’s like a time traveling GPU going back to forgotten time in computer industry.

CPU designs are more important than ISAs but ISAs still matter and assembler has not been fully replaced. If anything, assembler optimizations gave x86(-64) games an edge while developers failed to make compilers smart enough for PPC to compete. An ISA with human friendly assembler like the 68k is an asset. Fat traditional RISC with "smart" compilers lost.


Alpha
PA-RISC
MIPS
88k
SPARC
PPC


RIP RISC. It's funny that I listed them worst to best code density which is close to the order they died but code density doesn't matter either because it wasn't supposed to with the RISC philosophy. You can keep ignoring history or you can reexamine the RISC propaganda.

Last edited by matthey on 23-Sep-2022 at 10:03 PM.