@kolla

Quote:

The ONLY reason it is “profitable” is because it is done for FREE

Full Ack.

Quote:

Let me tell you about the THEC64 Mini. It was in many stores here, they had lots of them (and other minis) piles sitting there for months and months, prices going down a couple of times too iirc. My kid got one from a friend for xmas 2020. Then when I wanted to buy another one for h4cking reasons, they were suddenly all gone.

Sorry to hear that. A missed opportunity. Something similar happened to me. PS1Classic was released, I got one right away.
It is not a bad piece of kit once you put Bleem on it. Absolutely great piece of kit.

Prices dropped to eventually be 40$ at the gameshop or 20$ if you bought a full priced game.

When I decided I wanted a couple more of them, one for a friend and a second one to keep in its original unopened box in the game-cave they were all gone
And nowadays they sell on ebay for above the initial launch price.

(near 200$ for this, that means that no matter what someone will make serious money and that alone (making money) tells me Hyperion is not involved in any form.)

Lesson to be learnt. Don't hesitate when product is available.
THEA500 on the other hand is well above impulse purchase price point. It is basically a respberry Pi in a nice small box, for 200$. I will pass.

Last edited by V8 on 23-Apr-2022 at 11:13 AM.
Last edited by V8 on 23-Apr-2022 at 10:39 AM.

@matthey

In my view the traditional amiga line based on 68k including more powerful NG 68k hardware like V4 and PiStorm must be seen different from something targetting "normal users" using something as main system.

For me the 68k line satisfies certain needs/interests like feeling 20 years younger with certain memories and with new hardware creating a different reality based "what would have been" if Commodore would have survived and amiga would have developed further and Motorola would not have dropped the 68k processor line. There is a market for that obviously and FPGA is best suited for it. Even if it would be "real hardware" you still have the shortcomings like no memory protection and 64bit and multicore support. Adding such features including adapted OS would change the concept of the platform taking away some of the advantages and break lots of software without adding new to it. I do not see any sense in it (from a commercial point of view). Concentrating on FPGA makes a lot more sense to me here.

What is currently called "NG" is only halfway in my view inheriting many of the shortcomings of the older amiga platform just running on faster hardware.

If you want a modern platform you need both modern hardware and modern OS as base. For that I hope Linux will become based on the work of Deadwood. I am already looking forward to do a distribution then based on Linux and using Scalos as desktop. We will see. But then not more than the look&feel of amiga will be left and some converted software and technologies. Most will be no longer from amiga but from Linux.

To gaming... I think best for playing old amiga games are unexpanded A500 or A1200 or WinUAE.
If you want to play some more demanding ports and also do some productivity best is WinUAE or V4/PiStorm. As main platform no option currently is suited in my view (even if some try that). For that I hope on a linux based system as I already mentioned. My own dream

Last edited by OlafS25 on 23-Apr-2022 at 11:15 AM.
Last edited by OlafS25 on 23-Apr-2022 at 11:13 AM.
Last edited by OlafS25 on 23-Apr-2022 at 11:06 AM.
Last edited by OlafS25 on 23-Apr-2022 at 11:05 AM.

OlafS25 Quote:

In my view the traditional amiga line based on 68k including more powerful NG 68k hardware like V4 and PiStorm must be seen different from something targetting "normal users" using something as main system. For me the 68k line satisfies certain needs/interests like feeling 20 years younger with certain memories and with new hardware creating a different reality based "what would have been" if Commodore would have survived and amiga would have developed further and Motorola would not have dropped the 68k processor line. There is a market for that obviously and FPGA is best suited for it. Even if it would be "real hardware" you still have the shortcomings like no memory protection and 64bit and multicore support. Adding such features including adapted OS would change the concept of the platform taking away some of the advantages and break lots of software without adding new to it. I do not see any sense in it (from a commercial point of view). Concentrating on FPGA makes a lot more sense to me here.

FPGA isn't good enough for a high performance CPU. Even emulation gives better CPU performance. Why go to the trouble of creating a more accurate FPGA Amiga "what would have been" and then leave it in a cramped FPGA with 1/10 the performance at 10 times the production cost or more? Why not lower the cost with an ASIC to where the lack of memory protection is not a problem because the competition is not small enough footprint to compete?

OlafS25 Quote:

What is currently called "NG" is only halfway in my view inheriting many of the shortcomings of the older amiga platform just running on faster hardware.

Right. "NG" is only halfway to extinction from losing compatibility, fattening up and trying to compete on the desktop. PPC is dead for the first half and now ARM and x86-64 AmigaOS bastardizations are being attempted for the last half of extinction.

OlafS25 Quote:

If you want a modern platform you need both modern hardware and modern OS as base. For that I hope Linux will become based on the work of Deadwood. I am already looking forward to do a distribution then based on Linux and using Scalos as desktop. We will see. But then not more than the look&feel of amiga will be left and some converted software and technologies. Most will be no longer from amiga but from Linux.

No thanks to Linux with an Amiga skin. It's not Amiga.

OlafS25 Quote:

To gaming... I think best for playing old amiga games are unexpanded A500 or A1200 or WinUAE. If you want to play some more demanding ports and also do some productivity best is WinUAE or V4/PiStorm. As main platform no option currently is suited in my view (even if some try that). For that I hope on a linux based system as I already mentioned. My own dream

I prefer to have Amiga dreams that expand the Amiga user base with competitive retro hardware. Current gaming hardware is only halfway to that goal too. Linux not required.

@matthey

Quote:

Why go to the trouble of creating a more accurate FPGA Amiga "what would have been" and then leave it in a cramped FPGA with 1/10 the performance at 10 times the production cost or more? Why not lower the cost with an ASIC to where the lack of memory protection is not a problem because the competition is not small enough footprint to compete?

Please go ahead then - YOU CAN DO IT!!!

(oh, but you need that expensive FPGA anyways for that damn chipset... or do you want to make ASIC out of them too? What to do, what to do!)

Last edited by kolla on 24-Apr-2022 at 12:35 AM.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

Linux is not the answer!
I installed Unbutu and tried to get AMD openCL drivers installed but the stupid compiler option change in gcc 9.0.4 over 9.0.0 makes the script fail.

I had to dig to find it.

Most annoying OS ever!

@Lou
Quote:

Linux is not the answer!

But emulation under Linux could be.

I am pretty happy with the Amiga Forever KXLight install.

I would not consider it the best, but it is a strong alternative for anyone who has an old PC around.

Last edited by Nonefornow on 24-Apr-2022 at 02:15 AM.

@matthey

Amiga NG is something like Mac Os 9.
The same Amiga as from Commodore only better because cheaper and many times faster.

As a software devloper I don't care what cpu is in Amiga.
Classic or current NG Amiga may work on any cpu that work in 32 bit big endian mode.
It may be anything. 68k, ppc, mips, sparc even Intel Itanium or IBM Mainframe.
But not x86, ARM, Risc-V.

68060 was too expensive and too slow twenty five years ago and still is.

So It was good decision to switch to ppc and after so many years You should accept that.

@ppcamiga1

you are software developer for amiga? What have you done already?

and no switching to ppc was not a good decision looking back

if ever switch should have been to ARM or X86

The trick is to do what Apple did (and just about everyone else too) - switch to hardware architecture agnostic (ie portable) code, so that architecture becomes an implementation detail.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

ppcamiga1 Quote:

Amiga NG is something like Mac Os 9. The same Amiga as from Commodore only better because cheaper and many times faster.

AmigaNG hardware may be cheaper after adjusting original Amiga hardware prices for inflation but it is expensive and slow relative to modern competition which explains why it has only sold a few hundred units a year on average. Sadly, THEA500 Mini is more competitive and likely outsold all AmigaNG hardware ever in its first week available for sale.

ppcamiga1 Quote:

As a software devloper I don't care what cpu is in Amiga. Classic or current NG Amiga may work on any cpu that work in 32 bit big endian mode. It may be anything. 68k, ppc, mips, sparc even Intel Itanium or IBM Mainframe. But not x86, ARM, Risc-V.

lol. You don't care what CPU as long as it is a dead CPU architecture. Endianess and 32 bit support are just two factors of compatibility. For best Amiga compatibility, stick with the 68k. All the other big endian architectures are dead so the best choice to resurrect is the 68k.

ppcamiga1 Quote:

68060 was too expensive and too slow twenty five years ago and still is. So It was good decision to switch to ppc and after so many years You should accept that.

The 68060 had about 40% better integer performance and similar floating point performance to the Pentium at the same clock speed using the ByteMark benchmark.

https://amigaworld.net/modules/newbb/viewtopic.php?topic_id=44391&forum=25#847418

Early ARM cores were especially weak and had minimalist features. It was the low price and low power of the minimalist cores that made ARM popular for embedded use.

1994 68060 - 1.52 DMIPS/MHz
1994 ARM610@30MHz - 0.65 DMIPS/MHz (Acorn RISC PC - no FPU)
1995 ARM710@40MHz - 0.68 DMIPS/MHz (Acorn RISC PC - no FPU)

Even the later StrongARM CPUs at around 200MHz used in the Acorn RISC PCs could not match the DMIPS/MHz. ARM and Pentium were clocked up where the best performer, the 68060, was relegated to the basement due to a political decision to switch to PPC. Ironically, the longer 8 stage pipeline of the 68060 should have made it much easier to clock up than existing Pentium, PPC and ARM CPUs at that time.

kolla Quote:

The trick is to do what Apple did (and just about everyone else too) - switch to hardware architecture agnostic (ie portable) code, so that architecture becomes an implementation detail.

Yet Apple has moved to an integrated SoC and consoles have moved to integrated SoCs more like CBM had with the Amiga before procrastinating on enhancing and integrating the Amiga hardware. RJ Mical and Dave Needle integrated 3 custom chips into 1 chip of the 3DO in about 1 year (1993-1994) in comparison to the Amiga not integrating any of the Amiga custom chips together in nearly 10 years (AA+ was to have integrated 2 custom chips into 1).

Revisiting the Panasonic 3DO FZ-10 32 bit console - is it underrated? Review, Teardown, Games | MVG
https://www.youtube.com/watch?v=l01No0chbqw

This video has many Amiga references by the way as Modern Vintage Gamer is a huge Amiga fan picking the Amiga 1200 as the one retro computer he would keep if he had to choose just one.

3DO was an early ARM user.

1993 ARM60@12.5 MHz (3DO - no caches, no FPU, 1st support of 32 bit addressing and hardware multiply)

We can see just how Spartan ARM CPUs were at that time with the 3DO CPU being closer to a 68020 than a 68030, 68040 or soon to arrive 68060.

- Horses are nice.
- Yes, but donkeys are nicer, because dinosaur.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

MISTER vs THE A500 Mini

https://www.youtube.com/watch?v=9UFColj26NM

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

@ppcamiga1

What on earth are you developing that requires a CPU is 32 bit and big endian but other than that doesn't care which? I've never heard anything so preposterous. And that's coming from someone that does stupid preposterous things for fun whenever possible.

Let's be clear. Speaking as someone that has worked on OS4 in the past and developed code for both WarpOS and PowerUp before that, PPC was chosen originally because it was the lowest hanging fruit for a CPU upgrade from 68K. Apple had proven that emulation was feasible and Phase5 kicked the whole thing off by bringing out the Cyberstorm / CyberVision PPC cards.

There's no doubt that having the same endianness as 68K simplified the design of software that used both PPC and 68K in the PowerUp/WarpOS environment, allowing data to be shared between the processors without conversion.

Nevertheless, UAE and Amithlon have also shown that even on little endian CPUs, compatible and performant 68K emulation is also possible.

Last edited by Karlos on 25-Apr-2022 at 01:29 PM.
Last edited by Karlos on 25-Apr-2022 at 01:28 PM.

_________________
Doing stupid things for fun...

@NutsAboutAmiga

Quote:

MISTER vs THE A500 Mini

No it’s not, that’s the MiST vs the THEA500 Mini.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@kolla

I tend to think it's funny that Apple ARM models are referred to as Apple Silicon. How is it Apple silicon? They didn't design the CPU. In that case why wasn't the PPC called Apple Silicon as well? They forged it in the AIM alliance. Perhaps the strangest thing is there are now two CPU types again with three common OS. When PPC had died off and only x86 was left. Now ARM comes in to really replace PPC for Apple. They've already gone down this road. They went to Intel and secured the future. But now they are backtracking and using another RISC again. It's just another PPC really. If PPC failed to compete with x86 how do they expect ARM to? Just my thoughts.

@Hypex

Actually PPC and ARM are both called RISC, but only ARM is RISC in spirit IMHO. PowerPC certainly didn't have a "reduced instruction set" as was understood back in the day. It was big and had plenty of complex operations. It just didn't operate directly on memory. A big register file with load/store access to memory.

By comparison, ARM was lean and had plenty of interesting behaviours for the day, such as conditional execution of instructions to simplify the pipeline design (less branchy code, for example).

The lesson to take from Apple, for all the crap I don't like, is that they have mastered vertical integration. If RISC-V overtakes ARM, they'll probably migrate to that.

_________________
Doing stupid things for fun...

Hypex Quote:

I tend to think it's funny that Apple ARM models are referred to as Apple Silicon. How is it Apple silicon? They didn't design the CPU. In that case why wasn't the PPC called Apple Silicon as well? They forged it in the AIM alliance. Perhaps the strangest thing is there are now two CPU types again with three common OS. When PPC had died off and only x86 was left. Now ARM comes in to really replace PPC for Apple. They've already gone down this road. They went to Intel and secured the future. But now they are backtracking and using another RISC again. It's just another PPC really. If PPC failed to compete with x86 how do they expect ARM to? Just my thoughts.

Apple invested in the early development of ARM and PPC but did not actively participate. They became a fabless semiconductor designer later (~2000?). There are ARM reference designs available but Apple has made major changes to the CPU cores. ARM is cheap providing an improved profit margin and Apple can integrate the cores into its custom SoCs. Sony and Microsoft paid AMD to make their console x86-64 SoCs rather than designing them in house. Before that, IBM designed the PPC chips for consoles as it was not easy to license PPC until the architecture was already dead. Also, ARM has the development process streamlined for fabless semiconductor development.

Karlos Quote:

Actually PPC and ARM are both called RISC, but only ARM is RISC in spirit IMHO. PowerPC certainly didn't have a "reduced instruction set" as was understood back in the day. It was big and had plenty of complex operations. It just didn't operate directly on memory. A big register file with load/store access to memory.

If a "reduced instruction set" is a requirement of RISC, then AArch64, the most recent legacy ARM architecture replacement must not be RISC either. Some RISC purists would say ARM Thumb2, which won ARM the embedded market using a variable length encoding, is not RISC either. The variable length encoding is popular with RISC-V also. Is RISC dead?

Karlos Quote:

By comparison, ARM was lean and had plenty of interesting behaviours for the day, such as conditional execution of instructions to simplify the pipeline design (less branchy code, for example).

ARM was lean and weak alright. I suppose the conditional execution is "interesting" but it may be what sunk the legacy ARM architecture. Like branch delay slots which the legacy architecture also uses, conditional instruction execution works better for certain core designs reducing flexibility of core designs. The conditional instruction execution was not worth it considering the poor code density. Code density was likely similar to PPC but PPC had twice as many GP registers, hardware MUL/DIV, often a FPU, etc. The following chart shows stats on early simple and weak ARM "RISC" cores.

Year | Microarchitecture | Pipeline Depth | DMIPS/MHz | Frequency (MHz) | L1 Cache
1985 ARM1 3 0.33 8 N/A
1992 ARM6 3 0.65 30 4kiB unified
1994 ARM7 3 0.9 100 0–8kiB unified
1999 ARM9E 5 1.1 300 0–16kiB I+D
2002 ARM11 8 1.25 700 4–64kiB I+D
2005 Cortex-A8 13 2.0 600-1000 32kiB I+32kiB D
2009 Cortex-A9 8 2.5 1000 16–64kiB I+D
2011 Cortex-A7 8 1.9 1500 8–64kiB I+D
2011 Cortex-A15 15 3.5 2000 32kiB I+D
2012 Cortex-M0+ 2 0.93 60–250 None
2012 Cortex-A53 8 2.3 1500 8–64kiB I+D
2012 Cortex-A57 15 4.1 2000 48kiB I+32kiB D

https://www.sciencedirect.com/topics/computer-science/stage-pipeline

1994 68060 8 1.52 50-60 8kiB I+8kiB D

I couldn't find an ARM core that matched the 1994 68060 DMIPS/MHz performance for another decade after the 68060 with the 13 stage in order Cortex-A8 in 2005. The Cortex-A8 had 32kiB I+32kiB D, 512kiB L2 and more complex branch prediction for the longer 13 stage pipeline. The fab process improvements of a decade make a big difference as well. The 2012 Cortex-M0+ with just 2 pipeline stages has 74% of the DMIPS/MHz performance of the 2002 ARM11 with 8 stages due to process improvements even though the shorter pipeline limits the max frequency to roughly 1/3 of the ARM11. It's pretty easy to see that more pipeline stages allows higher frequencies from the chart above (the 68060 advantage of a longer pipeline than the competition did not allow it to be clocked up though). The simple in-order ARM RISC cores are weak sauce and it was complex OoO cores like the Cortex-A9 where ARM started to have some performance while the simplicity and power advantage disappears.

Karlos Quote:

The lesson to take from Apple, for all the crap I don't like, is that they have mastered vertical integration. If RISC-V overtakes ARM, they'll probably migrate to that.

RISC-V is about staying simple and flexible to scale smaller for the embedded market. Apple may use RISC-V highly customized deeply embedded cores but I doubt they will replace the main CPU cores of their SoCs anytime soon.

Apple Plans To Build Chips Using RISC-V Architecture
https://techlog360.com/apple-plans-using-risc-v-architecture/

Last edited by matthey on 25-Apr-2022 at 11:27 PM.

@matthey

To be clear, I don't think the RISC/CISC concept applies today at all. The architectural merits of each have been adopted by all major vendors still in the game. I tend to categorise modern implementations as load/store or not, since almost every other attribute is not unique to either implementation. However, when the term first became common, the idea that it encapsulated simple instructions that could execute with minimal resources, in conjunction with a load/store architecture to be run on simplified hardware is what what people thought of as RISC. Simple instructions that can achieve per cycle (or better) throughput.

Comparing the 68060 to contemporary ARM designs is a bit disingenuous given the different use cases at the time. You wouldn't put a 68060 in something designed for low power application. Nobody talked about performance per watt much then but the concept was still well understood. If it ran on a battery you probably chose ARM.

_________________
Doing stupid things for fun...

@all

ARM conquers the World!

Amiga conquer ARM!

Extremeley energy efficient ARM Amigas!

That's it!

Throw away your PCs and get them Recyceld and go UBER Raspberry Pi and UBER ARM Amigas!

Amiga Forever!

Save the Planet NOW!

Last edited by QBit on 26-Apr-2022 at 01:48 AM.
Last edited by QBit on 26-Apr-2022 at 01:47 AM.
Last edited by QBit on 26-Apr-2022 at 01:37 AM.

Karlos Quote:

To be clear, I don't think the RISC/CISC concept applies today at all. The architectural merits of each have been adopted by all major vendors still in the game. I tend to categorise modern implementations as load/store or not, since almost every other attribute is not unique to either implementation. However, when the term first became common, the idea that it encapsulated simple instructions that could execute with minimal resources, in conjunction with a load/store architecture to be run on simplified hardware is what what people thought of as RISC. Simple instructions that can achieve per cycle (or better) throughput.

I agree that the RISC/CISC concept does not apply today. CISC was made a bad word by propaganda and most RISC principals were abandoned. The only way to differentiate RISC anymore is by load/store memory accesses.

RISC principals
1. load/store memory access (RISC maintains)
2. single cycle throughput instruction pipeline (RISC won, CISC improved)
3. minimal reduced instruction set (RISC abandoned, CISC won)
4. minimal addressing modes (RISC abandoned, CISC won)
5. no microcode (RISC abandoned, CISC won)
6. clock cycle time should be minimized (RISC abandoned, CISC won)
7. fixed length encoding (RISC abandoned, CISC won)

RISC has 1 win from introducing a single cycle throughput instruction pipeline but CISC adopted and improved it by pipelining reg-mem instructions which access memory. The improvement is not possible with load/store memory accesses giving a performance disadvantage. CISC has 5 wins as 5 of the RISC principles are not even considered RISC principles anymore. The remaining defining characteristic of RISC is load/store memory access which remains more popular than reg-mem for new designs.

load/store advantages
+ uses less encoding space allowing for more instructions and registers (but uses more registers)
+ simpler and reduced design time

reg-mem advantages
+ high performance memory accesses, reg-mem instructions pipelined & load-use stalls avoided
+ fewer registers used
+ fewer instructions executed
+ better code density

The 68060 outperformed the Pentium which went on to rule multiple high performance CPU markets. One of the biggest advantages of the x86-64 architecture is the excellent memory performance yet nobody important seems to design reg-mem cores anymore.

Karlos Quote:

Comparing the 68060 to contemporary ARM designs is a bit disingenuous given the different use cases at the time. You wouldn't put a 68060 in something designed for low power application. Nobody talked about performance per watt much then but the concept was still well understood. If it ran on a battery you probably chose ARM.

How about Acorn RISC PC vs 68k Amiga or 3DO vs Amiga CD32? The 68k and ARM competed although not often the 68060 with ARM. The 68060 was used where high performance was needed and it was several years after the 68060 was practically abandoned before ARM had a competitor for it. The 68060@50MHz was only 3.9W max and average power draw would be lower. It was significantly lower power than the Pentium.

Pentium@75MHz 80502, 3.3V, 9.5W max
68060@75MHz 3.3V, ~5.5W max*

* estimate based on 68060@50MHz 3.9W max, 68060@66MHz 4.9W max

The 1994 ARM710@40MHz used in the Acorn RISC PC shows .425W on wikichips but I don't know if that is max, average, TDP etc. It is no doubt impressively low power but it is still a much weaker core at 0.9 DMIPS/MHz compared to the 68060 1.52 DMIPS/MHz. The 68060 was low enough power to be used in a laptop and would have been a better choice for a Mac laptop than an early PPC CPU. ARM was likely a better choice for the simplest battery operated devices that didn't require much processing power. Even PDAs and calculators have sometimes used the CMOS 68000 (Ti-89 and Ti-89 Titanium graphing calculators) and DragonBall (Palm OS PDAs) though.

https://en.wikichip.org/wiki/vti/vl86cx/vy86c710

Last edited by matthey on 26-Apr-2022 at 04:23 AM.
Last edited by matthey on 26-Apr-2022 at 04:22 AM.
Last edited by matthey on 26-Apr-2022 at 04:16 AM.