@Kronos

Quote:

Kronos wrote: @Hammer Again 2kb off of topic PC parts babble intervened with some bad takes. .

Again, engineering talent matters, not just money. I cited examples.

Apple M1's key engineering talent is at Qualcomm e.g. Oryon CPU preview benchmarks are pretty good.

Samsung has key engineers from AMD's Bulldozer and the results weren't good.

Both Samsung and Apple have lots of money.

Quote:

The programable logic of various kinds used in CPU cards is just a fraction of would have been needed to emulate AGA or better GFX. /quote] The Amiga market wasn't developed into a clone hardware standard with a self-survival redundancy. IBM set the standard, and PC cloners followed until NEC created VESA. [quote] A\Box was vaporware and is as such irrelevant. There is a reason why their next (and still vaporware) design the "Pre\Box" was much more conventual with a PCI based (of the shelf) GFX card.

Phase 5's excuse was that the Amiga market wasn't large enough for A/Box R&D ROI.

Last edited by Hammer on 07-Jan-2024 at 08:35 AM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Hammer

Quote:

Hammer wrote: Phase 5's excuse

At least you got that one right

ESCOM did not have the resources to pull it of and where smart enough to try.
Phase5 were a bunch of skilled guys with limited funding (and no way to secure big $),

Apollo has done the "easy" part 20 years after the fact but clearly only have tiny part of the resources and skills needed to make that happen in the 90s.


Billion and even trillion $ companies poaching talent to design chips today is irrelevant.
Startups having access to venture capital and the right contacts is irrelevant.

Neither operate at the same tiny scale as anything Amiga has since C= went bust hence noone is putting that kind of money into "Amiga".

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

@Kronos

Quote:

Oh boy....

Quote:

Petro's big "plan" was produce more obsolete A1200s 3 years after they triggered a "meh" response on initial release.

It was a good idea to get it selling again but it really needed an update. The floppy interface needed converting from Shugart to PC with disk ready logic. The CIA chips needed updating with a compatible replacement that provided faster serial, parallel and floppy transfer. Especially for using more common and hackless HD floppy drives at proper speed. That's not getting into AGA though an on board simple 8 bit RTG with a real packed productivity mode would have helped. Which really needed a matching 16 bit mode to compete with the Falcon. Most of those would have been considered a basic upgrade but even that would have needed time and resources.

Quote:

Made a bit over 200000 units, failed to sell them all and came up with the conclusion that Amiga needed another AGA machine with a slightly less obsolete CPU and plenty of extra compatibility issues backed in (the Walker).

The Amiga Hoover to suck Amiga users back in. It didn't look like an Amiga. But with a 68030 as standard it also needed to have AAA on board and working for a real A1200 follow up.

Quote:

If PPC had been center stage he would have gone after proDad who had an alpha OS and plans for legacy free HW which could have been something like early MorphOS+Pegasos only 7 years earlier.

All their was is coprocessor kernels for a speed boost. But no full OS. It may have been put in the spot light but by comparison it was back stage.

...

Quote:

Only "development" that was done was the incomplete floppy fix, without it even more games would have failed making the A1200 an even harder sale....

I installed one of those Escom floppy drives in my A1200 and lost my drive ready line. Broke a few games. And I couldn't get State of the art working either.

Last edited by Hypex on 07-Jan-2024 at 01:50 PM.

@agami

Quote:

agami wrote: @Karlos Just thought I'd check, because your logical and dignified response indicates that you are either unaware of the above fact, or for some unfathomable reason you've decided to give him yet another chance at civil discourse.

It's more for the benefit of anyone else that might be thinking along similar lines. I honestly think properly compatible PPC emulation on ARM is likely to disappoint. I may be wrong, but my experience of PPC emulation on x64 has not really enthused me.

_________________
Doing stupid things for fun...

@Karlos

Quote:

You can have your cake and eat it. Dirty self modifying 68K legacy code still runs, it just takes a performance hit because when there are code changes, the translation cache is invalidated for any traces that span it. It's not as if this doesn't work just fine in existing JIT emulations - just the performance penalty. Except, even with the penalty and even on a 603e, you are dealing with such an edge case for any system friendly code that you won't encounter it most of the time anyway.

They must have implemented this logic in OS4. A static emulator has no such complications.

Quote:

However, "clean" code will not have this problem and by by sticking to a well documented 68020 subset, you can tune that for best performance. You may even be able, with some effort and strict guidelines, get to a point where translation can be done AOT.

I think AOT would be a good idea. Or an ALT. At load time.

Quote:

Well most things that required WarpOS have moved on to OS4 in one way or another and should therefore be in a recompilable state. As it goes for WO2097, you don't need to worry, there's a 68K port in the works already.

Most, but just like on the real thing, the most important are the games. Only a few have OS4 ports. Given WipeOut needed a CPU running about 160Mhz and hardware 3d it's going to need a super 68K to run on. And relying on software 3d isn't the Amiga way.

Quote:

I see a lot less value in providing backwards compatibility with PPC primarily because most PPC software that is maintained still, ought to be recompilable - especially to another 32-bit/Big Endian target.

More value would be in the classic PPC ports.

...

Quote:

I may be wrong, but my experience of PPC emulation on x64 has not really enthused me.

It's funny that. PPC models we have now can barely touch a PPC from even 15 years ago. It's almost 20 since the G5. x64 has increased massively in power. And by comparison PPC is a simpler design. In the least instruction decoding is simpler. I would have expected x64 to kill PPC in emulation and run rings around an old Ghz PPC.

Last edited by Hypex on 07-Jan-2024 at 02:01 PM.

@Hammer

Quote:

Note why I stated "other RISC and X86 competitions".

At the time X86 lacked bi endian support. But it can be worked around with a custom compiler. Which has been done before.

Quote:

ARMv3 up to ARMv5 only supports BE-32 e.g. ARM6 with 12 Mhz and 10 MIPS. ARMv3 was introduced in 1991.

A few years before Amiga PPC.

Quote:

ARMv6 introduces BE-8 e.g. ARM11 with 532–665 MHz. Introduced in October 2001.

A few years after classic PPC and around when PPC was continued for OS4.

@agami

Quote:

Having the ownership question taken out of the equation frees up any reticent and reluctant developer to port and/or reimplement any functions without fear of potential litigation.

That's already been done with Enhancer 54. In fact Enhancer now replaces commands. But, when you test replacements and they act faulty, you can tell they aren't genuine. They then need more work to work as expected. But at that point they aren't original.

Quote:

The same loss of enthusiasm can be seen in score-topping arcade game players when the machine is periodically and infrequently reset, wiping the leaderboard.

That would be annoying. Now days taking a screen shot is easy enough. But reinventing the wheel is not enjoyable.

@Hypex

"it needed this, it needed that"

As soon as it would have meant making new chips, let alone changing any design it needed that NOT to be done.

It needed work on the SW as it was quite clear that "two much HW" was the problem all along.

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

@Hypex

Quote:

I think AOT would be a good idea. Or an ALT. At load time.

You lose a key benefit of JIT, which is folding out of runtime invariant conditionals. Also, most applications, most of the time are not compute bound, they are waiting for something. It's better to interpret the code and gather some information about what needs to be translated. That could also be guided by some sort of hint mechanisms but you only have to look at trance ans Emu68 to see that AOT is not really essential and even potentially undesirable (e.g. statically compiling a bunch of conditional code that only goes down one particular path at runtime.

_________________
Doing stupid things for fun...

@Hypex

Quote:

They then need more work to work as expected. But at that point they aren't original.

I have to agree, that bugging me as well, while they have technical fixed all the bugs and works as it should. But the text output is not the same, so there is possibility it can break in scripts, if text output is used for other things, then reading it, like piping into another command for example.

Last edited by NutsAboutAmiga on 07-Jan-2024 at 04:32 PM.

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

@Thread
Quote:

A\Box was vaporware ...

They really worked on it, (I knew people who knew people), they had the knowledge for GFX, CPU-Cards with RAM and HD ... but they overestimated their Know-How (for proprietary GFX and busses), at the time they announced the Pre\Box they might have gone back to something like the standard designs developed from the AIM initiative, for IBM and Apple RISC systems . Others like PIOS did it or like it was sold from UMAX as 'Mac compatible'.

When P5 switched to Pre\Box it was already to late for them, all the hopes for PIOS and others where destroyed when Steve Jobs returned and buried the Mac compatibles.

After Apple stopped, no one build PPC motherboards for desktops any more, this CPU drifted into obscurity.

Last edited by OneTimer1 on 07-Jan-2024 at 04:47 PM.

@Kronos

Quote:

Kronos wrote: @OneTimer1 Back into w##nking of to obscure alt history...

You are right, it's only questioning what could have been.

Even if Amiga Tech. had survived, they would have buried the ECS/AGA in the next years, that's why I'm always using Apple as example, they survived and the later PPC Macs had no 68k emulation as default.

Edith:
would -> what

Last edited by OneTimer1 on 07-Jan-2024 at 05:24 PM.

@OneTimer1

Quote:

OneTimer1 wrote: but they overestimated their Know-How (for proprietary GFX and busses), .

If they really thought they could pull that of they either had some big partner in the back hand (something along the line of a game console that never was) or had completely lost the plot.

GFX had turned to proper GPUs at that time and anything that was "just" a GFX chip was obsolete.

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

@OneTimer1

Quote:

the later PPC Macs had no 68k emulation as default.

That is only kinda correct, 68k vanished as support for non-OSX PPC SW was disabled.
Looooooooooong after the last 68k Mac and long after any Apple specific GFX chip.

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

@Kronos

PowerPC support was as early as MacOS7.6.x, MacOS8, back then OS booted on JIT compiler.
The first PowerPC only OS was MacOS9.x, it also offered backward compatible with 680x0.
Then they moved to MacOS X, MacOS X has more in common with BSD/UNIX then MacOS9.

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

@NutsAboutAmiga

Erm, yes and 2x2=4.

Just so you know.

OSX10.4 and below still had the "Classic" environment which allowed OS9 and earlier PPC and 68k SW to run.

10.5 removed that and only supported proper OSX SW, either PPC or "universal" (PPC+x86 in 1 file).

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

@Hypex

Quote:

Hypex wrote: @agami That's already been done with Enhancer 54. In fact Enhancer now replaces commands. But, when you test replacements and they act faulty, you can tell they aren't genuine. They then need more work to work as expected. But at that point they aren't original.

But they don't need to be the "original"

Every mainstream(ish) platform that has been around for more than a decade is essentially and ostensibly a "Ship of Theseus".

_________________
All the way, with 68k

@Hypex

Quote:

x64 has increased massively in power. And by comparison PPC is a simpler design. In the least instruction decoding is simpler. I would have expected x64 to kill PPC in emulation and run rings around an old Ghz PPC

So I have to preface this with "in my experience". For various reasons the PPC is a bit of a PITA to emulate well, at least 603/604. I had this same conversation recently with a friend. The experience is limited to OS4/WarpOS in UAE, but it seems to depend on integer software emulation for the FPU for best compatibility. There may also be mmu stuff clogging it up.

Maybe it's just that particular combination, I dunno.

_________________
Doing stupid things for fun...

@Kronos

Quote:

ESCOM did not have the resources to pull it of and where smart enough to try.

Did you expect technical miracles from Commodore Germany? This is the wrong Commodore division to evolve the classic Amiga.

https://i.imgur.com/jmUEnpF.jpeg
Escom clone PC motherboard with chipset design from SIS.

ELITEGROUP SI54P AIO motherboard, Revision 2.1, Socket 5 board with SiS 85C501 /-502 /-503 chipset.

http://ftp.bambi-amiga.co.uk/amigahistory/escomhist.html
During 1994, Escom sold 410,000 PC clones.
Around 1995, Escom reached £200 million annual turnover and it was short-lived. Unlike DELL, Escom expanded into a high-overhead cost bricks-and-mortar store chain business model. Escom went bankrupt in July 1996.

For comparison, DELL's revenues in Jan 1994 were USD $2.8 billion and in Jan 1995 were $3.4 billion. DELL has bid for Commodore in 1994.

Quote:

Phase5 were a bunch of skilled guys with limited funding (and no way to secure big $),

Phase 5's R&D ability is still inferior to 3DO's chip design R&D ability.

With the backing of Samsung and Pansonic, the 3DO team still failed since they hadn't learned the lesson from beyond game console pricing A1000. Commodore has cost reduced A1000 into A500.

Furthermore, the 3DO team selected the failed quadrilateral polygon-based four-sided sprite engine. Sega Saturn and NVIDIA NV1 have similar quadrilateral polygon mistakes.

NVIDIA has US Silicon Valley venture capitalist support which enables NVIDIA to survive the NV1 mistake and has key engineering talent from SUN GX and HP to quickly evolve NV2's quadrilateral polygon mistake into NV3's triangle polygon-based 3D accelerator. NVIDIA already has software driver teams for OpenGL/Direct3D during NV1/NV2.

Sony's Kutaragi and CEO Ohga leadership led to PlayStation's success.

The sting:
Sony's purchase of Psygnosis marked another turning point for the PlayStation as it played a vital role in creating the console's development kits. While Sony had provided MIPS R4000-based Sony NEWS workstations for PlayStation development, Psygnosis employees disliked the thought of developing on these expensive workstations and asked Bristol-based SN Systems to create an alternative PC-based development system.

Ian Hetherington, Psygnosis' co-founder, was disappointed after receiving early builds of the PlayStation and recalled that the console "was not fit for purpose" until his team got involved with it.[67] Hetherington frequently clashed with Sony executives over broader ideas; at one point it was suggested that a television with a built-in PlayStation be produced.[68] In the months leading up to the PlayStation's launch, Psygnosis had around 500 full-time staff working on games and assisting with software development.

During a 1993 meeting, Commodore International's CEO insulted Psygnosis.
----
Rise of 3DFX...

3DFX was founded in the spring of 1994 and it has three important engineers, Scott Sellers, Ross Smith and Gary Tarolli, who came from Silicon Graphics. Gordon Campbell, coming after pioneering fabless hardware design at Chips&Technologies. 3DFX received $5.5 Million in initial funding and began to work on the first architecture named Voodoo.

Both NVIDIA's RiVA 128 and 3DFX Voodoo are 128-bit 3D accelerators.

Silicon Graphics engineers who were involved in Nintendo 64 have founded a company, ArtX which was purchased by ATI. ArtX and ATI's existing Radeon design teams designed the Radeon 9700 series.

Silicon Graphics was bleeding engineering talent.

Last edited by Hammer on 09-Jan-2024 at 12:48 AM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Karlos

Quote:

Let's not forget the fact it looked like a retrofuturistic vacuum cleaner.