@Karlos

Quote:

If the process runs long enough, you can just measure it using the standard C routines which are accurate enough, even under emulation. We are talking about very obvious factors difference in performance. If all versions are affected the way mine was, you can't possibly miss it.

You might also be interested In buddhabrot fractal which differs slightly - it does not compute how many steps are necerrary to escape, but instead “draws” the escape path itself. Not only it stresses the cpu with computation but also memory since it has to store the paths and then draw them through saturated addition in framebuffer.

You can find AmigaOS example in my Emu68-tools package :)

@michalsc

I chose Mandelbrot set area calculation because I wanted something that wouldn't depend on memory access and could fit in the CPU cache effectively, but equally would do a lot of FP as I was trying to figure out if the FPU was as slow as it seemed.

_________________
Doing stupid things for fun...

@Hypex

Quote:

For FPU code there is a 64 bit FPU in PPC which underlying x87 code simulating it should handle easy in 80 bit or now days in SSE. That's if the code is optimised to convert FPU code to FPU code or best match.

AVX-512's 32 FPRs are a better fit for PowerPC FPU's 32 FPRs. Intel has backed out of AVX-512 for its mainstream later AlderLake (fused AVX-512 off), Raptor Lake, and Meteor Lake SKUs.

AMD's Zen 4 supports AVX-512 with optional Intel Ice Lake and BF16 extensions.

X64 SSE / AVX has 16 FPRs.

AVX 2 FMA3 is a better fit with PowerPC FPU's FMA3.

PowerPC 970 has a superior scalar FPU and 128-bit SIMD, but it has flaws e.g. slower GPRs and FPRs data exchange that is prevalent in PC games and X86 has superior code density.

From my POV, PowerPC 970 FPU (FMA3) and VMX are like AVX-512 (32 FPRs, FMA3) with pack math 128-bit wide. For the early 2000s, the superior FPU hardware didn't win the market.

POWER9's specs are reasonable and competitive in the 14 nm process node era. Platform cost is not competitive. IF POWER9 is available for the AM4 platform, I would have purchased a POWER9 CPU.

Last edited by Hammer on 10-Jan-2024 at 02:28 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)

@Hypex

Quote:

Hypex wrote: @agami Quote: Every mainstream(ish) platform that has been around for more than a decade is essentially and ostensibly a "Ship of Theseus". Sure, but they have a parent company, the Amiga brand hasn't had any since Commodore. The Amiga code base lacks continuance. Without any original developers working on them. The only leg it has to stand on is the original sources from which modern code is descended from. Take away that code and you take away any genuine link to AmigaOS. Otherwise people would have forgot about OS4 years ago and AROS would have taken over. Which builds from the ground up a portable replacement using a new crew of Theseus. Or replaced Windows with ReactOS.

ReactOS doesn't run modern Windows 7/8/10/11 games like HoloISO which is directly based from SteamOS 3.x (Steam Deck) which is based on Arch Linux.

ReactOS is stuck in the Windows NT 4/2000 era.

_________________
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)

@Kronos

Quote:

We were talking about a no-Amiga-chipset Amiga which yes could have been done is SW. 3rd party (MacroSystems) did it without having access to any sources so no reason ESCOM-Amiga couldn't have done it instead of wasting resources to put rotten meat (AGA) into new sausages (A1200 repro, Walker).

Amiga's core revenue market is associated with games. DraCo is a tiny market-niche workstation solution and it died like market niche MIDI port equipped Atari ST/STE/Falcon.

The "rotten meat" supporter like Apollo-Core has more than 10,000 V2/V4 units sold which beats DraCo.

Gaming PC-centric hardware vendors (NVIDIA, ATI/AMD, Intel) with larger economies of scale have defeated non-X86 market niche workstation vendors like SGI, SUN SPARC/GX 3D, MIPS, DEC Alpha, HP RISC, AIM PowerPC and 'etc'.

Last edited by Hammer on 10-Jan-2024 at 02:59 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

As always missing the mark.

C= sold millions of low margin A500s in 90/91 making little to no money..
They did sell the few A1200s they made and went down. All while very few newsworthy Amiga 1st games were released.

Years later ESCOM failed to sell 200k A1200s at the bottom of the market while games were either rehashed A500 concepts or mediocre PC ports.

Hence they had to fail.


Apollo selling 10k sounds nice, that is only fraction of what would be needed for a viable market and those numbers or doomed to drop once they hype wears down and saturation sets in.
Specific SW? 3 games and a bunch of emulators far less than what we got in the early years of NG.

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

@Kronos

Quote:

Kronos wrote: @Hammer As always missing the mark. C= sold millions of low margin A500s in 90/91 making little to no money..

False.

https://arstechnica.com/gaming/2017/01/a-history-of-the-amiga-part-10-the-downfall-of-commodore/

Commodore's profit and loss
1988 to 1989 = profit,
1989 to 1990 = tiny loss
1990 to 1991 = profit, +800,000 Amiga unit sales at end of 1990.
1991 to 1992 = profit, 1 million Amiga unit sales at end of 1991.
1992 to 1993 = large loss

Commodore canceled its best-selling Amiga 500 model in 1992 and incurred a large-scale loss at the start of 1993. LOL

The inferior Amiga 600 couldn't replace the best-selling Amiga 500 and couldn't stop the bleeding.

Commodore Germany's PC division couldn't stop the bleeding.

The Amiga 2000/3000/4000 Video Toaster market niche wasn't enough to save the Commodore.

At the start of 1994, Commodore was almost breaking even, but EPA and XOR patent problems pushed it into oblivion. There were large-scale layoffs from 1993 to 1994.

Last edited by Hammer on 10-Jan-2024 at 06:53 AM.
Last edited by Hammer on 10-Jan-2024 at 06:45 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

Yey, not making products or having any employees while selling of inventory is niece way of breaking even.

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

This is all woefully irrelevant.

Can we get back to drubbing PPC as a choice in 2024 now please?

_________________
Doing stupid things for fun...

@kolla

Lack of proper KVM support. A Power9 server may support it. But the average G4 or X5000 tends to lack proper support and usually emulates the PPC on the PPC.

@Kronos

Quote:

MIDI port is just a fancy serial port. The only reason it was so essential for the ST is that it had by default.

That put it in the map. Most musicians in the 80's make mention of Atari and Cubase. The Amiga missed out on Cubase and you can't use ProTracker alone to make a real song.

Quote:

That ship had long sailed for Amiga as the existing install did not have one.

Adding MIDI would have helped to improve the sound ability without touching Paula.

Quote:

"The user base" wanted better HW, most didn't care about who made they chips used.

Except the gamers who only like original hardware.

Quote:

We were talking about a no-Amiga-chipset Amiga which yes could have been done is SW.

An Amiga compatible PC floppy drive cannot be controlled in software alone with a standard controller.

Quote:

3rd party (MacroSystems) did it without having access to any sources so no reason ESCOM-Amiga couldn't have done it instead of wasting resources to put rotten meat (AGA) into new sausages (A1200 repro, Walker).

The Amiga memory map was common knowledge.

@Hypex

We lifted off on a spot about what ESCOM could have done better/different.

Adding a MIDI port at that point was pointless.

OCS/AGA games on a naked A1200 felt like a decade old compared to PC hence few good ones were made and even less bought as most of the penny pinching MagicPack customers had no budget for SW.

-> Redoing obsolete HW was doomed to fail and would only have made sense if they could be made super cheap and be sold with huge margins. Which wasn't the case.

A "cost reduced Draco", so 68030/40 with ISA slost and off the shelf PC parts would have been compatible enough with productivity SW and it would have been close enough to a PC so they we might have gotten more game ports at a better quality then what we got for AGA.

99% chance of failure, but still better than the 100% they had the A1200.

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

@Hammer

Quote:

Built-in MIDI port is a market niche and it didn't save Atari's Digital Research GEM-based TOS platform.

It didn't need too. it just needed to run Cubase, as basic as it looked.

Quote:

My A1200-PiStorm also has retro MacOS 68K with RTG.

I didn't think to try that. But RTG complicates my Pi. I'm always missing a cable to go from native to RTG. By comparison my A4000 is luxury. Just need one cable to the monitor and the hardware does the rest.

Quote:

AmigaOS's two-button mouse design is closer to Windows 11's two-button mouse design, hence I'm comfortable with flipping between AmigaOS and Windows 11.

Well, I upgraded the laptop to 11, but I still find Windows just as annoying as the previous version.

Quote:

Amiga's Floppy IO performance is linked with Paula. When I owned the A3000, its half-speed high-density floppy drive was satisfactory since SCSI hard disk performance was good.

It is linked to Paula. But the drives needed hacking. Would have made more sense using newer fab processes allowing greater speed so it could keep up.

Quote:

My A1200's PCMCIA has laptop's WiFi card and it works fine.

A rarity to find these days though I haven't looked in a while. I found it hard 20 years to even find a card that was 16 bit. I've still got a collection of ethernet cards that if they do fit don't have a driver.

@Hypex

Quote:

Hypex wrote: @kolla Lack of proper KVM support. A Power9 server may support it. But the average G4 or X5000 tends to lack proper support and usually emulates the PPC on the PPC.

https://www.linux-kvm.org/page/Processor_support#PowerPC:

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@Kronos

Quote:

99% chance of failure, but still better than the 100% they had the A1200.

AGA was good for some things, notably T-Zero, Slam Tilt, Napalm and Banshee! Those did differentiate the Amiga enough from the PC to warrant some value! It's a shame mist people wanted Doom but didn't wish to upgrade their Amigas to stand a chance of having a SKU capable of playing it!

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

@BigD

Only haters bash the chipset. It's an easy target, but let's be honest AGA, flawed as it was, sold more Amiga than any PPC ever has

_________________
Doing stupid things for fun...

@BigD
Quote:

BigD wrote: AGA was good for some things, notably T-Zero, Slam Tilt, Napalm and Banshee! !

|[Hammer-mode]

In 96 I had an AMD K5@133 Matrox-Mystique and played WingCommanderIV.
Yep felt and looked about a decade ahead of those AGA games

|[/Hammer-mode]

@Karlos
Don't have an issue with the chipset, just honest enough to see that it was obsolete by the time AGA came around and that C= couldn't wouldn't and probraly shouldn't have fixed it.
If they had done everything 2 years earlier, so AGA in 90 and AAA in 93 it might have been o.k.

AGA was already "retro" von ESCOM came around and back then nobody wanted the Amiga to be retro.

Hence sales sucked.

How many PPC units sold into that dead market is irrelevant, they tried to bring back Amiga as non retro system which didn't have a chance but was at least a try.

Nowadays their is a hype about retro, emulated retro, fake retro and alt-history retro all nice and funny but I just never got a hard one over any HW past or present.

Last edited by Kronos on 10-Jan-2024 at 05:05 PM.

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

@Kronos

Yeah, hater would say that :P

J/K

I wasn't really expecting a response to that obvious sh1tpost

_________________
Doing stupid things for fun...

@Hammer

Quote:

The Amiga 2000/3000/4000 Video Toaster market niche wasn't enough to save the Commodore.

Apple endured better fate while relying on just as niche markets.
I know that quikpak and similar companies were still selling A4000T to that market in the late 90s and Amigas were used on smaller TV stations even well after that. It certainly wasn't enough to support the entire platform, but it was very expensive hardware and probably far more profitable per unit than selling A500s and like.

In the second half of the 90s I'd argue, with hindsight, the best market for anyone that ended up with the Commodore/Amiga IP was handhelds and such.

_________________

@WolfToTheMoon

Apple had several niche markets, CBM Amiga only really had two. And for both Amiga niches, CBM was not selling "the product", others were (SCALA and NewTek)

Eventually, Apple also had the benefit of Steve Jobs and his... hiatus projects.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC