@Karlos

Quote:

Karlos wrote: @Kronos Nobody is claiming AGA or the 020 were ground breaking in 1992. They weren't and there are pages in every forum about how too little to late they were. You know it. I know it. You know I know it and I know you know I know it However, the capabilities of the average PC are grossly overestimated in a distorted recollection of the era. Average PCs like average anything do not stand out in your memory, only the exceptional things. The vast majority of PCs that everyday people owned in 1992 were absolutely shit. Many were still grinding away on 286, a significant proportion still didn't have VGA or any audio capabilities at all. Why? Because you are dealing with machines people already had, sometimes for a few years and they generally weren't used for anything that needed better visuals or sound.

The "average PC" argument is flawed when minority gaming PC outnumbers A1200/A4000 unit sales.

IDsoftware sold 4 million copies of PC Doom and Doom 2. IDsoftware estimated a further 15 million Doom copies.

Can you say the Amiga AGA platform with 68030 @ 50 Mhz has a 1 million install base?

You're not going to win a "gaming PC" sales game.

_________________
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:

Karlos wrote: "So anyway, I bought the PC..." Didn't happen of the year award. Doom came out, late 93, coinciding with a relative decrease in the cost of entry level 486 as it ran through numerous clockspeed increases and then the appearance of the Pentium. Graphics and sound capabilities improved, "CDRom/Multimedia" became all the rage. By the end of 1993, there was no question that things had moved on rapidly. Except for one. None of the new shiny PC stuff ran the software I wanted to use. And so, I expanded the machine, part by part. First with a HD and RAM expansion, then later with one of the first 040 cards. I needed to use a few MS applications by then, but I ran them just fine in shapeshifter. When DoomAttack was released I was amused to discover it ran at perfectly playable frame rates on AGA on my 040, complete with multichannel sound. Which means that had it have been available in 1993, the A4000/040 could probably have ran it too. Towards the end of the decade, I had a BPPC/BVPPC, CGX 4.2, OS3.5, network. I still ran shapeshifter for the annoying Uni applications I needed to run, while the rest of my time was messing around in OctaMED, programming and generally enjoying the machine I had. It was only formally retired as my primary machine ~2007. I just didn't have any PC envy the whole time.

John Carmack is aware of the A4000/040 with RTG card capability, but his argument is the installed base with Amiga's 68040+RTG specification.

68040's FPU is useless for Doom.

Commodore didn't offer a cost-reduced CD32 (AGA Southbridge small single board) with a single Zorro 3 slot (for RTG card), 68LC040 @ 25 MHz, and 4MB 32-bit Fast RAM.

My proposed single-slot Zorro 3 Amiga 2400/LC040 prevents it from competing against big-box Amiga 4000s. This is my counter against sub $1000 USD H2 1993 Apple Quadra 605 (680LC40) SKU.

There's a mass production problem with Amiga's Zorro graphics card market.

Mass-produced Zorro III install base helps Amiga's 3rd party RTG card market's "chicken and egg" problem e.g. Phase 5 can't risk manufacturing 500,000 CyberGraphics for 150,000 Amiga 2000/3000/4000 market.

AGA's smaller 3rd party market killed GVP's business.

If Commodore sold Amiga chipsets to 3rd party OEM box shipper cloners (i.e. AMD/NVIDIA business model), they could have covered Amiga 2400 specifications.

AMD, NVIDIA, Intel, and Qualcomm engage multiple allied 3rd parties to manufacture end products and ship their CPU, GPU, and chipsets.

Last edited by Hammer on 25-Jan-2024 at 04:17 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: The "average PC" argument is flawed when minority gaming PC outnumbers A1200/A4000 unit sales. IDsoftware sold 4 million copies of PC Doom and Doom 2. IDsoftware estimated a further 15 million Doom copies. Can you say the Amiga AGA platform with 68030 @ 50 Mhz has a 1 million install base? You're not going to win a "gaming PC" sales game.

True, but I argue that the situation you've described above was not reality in most markets until 1994.

In 1992 and for most of 1993
1. In the largest PC market, USA, the gaming market was dominated by consoles.
2. By most indications, UK was very similar to AU, or vice versa, and there too gaming was dominated by game consoles (to a lesser degree than USA) and "home" computers from Commodore and Atari.
3. Gaming in the rest of west and east Europe was also much more skewed toward "home" computers from the likes of Commodore and Atari, which were more compatible with the standards of living in those countries.
4. A PC sold which is more capable of running games than an A1200 is not automatically a "gaming PC". So even if more 486 PCs were sold to deep pocketed early adopters in 1992 and 1993 than total A1200 sales, fewer of them were intended for gaming than the A1200.
I wouldn't be surprised if Apple sold more mac Performa 450s (LC III) units than A1200s in 1993, but that still didn't make it a target for FPS game developers.

id Software didn't sell 4M Doom units in December 1993. And not even when Doom II came out in May 1994. It took some time for that number to be reached. Thanks to a confluence of contributing factors all culminating in 1994.

Last edited by agami on 25-Jan-2024 at 07:18 AM.

_________________
All the way, with 68k

@agami

Quote:

agami wrote: @Hammer Quote: Hammer wrote: The "average PC" argument is flawed when minority gaming PC outnumbers A1200/A4000 unit sales. IDsoftware sold 4 million copies of PC Doom and Doom 2. IDsoftware estimated a further 15 million Doom copies. Can you say the Amiga AGA platform with 68030 @ 50 Mhz has a 1 million install base? You're not going to win a "gaming PC" sales game. True, but I argue that the situation you've described above was not reality in most markets until 1994. In 1992 and for most of 1993 1. In the largest PC market, USA, the gaming market was dominated by consoles.

Consoles are not monolithic platforms.

Mainstream 16-bit game consoles from 1988 to 1993 didn't deliver PC's "full 32-bit" gaming experiences. This factor is important.

Commodore gimped A1200's 14 Mhz 32-bit 68EC020 CPU and 16-bit (Alice) / 32-bit (Lisa) AGA with a shared bus design.

Amiga's "32-bit" didn't deliver PC's "full 32-bit" gaming experiences.

Amiga's "fake 32-bit" AGA platforms were squeezed out by SNES, Genesis, and failing prices full 32-bit gaming PCs.

Gaming PCs delivering a "full 32-bit" gaming experience are important to differentiate themselves from 16-bit consoles.

https://www.statista.com/statistics/1101900/unit-sales-snes-region/
Super Nintendo Entertainment System (SNES) lifetime unit sales worldwide as of September 2023
North America is 22.88 million.
Japan is 17.17 million
Europe is 8.15 million
The rest of the World is 0.9 million

https://vgsales.fandom.com/wiki/Fourth_generation_of_video_games
SNES install base
1990
1991, 2 million
1992, 6.9 million (+4.9 million)
1993, 11.3 million (+4.4 million)

For 1992, Doom-capable "gaming PC" had extra time to build its install base
1. Intel's 32-bit 486 was released on April 10, 1989.
Intel 486SX-25 and 486SX-33 were released in 1991.

2. Intel full 32-bit 386 was released in 1985 and mass production in 1986.
Intel's 386 CPUs had topped out at 33 MHz in 1989.
AMD introduced 386DX-40 MHz versions in March 1991.

https://arstechnica.com/features/2005/12/total-share/7/

The golden age of PC gaming had arrived, with classics like Wing Commander, DOOM, Ultima 7 and System Shock pushing the boundaries of computer entertainment. At the same time, the PC became easier to use with the release of Windows 3 in 1990 and 3.1 in 1992. The latter operating system proved incredibly popular, pushing PC sales back from their minislump in 1991. By 1994 PCs and clones were selling at the incredible rate of 37 million units per year.


-------------

Pentium was released in 1993 and Sony PSX was released in Q4 1994 in Japan and Q4 1995 in the USA.

Pentium CPU is needed for Pentium era games which includes PSX game ports.

Last edited by Hammer on 25-Jan-2024 at 08:00 AM.
Last edited by Hammer on 25-Jan-2024 at 07:57 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

Can you post anything without littering it with guff?

Quote:

John Carmack is aware of the A4000/040 with RTG card capability, but his argument is the installed base with Amiga's 68040+RTG specification.

Doom was created on a 68040. Carmack pointed out at the time that it takes the full processing power of the 040 to run the game even without addressing the planar display issue. The Amiga userbase able to run Doom on a graphics card in 1993 was miniscule and efficient chunky to planar routines were still some years away. And please, don't mention Akiko. The BS around it is bad enough already. It's a PIO register set you have to write to, read back from and then write out again to chip. It was faster at C2P on the CD32 only because the installed CPU / no fast memory was so weak already.

Quote:

68040's FPU is useless for Doom.

I fail to comprehend the reason why you feel the need to point this out. Who ever suggested otherwise?

_________________
Doing stupid things for fun...

Apple dug PPC’s grave when they switched to Intel then their own CPU architecture. PowerPC for Desktops is living on borrowed time. I own a X5000 myself so this is not comforting news, but it is the reality. The X5000 is still a nice machine but it’s higher priced than a Vampire V4.

Last edited by DiscreetFX on 25-Jan-2024 at 06:56 PM.
Last edited by DiscreetFX on 25-Jan-2024 at 05:27 PM.
Last edited by DiscreetFX on 25-Jan-2024 at 05:26 PM.
Last edited by DiscreetFX on 25-Jan-2024 at 05:25 PM.

_________________
Sent from my Quantum Computer.

@DiscreetFX

PowerPC for desktop died when Steve Jobs returned and cancelled the MacOS licensing program, this pretty much ended CHRP.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

People seem to forget when technology was available and how much it cost. High end x86 CPUs could cost more than a low end Amiga computer. Tech changed quickly in the early to mid 90s. Typical hardware being purchased in mass looked something like the following.

1985-1990 68000+OCS vs 286+EIDE
1991 68000+ECS vs 386+EIDE
1992 68EC020+AGA vs 386+custom bus
1993 68EC020+AGA vs 386+custom bus+VRAM
1994 68EC020+AGA vs 486+custom bus+VRAM
1995 68EC020+AGA vs Pentium+PCI+VRAM

x86 hardware had a growing clock speed advantage where the 68k originally had a small clock speed advantage. C= fell behind at delivering value. Economies of scale and cost are important but so is value. Higher end 68k CPUs were available and a 68EC030@28MHz would have been reasonably priced at the time AGA was introduced. Combined with double the chip memory bandwidth from a 28MHz clocked Amiga chipset, this would have kept the Amiga viable through 1994 and would have kept AGA from having the reputation of being "slow". The 68EC030 was produced in adequate scale for economies of scale being used in the embedded market. The 68k was still out selling x86 even in 1997 after the fall of C=, elimination of Atari and architecture switch away from 68k of Apple.

RISC Volume Gains But 68K Still Reigns
https://websrv.cecs.uci.edu/~papers/mpr/MPR/19980126/120102.pdf Quote:

Motorola’s 79.3 million units put it on top, as usual. Its 68K line has been the embedded 32-bit volume leader since it created the category. As the figure shows, sales of 68K chips were about equal to worldwide sales of PCs. Taken together, that’s one new 32-bit microprocessor for every man, woman, and child living in the United States.

In 1997, ARM was only at 10 million units sold and PPC at 3.5 million with more units sold for embedded use than Macs. The x86 architecture was winning the economies of scale war for high end high margin CPUs though. Performance as well as value are important for the desktop market.

kolla Quote:

PowerPC for desktop died when Steve Jobs returned and cancelled the MacOS licensing program, this pretty much ended CHRP.

Apple digs the PPC grave
Steve Jobs returned and stopped Mac clones. Apple practically put the PPC CPU developer Exponential Technology out of business and bought the successor Intrinsity.

IBM put one foot of PPC in the grave
PPC G5 was a huge and expensive disappointment.

Apple kills PPC
Apple switched from PPC to x86

IBM puts the other foot of PPC in the grave
Cell based PPC consoles were a big disappointment.

Apple buries PPC
Apple buys PA Semi, the last PPC CPU developer with potential.

Exponential Technology was early enough they could have made a difference for PPC. The x704 3 way superscalar CPU could clock from 400 to 533MHz which is comparable to Alpha CPUs at the time while the PPC 604e was limited to 233MHz. This was possible with tiny 2kiB I+D L1 caches, an on die 32kiB L2 cache and using dynamic logic with a more expensive BiCMOS process that is able to use static and dynamic logic at the same time. This takes advantage of the faster switching times of NMOS transistors while using CMOS to reduce power. Several early Pentium cores use BiCMOS which was partially responsible for the Pentium out clocking the CMOS only fully static 68060 and PPC designs (BiCMOS extra expense for embedded use was not wanted by Motorola/Freescale). Higher clock speeds are attractive to desktop users but static designs are easier to develop, cheaper to produce and fully static designs can be clocked from maximum all the way to zero. Intel tried to return to a static design with the Pentium Pro but the Pentium 4 and successors used dynamic logic to increase maximum clock speeds.

https://web.archive.org/web/19970110114750id_/http://ada.computer.org:80/pubs/micro/articles/m60032.pdf Quote:

The P55C will mark Intel’s shift away from the BiCMOS process technology of earlier Pentiums. The 0.28-micron (drawn gate size) process enables Intel to reduce the supply voltage from 3.3 to 2.8 V, which significantly reduces power consumption. At this low voltage, however, bipolar transistors offer little benefit, making the extra process steps of BiCMOS unjustified. The supply voltage reduction will make higher clock rates practical for portable systems and will simplify cooling in desktop systems.

Intel returned permanently to static logic with Nehalem (early i7 after Core 2).


https://www.anandtech.com/show/2658/9

https://en.wikipedia.org/wiki/Dynamic_logic_(digital_electronics)
https://en.wikipedia.org/wiki/BiCMOS

PPC CPU cores have often been a disappointment due to cache requirements. The solution has been to add more caches and shrink the die. The disappointing PPC 603 and 604 were saved by doing this and turning them into the 603e and 604e. The PPC G3/G4 gave a 2nd wind to PPC by adding on chip L2 caches. The biggest disappointments were in the low end and high end leaving PPC with mediocrity. The CISC x86 architecture had the high performance market and ARM eventually found some RISC performance by abandoning RISC philosophies and adopting OoO designs while die shrinks allowed to reduce power below PPC cores ending the communication and automotive last embedded niches of PPC.

Last edited by matthey on 27-Jan-2024 at 12:55 AM.

@matthey

You edited your posting, what a disappointment.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@Karlos

Quote:

Doom was created on a 68040. Carmack pointed out at the time that it takes the full processing power of the 040 to run the game even without addressing the planar display issue. The Amiga userbase able to run Doom on a graphics card in 1993 was miniscule and efficient chunky to planar routines were still some years away. And please, don't mention Akiko. The BS around it is bad enough already. It's a PIO register set you have to write to, read back from and then write out again to chip. It was faster at C2P on the CD32 only because the installed CPU / no fast memory was so weak already.

FYI, Viona Development/GVP's EGS was shown in 1992, but Commodore publicly attacked EGS with the promise of a Commodore RTG solution, but Commodore was not able to deliver such a solution before its demise.

Piccolo and EGS 28/24 (Cirrus Logic GD5426) weren't cost-competitive when compared to PC's Cirrus Logic GD5426-based cards let alone the 1989 Tseng Labs ET4000AX.

The Amiga Zorro II market was tiny compared to PC's 16-bit ISA-equipped motherboard mass production rates.

Furthermore, Commodore canceled the A500 which has 3rd third-party Zorro II bus board (via Zorro 1 edge connector) upgrades for the inferior A600.

A500 with 3rd third-party Zorro II bus board and 040-500 - PPS (Progressive Peripherals & Software) 68040 accelerator is not cost-effective due to economics of scale issues and lack of official Commodore support.

What's needed in Xmas 1991 is a single-slot Zorro III Amiga AGA SKU in mass production that would fit between A1200 and A4000. AA3000+ AGA was functional in Q1 1991.
The idea is a single-slot AGP or AMD's A520/A620 single PEG (PCIe 16 lane for graphics) slot low-cost expandable gaming PC.

A500 has a single Zorro I slot and it was replaced by a laptop without a screen A600.

_________________
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)

@matthey

Quote:

People seem to forget when technology was available and how much it cost. High end x86 CPUs could cost more than a low end Amiga computer. Tech changed quickly in the early to mid 90s. Typical hardware being purchased in mass looked something like the following. 1985-1990 68000+OCS vs 286+EIDE 1991 68000+ECS vs 386+EIDE 1992 68EC020+AGA vs 386+custom bus 1993 68EC020+AGA vs 386+custom bus+VRAM 1994 68EC020+AGA vs 486+custom bus+VRAM 1995 68EC020+AGA vs Pentium+PCI+VRAM

1989 A500 was price competitive (about $1000 AUD, about 1/3 weaker than USD) while 1990 released A3000 ECS wasn't price competitive.

For the European market, the 1992-era stock A1200 delivers about 16-bit SNES gaming experience. SNES has the edge with arcade game ports. PC's Mortal Kombat 1 port quality is competitive against the SNES version.

The Amiga has an uncompetitive price problem above the stock A500 and A1200.

Amiga wasn't price-competitive by 1992.

https://archive.org/details/bub_gb_hqQJaNzN9IcC/page/n603/mode/2up
PC Mag 1992-08, page 604 of 664,
Diamond Speedstar 24 (ET4000AX ISA) has $169 USD.

Existing PC clone 386 with ISA slots can be upgraded with ET4000AX which improves VGA frame buffer performance.

386DX 25 Mhz is needed for a smooth frame rate Wing Commander (1990). Intel 386DX-33 and 486DX 20 to 25Mhz SKUs were released in 1989.

486SX 16 Mhz, 20 Mhz, and 25 Mhz SKUs were released in 1991. FPU wasn't a major factor for Wing Commander and Doom.

High-end PC CPU from 1989 to 1993 is 486 based on SKU. You don't need a 486 for Doom (full-screen, low details). Pentium was released in 1993.

386DX-33+ET4000AX can play Mortal Kombat 2 better than SNES.


Quote:

x86 hardware had a growing clock speed advantage where the 68k originally had a small clock speed advantage. C= fell behind at delivering value. Economies of scale and cost are important but so is value. Higher end 68k CPUs were available and a 68EC030@28MHz would have been reasonably priced at the time AGA was introduced. Combined with double the chip memory bandwidth from a 28MHz clocked Amiga chipset, this would have kept the Amiga viable through 1994 and would have kept AGA from having the reputation of being "slow". The 68EC030 was produced in adequate scale for economies of scale being used in the embedded market. The 68k was still out selling x86 even in 1997 after the fall of C=, elimination of Atari and architecture switch away from 68k of Apple.

For A1200, 7.1 MB/s Chip RAM is about 3.5 Mhz 32-bit or 7.1 Mhz 16-bit. Alice's memory controller has 1985 Agnus-like behavior.

Commodore management didn't allow any new Amiga graphics IP to be designed on a proven 25 Mhz 32-bit Ramsey memory controller. Recall, "Read my lips, no new chips" during A3000's development.

Alice's memory controller wasn't designed with a 020/030 bus performance.

28 Mhz 32-bit memory controller would be delivering about 112 MB/s.

1st gen PCI chipset has about 25 MB/s bandwidth, but Intel has improved its PCI chipsets until the need for AGP.

There's very little real-world IPC difference between 68020 and 68030, hence 28 Mhz 68020 would do the job. The fastest 68020 is a 33 Mhz variant.

Last edited by Hammer on 27-Jan-2024 at 12:35 PM.
Last edited by Hammer on 27-Jan-2024 at 12:17 PM.
Last edited by Hammer on 27-Jan-2024 at 12:02 PM.

_________________
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)

@matthey

Quote:

RISC Volume Gains But 68K Still Reigns https://websrv.cecs.uci.edu/~papers/mpr/MPR/19980126/120102.pdf Quote: Motorola’s 79.3 million units put it on top, as usual. Its 68K line has been the embedded 32-bit volume leader since it created the category. As the figure shows, sales of 68K chips were about equal to worldwide sales of PCs. Taken together, that’s one new 32-bit microprocessor for every man, woman, and child living in the United States.

The embedded argument is useless for the desktop computer market.

https://arstechnica.com/features/2005/12/total-share/8/
_ By 1998, PCs were closing in on sales rates of 100 million units per year,_

X86 has a unified ecosystem with common motherboard form factors.

A Motorola 68030 CPU is useless without an adaptor board for A1200 or A500.

A Motorola 68040 CPU is useless without an adaptor board for A1200 or A500.

A Motorola 68060 CPU is useless without an adaptor board for A1200.

Shall I post the Gateway 2000 price list against Commodore again?

I can purchase a low-cost Power8 CPU, but it's useless without a cost-competitive motherboard!

https://techmonitor.ai/technology/motorola_plans_to_sample_the_68060_next_quarter

Motorola Inc yesterday finally launched the long-promised 68060 follow-on to the 68040, claiming that it matches the performance of the Intel Corp Pentium at less than half the price – it costs $263 at 50MHz when you order 10,000 or more and will sample next month

There are also cheaper 68LC060 and 68EC060 variants of the new part, which omit the memory manager, and both the memory manager and the floating point unit; they cost $169 and $150 respectively for 10,000-up.


From 1994 to 1996, Amiga-related 68060-based accelerator cards weren't below $500 USD.

In 1996, I purchased my Pentium 150-based PC clone for about $1500 AUD.

Prove to me that a 68060-based Amiga bundle was offered at AUD 1500 in 1996.

There's a very large gap between Motorola's 68060 wholesale price and Phase 5's asking price.

https://amiga.resource.cx/adcoll/preview/Phase5_1995-08.jpg
Blizzard 1260 Turbo card has DM 1249 which is about $832 USD or $1040 AUD.

https://www.poundsterlinglive.com/bank-of-england-spot/historical-spot-exchange-rates/usd/USD-to-DEM-1996

https://www.poundsterlinglive.com/bank-of-england-spot/historical-spot-exchange-rates/usd/USD-to-AUD-1996

1 USD to AUD is about 1.25.

1 USD to DM is about 1.5.

In 1995, Pentium was at 133 Mhz (June 1, 1995) while Pentium Pro was at 200 Mhz (November 1, 1995).

I purchased my Pentium 150-based PC clone in 1996. Pentium class games such as Quake and Duke-Nukem 3D were the triggers. Sony PSX ports require Pentium class PC.

In 1996, Pentium was at 200 Mhz (June 10, 1996). Pentium Pro and Pentium 200Mhz cause price pressures for older Pentium SKUs.

Last edited by Hammer on 27-Jan-2024 at 11:54 AM.
Last edited by Hammer on 27-Jan-2024 at 11:26 AM.
Last edited by Hammer on 27-Jan-2024 at 03:16 AM.
Last edited by Hammer on 27-Jan-2024 at 02:28 AM.
Last edited by Hammer on 27-Jan-2024 at 02:27 AM.
Last edited by Hammer on 27-Jan-2024 at 02:26 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: @agami https://www.statista.com/statistics/1101900/unit-sales-snes-region/ Super Nintendo Entertainment System (SNES) lifetime unit sales worldwide as of September 2023 North America is 22.88 million. Japan is 17.17 million Europe is 8.15 million The rest of the World is 0.9 million https://vgsales.fandom.com/wiki/Fourth_generation_of_video_games SNES install base 1990 1991, 2 million 1992, 6.9 million (+4.9 million) 1993, 11.3 million (+4.4 million) For 1992, Doom-capable "gaming PC" had extra time to build its install base 1. Intel's 32-bit 486 was released on April 10, 1989. Intel 486SX-25 and 486SX-33 were released in 1991. 2. Intel full 32-bit 386 was released in 1985 and mass production in 1986. Intel's 386 CPUs had topped out at 33 MHz in 1989. AMD introduced 386DX-40 MHz versions in March 1991. https://arstechnica.com/features/2005/12/total-share/7/ The golden age of PC gaming had arrived, with classics like Wing Commander, DOOM, Ultima 7 and System Shock pushing the boundaries of computer entertainment. At the same time, the PC became easier to use with the release of Windows 3 in 1990 and 3.1 in 1992. The latter operating system proved incredibly popular, pushing PC sales back from their minislump in 1991. By 1994 PCs and clones were selling at the incredible rate of 37 million units per year.

Thanks for providing the data to support my point.

Looking back, late '92/early '93 and 1994 may appear very close together, likening them to how two consecutive years for today's mature computing ecosystem are almost the same.
But looking forward from 1992 (the year the A1200 went into production), a time when there was still serious fighting over hardware and software standards, 1994 was just as unknown as 2000.

For Commodore, a company that was in debt and struggling to find meaningful profitability, the A1200 was about as logical/practical a move as they come. Alas it and subsequent CD32 gambit did not pay off. And I'd go as far to say the world was poorer for it.

_________________
All the way, with 68k

@agami

Quote:

agami wrote: Thanks for providing the data to support my point.

Your "game console" argument is a red herring since they (SNES, Genesis) are not the Amiga platform.

"Game consoles" are not a single monolithic platform.

For Amiga's revenue core European market, both SNES and AGA platforms arrived in Q4 1992.

Commodore wasted time on a 32-bit 020-based CPU with ECS "A1000 Jr" when AA3000+'s AGA chipset was completed in Q1 1991. "More than six months" was wasted.

AA3000+'s R&D was on top of A3000's "no new chips" time-wasting.

Extra engineering effort was needed to cut down AA3000+ into A1200 and A4000 (inserted PIO IDE controller and removed SCSI controller, Amber flicker fixer, and AT&T 3210 DSP). The A4000 motherboard form factor changed and it's not compatible with existing A3000 and A2000 cases.

Meanwhile, C65's 256-color (8-bit planes) display chipset was completed in Q4 1990. This design was cloned as MEGA65 by Commodore fans.

On paper spec, AT&T 3210 DSP inclusion enables AA3000+ to rival 1993 Pentium 60's FP32 MFLOPS.

During 1993, the A4000/040 wasn't competitive against a similarly priced Pentium 60 PC clone.

AA3000+ PCB design was recently cloned by Amiga fans as a drop-in A3000 motherboard replacement. DaveH wanted AA3000+ as a drop-in motherboard upgrade path for existing A3000 owners.

Quote:

Looking back, late '92/early '93 and 1994 may appear very close together, likening them to how two consecutive years for today's mature computing ecosystem are almost the same.

With PCs, buying a faster VGA card benefits gaming PCs. EGS RTG is a deadweight with Amiga gaming.

Only Commodore makes it possible for the full "32-bit CPU" platform to be dead-ended when AGA Amigas was released i.e. buying 030 accelerated OCS/ECS Amiga and then buying another 030 accelerated Amiga. Meanwhile, 386DX PC can upgrade with ET4000AX.

Amiga's big box game console topology doesn't work on expandable desktop computers.

Quote:

But looking forward from 1992 (the year the A1200 went into production), a time when there was still serious fighting over hardware and software standards, 1994 was just as unknown as 2000. For Commodore, a company that was in debt and struggling to find meaningful profitability, the A1200 was about as logical/practical a move as they come. Alas it and subsequent CD32 gambit did not pay off. And I'd go as far to say the world was poorer for it.

A500's cancellation and the move from Hong Kong to the Philippines manufacturing plant were Commodore's self-inflicted problems and damaged A500's 3rd party add-on market.

_________________
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)

@agami

Quote:

agami wrote: True, but I argue that the situation you've described above was not reality in most markets until 1994. In 1992 and for most of 1993 1. In the largest PC market, USA, the gaming market was dominated by consoles.

That's a red herring.

Quote:

2. By most indications, UK was very similar to AU, or vice versa, and there too gaming was dominated by game consoles (to a lesser degree than USA) and "home" computers from Commodore and Atari.

That's a red herring with your game console argument.


Quote:

3. Gaming in the rest of west and east Europe was also much more skewed toward "home" computers from the likes of Commodore and Atari, which were more compatible with the standards of living in those countries. 4. A PC sold which is more capable of running games than an A1200 is not automatically a "gaming PC". So even if more 486 PCs were sold to deep pocketed early adopters in 1992 and 1993 than total A1200 sales, fewer of them were intended for gaming than the A1200. I wouldn't be surprised if Apple sold more mac Performa 450s (LC III) units than A1200s in 1993, but that still didn't make it a target for FPS game developers.

Do you claim there were only 120,000 486 gaming PCs in 1992?

A1200 shipments in 1992 are in the 100,000s. Your argument is absurd.

Quote:

id Software didn't sell 4M Doom units in December 1993.

That's a red herring. I didn't claim there were "4M Doom units in December 1993".

I played the shareware version before I obtained the full Doom.

I did state PC's Doom (released in 1993) and Doom 2 (released in 1994) unit sales were 4 million.

Doom-capable gaming PCs are more than A1200/CD32 install base.

IDsoftware estimated another 15 million floating copies i.e. pirate copies.

Doom was competing against other games such as Star Wars X-Wing (1993) and IndyCar Racing (1993).

https://youtu.be/1B1jKjrRUmk?t=7
68030 50 Mhz (TF1230, C= AGA) VS 386DX-40 Mhz (ET4000AX) playing Doom.

There's a reason why Commodore UK pushed for CPU-accelerated A1200 game bundles.

https://youtu.be/KQDEKoRcXZc?t=334
i386DX-33 ET4000AX playing Doom.

CPU-accelerated A1200 game bundles would deliver a "32-bit" gaming experience that can't be done on SuperFX2-enhanced SNES.

Quote:

And not even when Doom II came out in May 1994. It took some time for that number to be reached. Thanks to a confluence of contributing factors all culminating in 1994.

A game has a self-life.

Last edited by Hammer on 27-Jan-2024 at 01:34 PM.
Last edited by Hammer on 27-Jan-2024 at 01:25 PM.

_________________
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: 68040's FPU is useless for Doom.

Because Doom does not use the FPU. It is based on an integer Raycasting Alogrithm.

Hammer Quote:

The Amiga has an uncompetitive price problem above the stock A500 and A1200. Amiga wasn't price-competitive by 1992.

The mass produced Amiga 1200 was at the low end of mid-performance in 1992, low end in 1993 and practically obsolete in 1994. Increasing the CPU performance was relatively easy while increasing the chipset performance was difficult, especially with the lack of vision and planning by upper management.

Hammer Quote:

There's very little real-world IPC difference between 68020 and 68030, hence 28 Mhz 68020 would do the job. The fastest 68020 is a 33 Mhz variant.

The cost difference of a 68EC020 and 68EC030 was quickly closing and the cost of a 68EC030 became low enough to be compelling. There were likely to be better quantity discounts on the 68EC030 that was not at the high end of clock ratings. The 68030 is ~25% lower power than the 68020 offsetting much of the clock increase to power and heat, the data cache further reduces memory traffic and the 68EC030 has a full 32 bit address bus instead of the 24 bit address bus of the 68EC020 (larger memory expansions could be supported while PCMCIA slot and memory expansion conflicts could be avoided).

Hammer Quote:

The embedded argument is useless for the desktop computer market.

It's a good thing the RPi Foundation doesn't market or sell "useless" embedded CPU/SoC chips for the desktop market. A-Eon sells the AmigaNOne with embedded CPU/SoC chips for the desktop market but they are obsolete even for the embedded market.

Back in the Amiga days, C= bought mostly embedded CPUs. This allowed them to leverage economies of scale from the embedded market. It worked well enough until they needed higher performance CPUs and there wasn't adequate economies of scale to reduce CPU chip costs with Apple leaving the 68k market. Well, C= upper management was so oblivious that they didn't get that they needed to increase CPU performance to remain competitive or they would have at least upped the clock speed of the CPU with the very popular and high production embedded 68EC030. The 68060 would have been super for desktop, laptop and console use but the high end embedded market wasn't large enough at that time to drop the price, especially without Apple.

Hammer Quote:

https://techmonitor.ai/technology/motorola_plans_to_sample_the_68060_next_quarter Motorola Inc yesterday finally launched the long-promised 68060 follow-on to the 68040, claiming that it matches the performance of the Intel Corp Pentium at less than half the price � it costs $263 at 50MHz when you order 10,000 or more and will sample next month

It would have been more accurate to say that the 68060 offered the per clock performance of the Pentium and costs less than half the price. A 100MHz Pentium P54C was planned by the time the 68060 launched in 1994.

68060@50MHz
90 DMIPS (1.8 DMIPS/MHz)
2.5 million transistors, CMOS
$308 (1000s quantity)

Pentium P54C@100MHz
138 DMIPS (1.38 DMIPS/MHz)
3.3 million transistors, BiCMOS (large die and BiCMOS raises production cost)
$995 (1000s quantity)

PPC601@66MHz
93 DMIPS (1.41 DMIPS/MHz)
2.8 million transistors, CMOS
$370 (1000s quantity)

https://websrv.cecs.uci.edu/~papers/mpr/MPR/ARTICLES/080502.pdf

Hammer Quote:

There are also cheaper 68LC060 and 68EC060 variants of the new part, which omit the memory manager, and both the memory manager and the floating point unit; they cost $169 and $150 respectively for 10,000-up.

The article above talks about actual 68060@66MHz silicon being tested in 1994 yet there was never a 68060 released with a rating above 60MHz (rare, not advertised & perhaps for a single customer) that I'm aware of despite 2 die shrinks and revision 6 CPUs often overclockable to 100MHz.

https://websrv.cecs.uci.edu/~papers/mpr/MPR/ARTICLES/080502.pdf Quote:

Initial devices will run at 50 MHz, with production quantities of 66-MHz units promised for late in 4Q94. Along with the full-scale processor, the company introduced two lower-cost variants: the 68LC060 (which lacks an FPU) and the 68EC060 (which has neither an FPU nor an MMU). ... Tests also showed that a 50-MHz ’060 had 2.1 times the performance of a 40-MHz ’040 running the Macintosh Speedometer benchmark; the 66-MHz version achieved a rating 2.8 times the ’040. These figures are from alpha-test sites using actual ’060 silicon installed in converted Macintosh machines and in Motorola’s IDP (Integrated Development Platform) evaluation boards.

The embedded LC and EC variants of the 68060 seemed to be the only priority after the full 68060 release in 1994 despite the paper talking about other enhancements. A 68060@66MHz would have directly competed with the PPC601@66MHz with a competitive advantage in performance and cost (due to smaller chip die size). The 8 stage 68060 pipeline should have provided a significant advantage in clock speed over the PPC601 with 4 stage pipeline as well (the 8 stage 68060@50MHz pipeline was deeper than the contemporary Alpha21064@300MHz with 7 stage pipeline). The fateful political decision to switch to PPC for the desktop had already been made though. The Pentium killer 68060 was locked in the Motorola basement and designated to low clock speed embedded duty, with one of the deepest pipelines and best DMIPS/MHz performance of the time.

CPU max clock rating @ ~0.5um chip process with pipeline stages
ARM710@40MHz 3-stage
PPC601+@120MHz 4-stage
PPC603@160MHz 4-stage
Pentium P54C@120MHz 5-stage
PPC604@180MHz 6-stage
HP PA-8000@180MHz 7-stage
Alpha 21064@300MHz 7-stage
MIPS R4400@200MHz 8-stage
68060@50MHz 8-stage

Last edited by matthey on 28-Jan-2024 at 04:14 AM.
Last edited by matthey on 28-Jan-2024 at 04:03 AM.
Last edited by matthey on 28-Jan-2024 at 03:53 AM.
Last edited by matthey on 28-Jan-2024 at 02:06 AM.
Last edited by matthey on 28-Jan-2024 at 01:31 AM.
Last edited by matthey on 28-Jan-2024 at 12:09 AM.

@matthey

Quote:

matthey wrote: and the 68EC030 has a full 32 bit address bus instead of the 24 bit address bus of the 68EC020 (larger memory expansions could be supported while PCMCIA slot and memory expansion conflicts could be avoided).

While the cost of the chips may have been near identical, the cost of running/handling those extra 8 address lines would have been real.

On a product that had to be penny pinched to make sense on the bottom of the market even at launch.

For the price of an A1200HD you could get a decent 286 or even 386, which yeah sure wasn't as sexy, required a proper monitor yaddayaddadada..... but was also seen as a "real computer" by parents who most of the times footed the bill.

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

@Kronos

Also lets not forget that C= was planning for midrange AGA-Amiga that would have started with a full 68020 (as evidenced by the solder pads on the A3630) which might have stopped them from making the A1200 "too good" to avoid cannibalization.

That was canceled at one point and kinda replaced with the A4000/030.

Last edited by Kronos on 28-Jan-2024 at 12:34 PM.

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

@Kronos

The PA6T was out in 2007 and out again the year after when it was discontinued. And by then 64 bit was increasingly becoming PC standard. In any case the most important feature it had is AltiVec which is said to exist from 1999 in the G4. Embedded PPC tends to lack AltiVec as well as using incompatible supervisor instructions. Including other differences from a full BookS core. So, by PPC standards, any PPC without AltiVec and a full BookS core is inferior. There's a good reason an e6500 core with AltiVec was chosen for the PPC notebook even though it's an embedded market CPU.

The X1000 was rather crippled in some respects. The PA6T was designed for DDR2 when DDR3 was around the corner. The X1000 was a better design than the XE but, being on par with a 10 year older Apple doesn't look good. The PA6T was likely the best chip on the whole board.

Intel were certainly more popular as a brand. PowerPC would have been more confusing than anything when they slapped it on the cases. A Mac that is a powerful PC? It can't even run PC applications. Nice try Apple!

But, many people blamed PowerPC for the high cost of Macs. Including Amiga people that blamed it for the high cost of the XE boards. Though, the Pegasos G4 was almost a 3rd of what an XE G4 cost, with a superior design. So, Apple go Intel, and their machines are still expensive!

I can imagine if an AmigaOne x86 was produced. The same people would complain it was too expensive and they could buy a PC cheaper. Of course, it would be more expensive just like Mac x86 was, but they all argued that it was PPC which made it expensive.