@matthey
"AI is over rated and I believe there is an AI bubble reminiscent of the 2000 dot-com bubble for several reasons.
1. energy hungry
2. fuzzy logic
3. no morality
4. insane valuations of AI related stocks like Nvidia"


The most insane thing I've sporadically read about it that seems to be true is that huge parts of it in Silicon Valley are going to be feed huge amounts of electricity produced from ..... drum roll COAL!
While the west fines anyone for their carbon emissions and it's industry is moving to China and India where coal plants are being built faster than the west is dismantling them, all in the name of reducing planetary emissions, it doesn't have any problems feeding AI coal produced electricity... Nevermind the Orwellian surveillance state that will come about with AI...

_________________

José

@Jose

Hardware will probably arise that will make it less of a burden on carbon emissions. We already have had massive developments in protein folding, who knows if it will be able to control fusion

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

Jose Quote:

The most insane thing I've sporadically read about it that seems to be true is that huge parts of it in Silicon Valley are going to be feed huge amounts of electricity produced from ..... drum roll COAL! While the west fines anyone for their carbon emissions and it's industry is moving to China and India where coal plants are being built faster than the west is dismantling them, all in the name of reducing planetary emissions, it doesn't have any problems feeding AI coal produced electricity... Nevermind the Orwellian surveillance state that will come about with AI...

Yes, most of the "democratic" West is turning into dystopian nations committing economic suicide. Stopping coal use, the dirtiest energy for both CO2 and other pollutants, made sense. The U.S. converted from mostly coal to mostly natural gas but instead of energy businesses being heroes, they were vilified as natural gas, oil and even nuclear were targeted as bad by the same radical leftist groups, who are also anti-business. Rising energy prices reduce industrial competitiveness as seen by European nations, especially industrialized Germany, since the start of the Ukraine war. The U.S. came to the rescue of Europe with LNG shipments which Biden fast tracked but he stopped new drilling permits, increased regulations beyond reason and more recently has tried to stop new LNG exports. We have such a glut of natural gas that energy businesses in Texas have asked to flare the natural gas (burn it off with no energy production). In the mean time, China has been importing Russian oil at a reduced price funding the Ukraine war. China doesn't just use coal. They use diverse energy sources which is a better strategy for energy security while the West green strategy shuts down non-green energy sources before replacing them. I expect U.S. and European combat readiness is practically zero without oil too.

Data centers already move to locations that have an energy cost advantage. AI high energy use is likely to encourage this more. China is likely to be a destination because of it. China is already ahead of other nations in the use of AI which it is using to spy on and control its own citizens as big brother. This is not limited to China itself as it has policing stations in Western nations. Electronic equipment has back doors, more than a few Chinese people have been caught stealing data, Chinese hackers have been very active probing Western security, China has been involved in organized crime including the drug trade and most Chinese businesses are majority owned by the state making dealing with any Chinese businesses risky yet they are a huge supplier of goods that is likely to grow as Western energy prices rise. We have our own dystopian problems too as the U.S. at least has become much more corrupt. The Ukrainian war started when the Biden family moved in on the turf of the crime lord Putin and his energy dominance of Eastern Europe. The Twitter files show state funded propaganda and election interference. There is political targeting with the judicial system, intelligence agencies, IRS, state controlled media, etc. Even our voting system is in question as tax payer dollars have payed for propaganda and censorship that violates free speech granted in the Bill of Rights. I don't think nuclear systems will be put in charge of AI like Skynet anytime soon but I'm not sure that will save us as not just AI lacks morality but our leaders too. I am much more concerned with humans destroying the world in WWIII than global warming destroying it.

pixie Quote:

Hardware will probably arise that will make it less of a burden on carbon emissions. We already have had massive developments in protein folding, who knows if it will be able to control fusion

Sure, AI energy use will be reduced and performance improved but it is starting at a much higher energy use than procedural code. OOP code used to have a big performance loss and still has a significant performance loss on low end hardware but indirect branch prediction and larger caches have helped reduced the performance loss at a cost. AI appears to have a much higher energy cost than OOP.

https://marz.kau.edu.sa/Files/320/Researches/73552_46720.pdf Quote:

The OOP approach can demean performance and raise the power consumption of software (as compared to the classical procedure programming) because of its additional abstraction and encapsulation layers and mechanisms.

I expect nuclear fusion is many years and likely decades away. Nuclear Fission may even be the answer but likely not with Uranium fuel which is expensive and rare causing the cost of nuclear power to go up as other energy sources become cheaper.

Last edited by matthey on 16-May-2024 at 06:28 PM.

Getting back to the topic of lack of competitive Amiga hardware, I believe the following article pretty well sums up just how little value modern AmigaNOne hardware has.

https://www.talospace.com/2020/01/another-amiga-you-dont-want.html Quote:

Another Amiga you don't want The Amigaphile community is possibly more rabid about its ecosystem than even us OpenPOWER dweebs, so right off I present my Amiga bona fides before getting stuck into it: in this house is a Amiga Technologies '060 A4000T running AmigaOS 3.9, an Amiga 3000 (with the tape drive, so I might even try Amix on it someday), and several A500s. My disenchantment with the current crop of PowerPC-based Amigas, however, is well-known. At their various asking prices you got embedded-class CPUs at mid-high desktop level prices with performance less than a generation-old Power Mac. (Many used and easily available models of which, I might add, can run MorphOS for considerably less cash out of pocket.) A-EON's AmigaOne X5000 in particular attracts my ire for actually running a CPU that doesn't have AltiVec to replace the X1000, which did. Part of this problem was the loss of availability of the PA6T-1682M CPU, the one and only CPU P. A. Semi ever officially produced under that corporate name before Apple bought them, but there were plenty better choices than the NXP P5020 which wasn't ever really intended for this application. Since 2014 or thereabouts a "small" Amiga (presumably along the lines of the earlier Sam440ep and Sam460ex family, which used PowerPC 400-series CPUs) was allegedly in development by A-EON and unveiled in 2015 as the original A1222 "Tabor" board. At that time it was specced as a mini-ITX board with 8GB of RAM and most controversially another grossly underpowered embedded core, a QorIQ P1022 at 1.2GHz. Based on the 32-bit e500v2 microarchitecture, it not only also lacks AltiVec but even a conventional FPU, something that for many years was not only standard on PowerPC CPUs but considered one of its particular strengths. As an embedded CPU, this was forgivable. As an embedded CPU in a modern general-purpose computing environment, however, this was unconscionable. Even by Amiga standards this was an odd choice and one with potential impacts for compatibility, and one a G5 or high-end Power Mac G4 would mop the floor with. The original Tabor was apparently manufactured in a small batch of 50 for testers. Between then and 2018 it's not clear what happened on A-EON's end, but their hardware partner had a change of management and various other components needed updating. Strangely, the CPU was not among them. Today, six years later, people may now publicly pre-order the "A1222 Plus," with no change in specs other than changing some board components, and expected to ship Q2 2020. If you're lucky and they're not sold out, your $128 will get you a AAA Bundle Package with some software and a certificate in a pretty box as a 20 percent down payment on the "special low" $450 purchase price. If you get your $128 in after they're gone, then it's just a deposit. The AAA Bundle Package was supposed to be available December 24 but the website wasn't even up until January 11. I'm willing to make some allowances for a cheap(er) modern Amiga because like our OpenPOWER systems, getting the price down some expands the market (ergo, Blackbird). However, that $450 price is almost certainly not the intended MSRP (nor has it been reported what that eventual MSRP is), and for low production volumes the CPU is not the major cost of a system. I don't know what actual considerations went into its design, but if A-EON chose this CPU deliberately as an attempt to make the price of the system lower, they chose wrong. I certainly don't want to pick on other boutique systems unnecessarily. After all, in many people's minds the Talos II I'm typing on is itself a pretty darn boutique system, and one where the out-the-door price can easily eclipse even the priciest AmigaOne configuration. If people want to have their fun and pay a crapload of money for it even if other people think it's junk, then as long as it's not hurting anyone else praise the Lord and pass the credit card. However, this is not that situation. Amiga has a long history in computing consciousness and most computer nerds would at least recognize the brand. Similarly, people still remember Power Macs, and while recollections vary in accuracy and fondness, there's a rather pernicious and commonly-held belief that Apple's migration to Intel somehow "proved" the inferiority of Power ISA. Today, along comes yet another underwhelming PowerPC-based Amiga to confirm their preconceived notions: because it's Amiga, it sticks in people's minds, and because it confirms their own beliefs about PowerPC, it reinforces their unjustified biases. Regardless of the fact that POWER9 systems like this one outpace ARM and RISC-V and compete with even high-end x86_64 offerings, the presence of Tabor and X5000 et al simply gives the resolutely underinformed yet another stick to beat Power desktops with. These Amigas suck compared to PCs, they reason, so OpenPOWER must suck too. At the very least modern Amiga systems need to beat these decade-plus-old Power Macs to be even vaguely taken seriously. If they must go embedded then at least an e6500-series processor would do better than anything that they're running, and while I certainly do appreciate the considerable AmigaOS porting work that would be necessary, going with a minimally-redesigned reference design board for a POWER8 or POWER9 in big-endian mode could still free up development resources for such a task that would otherwise be sunk into yet another quixotic product. As it is, these Amiga systems don't do the Amiga community any favours, and from this outsider's view their performance even seems to be going backwards. If cost were the main consideration, there are other ways of dealing with it, and it's a certainty that Tabor's eventual price will be much greater than $450. Given all that, plus the protracted gestation time for "Tabor Plus," don't the Amiga owners who would be willing to buy such a machine deserve something with a little more zip? The Amiga community needs new hardware, to be sure, and A-EON has at least filled the business opportunity however slow and abortive its progress in doing so; whatever small number they end up producing this time I imagine they'll eventually sell. That doesn't change the fact that weaksauce systems like this being sold as desktop machines continues to tar the architecture as underpowered and does nothing to expand the market beyond the shrinking circle of faithfuls. When there exist today at this very moment Power ISA workstations that can be every bit as utilitarian and functional as Intel workstations, the last thing any of our two communities need is yet another Amiga that people don't want.

How sad it is that the value has declined even more with current A1222 pricing, as predicted above. Every Amiga user at once, place bags over head. Irving Gould only tried to compete with a 68EC020@14MHz against 80386s and 80486s at more than double the clock speed. Trevor's hardware competitiveness and value is much worse than anything CBM ever produced and the guy won't give up no matter how embarrassing his AmigaNOne is and after more than 2 decades of failure with a broken business model.

@matthey

Thanks for sharing the link. Good read, even if it is from 4 years ago.
Or should I say: and especially since it's from 4 years ago.

_________________
All the way, with 68k

@matthey

Quote:

How sad it is that the value has declined even more with current A1222 pricing, as predicted above. Every Amiga user at once, place bags over head. Irving Gould only tried to compete with a 68EC020@14MHz against 80386s and 80486s at more than double the clock speed. Trevor's hardware competitiveness and value is much worse than anything CBM ever produced and the guy won't give up no matter how embarrassing his AmigaNOne is and after more than 2 decades of failure with a broken business model.

Due to a shared memory bus issue, A1200's 14 Mhz 32-bit 68EC020 operates like 7 Mhz 32-bit. Adding Fast RAM doubles A1200's 68EC020's performance.

PC 386DX-33/386DX-40/486SX/486DX PCs with VGA-type cards have discrete video memory.

Irving Gould only tried to compete with a 68EC020 @ 7 MHz effective against 80386s and 80486s with discrete video memory.

Other game consoles like Mega Drive, SNES, 3DO, PS1, and Saturn have discrete video memory.

Mega Drive's 68000 CPU has a 16-bit bus and custom chips have discrete fast 8-bit VRAM (the RAM type rejected by Commodore's soft drink CEO).

According to ex-Commodore UK's MD, A1200 has a healthy profit margin.

A similar spec CD32 with 2X speed CD-ROM drive has 299 UKP and A1200's 399 UKP.

A1200 should have 2MB Chip RAM and 1 Mb Fast RAM (and $20 DSP3210 @ 42 Mhz or 56 Mhz). It's about $4 difference between 68EC020-16 and 68EC020-25.

A1200's craftmanship wasn't good due to the maximum "PCJr" mindset from Bill "PCJr" Sydnes.

Ali fired Bill Sydnes after the A300(A600) debacle


https://www.youtube.com/watch?v=P4VBqTViEx4
Steve Jobs on the role of product and marketing people.


https://www.youtube.com/watch?v=eywi0h_Y5_U
Sales Steve Ballmer laughs at iPhone. Steve Ballmer's spills Windows Mobile's features but doesn't realize the craftsmanship of Windows Mobile is inferior to Apple's iOS.

Last edited by Hammer on 17-May-2024 at 10:55 AM.
Last edited by Hammer on 17-May-2024 at 10:49 AM.
Last edited by Hammer on 17-May-2024 at 10:43 AM.
Last edited by Hammer on 17-May-2024 at 10:40 AM.
Last edited by Hammer on 17-May-2024 at 10:39 AM.
Last edited by Hammer on 17-May-2024 at 10:35 AM.
Last edited by Hammer on 17-May-2024 at 10:32 AM.
Last edited by Hammer on 17-May-2024 at 10:30 AM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@pixie

IQuote:

pixie wrote: @Hammer If only they brought back crts again....

I still have my 1989-era C= 1084S monitor.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@ppcamiga1

Quote:

ppcamiga1 wrote: as long as regular tv display video from chipset chipset has some sense after swith to digital tv chipset is worth nothing no reason to use it on pc aros is still not good as windows so no reason to switch to pc only valid option is rtg/ng agami stop trolling start working on mui

Note that Amiga's video port outputs both digital and analog signals.

https://www.youtube.com/watch?v=F6RvQAVEU84
Amiga's digital video is similar to enhanced CGA e.g. Plantronics ColorPlus or Tandy 1000. Amiga's digital video can display all CGA 16-color space at the same time.

Amiga's digital video yields clear pixel graphics, useful for word processing.

https://www.youtube.com/watch?v=xNsK_F4JlG4
Meet the "Super CGA" Cards i.e. Plantronics ColorPlus.

Last edited by Hammer on 17-May-2024 at 11:14 AM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@Hammer

Quote:

Amiga's digital video yields clear pixel graphics, useful for word processing.

Did it give you extra modes on amiga though? I Think it would give you pretty much the same, but you now would be able to use a cga monitor. For example I had 1084s (the same being presented here) it wouldn't give me any advantage whatsoever, I would have the same crisp image but with less colors.

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

Hammer Quote:

Due to a shared memory bus issue, A1200's 14 Mhz 32-bit 68EC020 operates like 7 Mhz 32-bit. Adding Fast RAM doubles A1200's 68EC020's performance.

Good Point. The Amiga CPU was at a major disadvantage without fast mem or VRAM because of chip mem bus contention between the CPU and chipset. It's not quite as simple as saying it operated like it was half speed. There is no contention for ROM accesses and the 68EC020 instruction cache reduces memory traffic. The performance difference depends on memory performance as well. CPU performance could more than double with fast memory added so the Amiga 1200 CPU performance may have been roughly 1/4 of the low end but high clocked 80386 competition in the early 1990s. Some customers understood that at least a fast mem upgrade was needed but the lack of competitiveness gave a reputation of poor performance to the Amiga 1200 and CD32. Even loyal Amiga customers began to think the old Amiga chipset had become a bottleneck and looked to graphics cards often with VRAM that CBM had denied for Ranger and which would have completely solved the memory bus contention. Another issue was the switch from 2D games where the Amiga had dedicated hardware giving a major boost in performance to 3D games where the 2D hardware provided only a minor benefit and CBM was slow to add chunky modes which make 3D easier and reduces memory traffic. Amigas generally lacked standard high density floppies, hard drives and CD-ROM drives to deliver larger games. The Amiga became retro in the 1990s before it was popular to be retro.

Hammer Quote:

PC 386DX-33/386DX-40/486SX/486DX PCs with VGA-type cards have discrete video memory. Irving Gould only tried to compete with a 68EC020 @ 7 MHz effective against 80386s and 80486s with discrete video memory. Other game consoles like Mega Drive, SNES, 3DO, PS1, and Saturn have discrete video memory. Mega Drive's 68000 CPU has a 16-bit bus and custom chips have discrete fast 8-bit VRAM (the RAM type rejected by Commodore's soft drink CEO).

Most early consoles had separate memories for different purposes. The advantage was that some memory could be faster without requiring as much of it and most 8 and 16 bit CPUs could only address a 64kiB bank of memory anyway. The disadvantages are losing some memory in the partitioning and increased complexity in programming. The Amiga chipset memory could have been smaller and used VRAM if it had only been for a console but higher resolutions were used on the desktop. Jay Miner's Ranger design was an attempt to unify the Amiga memory but would have required more VRAM than most hardware. This would have been more expensive but offered better performance and was ahead of its time. The 1996 N64 has a large 4MiB unified memory subsystem and more powerful CPU than previous consoles which is what the Amiga needed to be competitive with less chipset hardware. CBM didn't upgrade anything though and let the Amiga slip from low end to obsolete. Modern consoles use unified memory and even a HSA to keep caches coherent and unified system wide.

Hammer Quote:

According to ex-Commodore UK's MD, A1200 has a healthy profit margin. A similar spec CD32 with 2X speed CD-ROM drive has 299 UKP and A1200's 399 UKP. A1200 should have 2MB Chip RAM and 1 Mb Fast RAM (and $20 DSP3210 @ 42 Mhz or 56 Mhz). It's about $4 difference between 68EC020-16 and 68EC020-25.

I agree with adding the 1MiB fast ram and either a 68EC020@28MHz or a 68EC030@28MHz. The 68EC030 is more compelling with the fast memory upgrade as the data cache and cache line bursts can be used in fast memory. More address lines for more address space are worthwhile for upgrades at a small cost. I'm not as big of fan of the DSP. DSPs are good for doing dedicated tasks but I don't like the idea of trying to make it general purpose for all programs to boost performance. DSPs have high theoretical performance but they are difficult to program which is the opposite philosophy of the 68k. I would rather upgrade the CPU and/or memory than add a DSP. I'm not saying it was wrong for high end systems but it makes less sense for small footprint affordable hardware.

Hammer Quote:

A1200's craftmanship wasn't good due to the maximum "PCJr" mindset from Bill "PCJr" Sydnes. Ali fired Bill Sydnes after the A300(A600) debacle

Has the A1222 debacle surpassed the A600 debacle? Not only is the A1222 price increase percentage higher but AmigaNOne compatibility is worse than A600 Amiga compatibility. The A1222 had many years of delays as the hardware value dropped and production cost increased. Then, Trevor may have been able to unload the bastard CPUs during COVID chip shortage price spikes. If he could have upgraded the hardware to later SoC chips with e500mc cores and standard FPUs, maybe the hardware would be a little bit of an upgrade from the Sam but it is a downgrade at a higher price and without a standard PPC FPU.

Trevor is the guy in charge more like Irving Gould though so maybe it is better to compare to the "no new chips" mandate that did more harm than the A600. Nobody could tell Irving he didn't know what he was doing. CBM with the 68k Amiga had an advantage over IBM, Apple and Atari in the PC market but squandered having a system 5 years ahead of the competition. The Amiga had a lost decade under Irving and now over 2 lost decades under Trevor. There is nothing Amiga fans can do but wear a bag over our heads and wait for extinction, maybe hope for extinction if the situation gets any more embarrassing.

Last edited by matthey on 17-May-2024 at 09:02 PM.

@Hammer

Quote:

According to ex-Commodore UK's MD, A1200 has a healthy profit margin.

How many A1200s were produced in UK?
What R&D expences did CBM UK have?

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@matthey

If Amiga fans don't like the A1222 they can source a X5000 or just buy a Vampire V4 for their classic Amiga or get a brand new Vampire V4SA and/or setup a MorphOS system. Lots of choices in today's Amiga hobby market.

Last edited by DiscreetFX on 18-May-2024 at 12:13 AM.
Last edited by DiscreetFX on 17-May-2024 at 11:33 PM.

_________________
Sent from my Quantum Computer.

@kolla

Quote:

kolla wrote: @Hammer Quote: According to ex-Commodore UK's MD, A1200 has a healthy profit margin. How many A1200s were produced in UK? What R&D expences did CBM UK have?

For 1993 context:
"Meanwhile our UK-based Consumer Products business was in a very healthy position – we had taken massive orders for the Amiga 1200, scheduled for specific volume deliveries from September through to Christmas from all the major retailers. It was a profitable product with a healthy margin for Commodore at both corporate head office and in the UK."

AGA Amiga install base
Germany:
Amiga 1200 = 95,500
Amiga CD32 = 25,000
Amiga 4000/030 = 7,500
Amiga 4000/040 = 3,800
Sub-total: 131,800

https://web.archive.org/web/20230726021525/http://www.bambi-amiga.co.uk/amigahistory/sales.html

UK:
Amiga 1200 (Oct - Dec 1992) = 44,000 (Amiga Format May 1993)
Amiga 1200 (Jan - Aug 1993) = 100,000 (Amiga Format September 1993)
Amiga 1200 (Xmas 1993) = 160,000 (Amiga Format 56 Feb 1994)
Amiga CD32 (Xmas 1993) = 70,000 (Amiga Format 56 Feb 1994)
A4000 sales weren't disclosed.
Sub-total: 374,000

AGA install base from Amiga's strongest markets has exceeded 505,800.

Last edited by Hammer on 20-May-2024 at 07:38 AM.
Last edited by Hammer on 20-May-2024 at 07:00 AM.
Last edited by Hammer on 20-May-2024 at 06:59 AM.
Last edited by Hammer on 20-May-2024 at 06:57 AM.
Last edited by Hammer on 20-May-2024 at 06:57 AM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@matthey

Quote:

Most early consoles had separate memories for different purposes. The advantage was that some memory could be faster without requiring as much of it and most 8 and 16 bit CPUs could only address a 64kiB bank of memory anyway. The disadvantages are losing some memory in the partitioning and increased complexity in programming. The Amiga chipset memory could have been smaller and used VRAM if it had only been for a console but higher resolutions were used on the desktop. Jay Miner's Ranger design was an attempt to unify the Amiga memory but would have required more VRAM than most hardware. This would have been more expensive but offered better performance and was ahead of its time. The 1996 N64 has a large 4MiB unified memory subsystem and more powerful CPU than previous consoles which is what the Amiga needed to be competitive with less chipset hardware. CBM didn't upgrade anything though and let the Amiga slip from low end to obsolete. Modern consoles use unified memory and even a HSA to keep caches coherent and unified system wide.

For 16-bit game consoles, 64 KB VRAM is enough for low-resolution gaming. ROM stores the game content.

N64 has 4MB Rambus RDRAM. RDRAM standard failed in the PC world due to cheaper DDR SDRAM competition. PC world rejected RDRAM and this shows Intel's influence limitation.

My comment's context is for the Fast Page DRAM era.

SGRAM and GDDR SDRAM displaced VRAM.

Quote:

I agree with adding the 1MiB fast ram and either a 68EC020@28MHz or a 68EC030@28MHz. The 68EC030 is more compelling with the fast memory upgrade as the data cache and cache line bursts can be used in fast memory. More address lines for more address space are worthwhile for upgrades at a small cost.

For most use cases, 68EC030-25's higher price with a tiny data cache is not worth it.

68EC030 doesn't have arithmetic DSP/RISC intensity.

https://archive.computerhistory.org/resources/access/text/2013/04/102723262-05-01-acc.pdf
Page 119 of 981

For 1992
68000-12 = $5.5
68EC020-16 PQFP = $16.06, it's $15 in 1993 Q1.
68EC020-25 PQFP = $19.99, it's $18 in 1993 Q1.

68EC030-25 PQFP = $35.94
68030-25 CQFP = $108.75

68040-25 = $418.52
68EC040-25 = $112.50
---
Competition

AM386-40 = $102.50
386DX-25 PQFP = $103.00

486SX-20 PQFP = $157.75
486DX-33 = $376.75
486DX2-50 = $502.75

Quote:

Has the A1222 debacle surpassed the A600 debacle? Not only is the A1222 price increase percentage higher but AmigaNOne compatibility is worse than A600 Amiga compatibility. The A1222 had many years of delays as the hardware value dropped and production cost increased. Then, Trevor may have been able to unload the bastard CPUs during COVID chip shortage price spikes. If he could have upgraded the hardware to later SoC chips with e500mc cores and standard FPUs, maybe the hardware would be a little bit of an upgrade from the Sam but it is a downgrade at a higher price and without a standard PPC FPU.

I wonder who advised Trevor on buying the "discounted" kitbash CPUs.

It's Freescale's stupidity for designing kitbash PowerPCs.

Quote:

Trevor is the guy in charge more like Irving Gould though so maybe it is better to compare to the "no new chips" mandate that did more harm than the A600. Nobody could tell Irving he didn't know what he was doing. CBM with the 68k Amiga had an advantage over IBM, Apple and Atari in the PC market but squandered having a system 5 years ahead of the competition.

Don't focus on IBM. The clones are the real threats. IBM didn't design fast VGA like ET4000.

Last edited by Hammer on 20-May-2024 at 07:52 AM.
Last edited by Hammer on 20-May-2024 at 07:52 AM.
Last edited by Hammer on 20-May-2024 at 07:51 AM.
Last edited by Hammer on 20-May-2024 at 07:35 AM.
Last edited by Hammer on 20-May-2024 at 07:21 AM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

Hammer Quote:

For 16-bit game consoles, 64 KB VRAM is enough for low-resolution gaming. ROM stores the game content.

The Amiga as a budget gaming PC competed against the consoles but it had to have more graphics memory for higher resolutions. CBM changed Agnus to support larger blits to better support the large flat address space advantage of unified chipset memory but rejected the cost of larger than usual VRAM that Jay Miner wanted with the Ranger chipset. This resulted in consoles in the late 1980s and 1990s often having a graphics performance advantage over the Amiga by using limited amounts of VRAM. Even for AGA where CBM borrowed from AAA which could use VRAM or DRAM, CBM rejected VRAM. The CD32 was the right idea and almost saved CBM but it would have been more competitive with more performance and this required either VRAM or 1MiB of additional fast memory. Interestingly, an earlier ROM based Amiga console would have avoided the contention between the CPU and chipset as the CPU accessing the ROM avoids the bottleneck. CBM gave the world the ROM based C64GS console in late 1990 showing just how out of touch they were.

Hammer Quote:

N64 has 4MB Rambus RDRAM. RDRAM standard failed in the PC world due to cheaper DDR SDRAM competition. PC world rejected RDRAM and this shows Intel's influence limitation. My comment's context is for the Fast Page DRAM era. SGRAM and GDDR SDRAM displaced VRAM.

Unified memory became more popular as the early memory bandwidth limitation was reduced by VRAM and then DDR SDRAM. The Amiga had unified memory without VRAM or fast memory meaning performance was reduced.

Hammer Quote:

For most use cases, 68EC030-25's higher price with a tiny data cache is not worth it. 68EC030 doesn't have arithmetic DSP/RISC intensity. https://archive.computerhistory.org/resources/access/text/2013/04/102723262-05-01-acc.pdf Page 119 of 981 For 1992 68000-12 = $5.5 68EC020-16 PQFP = $16.06, it's $15 in 1993 Q1. 68EC020-25 PQFP = $19.99, it's $18 in 1993 Q1. 68EC030-25 PQFP = $35.94 68030-25 CQFP = $108.75

The price of the 68EC030 was dropping faster than the 68EC020 and availability and discounts were likely to be better at the low end of the clock rating range for the 68EC030. The 68EC030 also has cache burst in addition to the small data cache. Cache line bursts are very effective for the instruction cache as most code is accessed sequentially. Even the tiny 256 byte data cache is effective at reducing memory traffic and long loads and stores to and from memory. Back then, it was common for the execution pipeline to stall on stores due to the lack of a store buffer.

Hammer Quote:

I wonder who advised Trevor on buying the "discounted" kitbash CPUs. It's Freescale's stupidity for designing kitbash PowerPCs.

Trevor said he was "recommended" the QorIQ P1022 by a business partner in one of the Amiga show videos. As I recall, the show was Amiga38 and the business partner was likely Ultra Varisys. It looks like a case of blind trust with no tech people at A-Eon to review the hardware choice. The 68k AmigaOS used very slow floating point math libraries but AmigaOS 4 made the PPC FPU standard.

https://wiki.amigaos.net/wiki/Math_Libraries#Introduction Quote:

The math libraries described in this section are implemented in 68K assembly language and are designed to only be callable from 68K code. They are provided for backwards compatibility with pre-4.0 versions of AmigaOS.

A standard FPU is very good and allows FPU code to be several times better performance than calling math libraries. The P1022 e500v2 cores use a different and incompatible PPC ISA and ABI. The ABI is especially important as the standard PPC ABI passes fp function arguments in FPU registers that don't exist with the e500v2 core resulting in corruption and alignment problems. Simply trapping and emulating the FPU instructions does not work as well as missing 68k FPU instructions using hardware FPU registers. The e500v2 GPR registers have to be stored, the FPU operation performed with results stored to FPU registers in memory and then the GPR registers restored for every instruction and this doesn't include trapping overhead. There is a run time patcher but it is very complex, predictably buggy and will likely never be reliable. One of the few things AmigaOS 4 has going for it was 3D support but standard 3D PPC programs run like molasses.

http://www.eliyahu.org/tabor/intro.html Quote:

That's a big deal. It wasn't in the 1980s, or even the early 1990s, when Commodore shipped several Amiga systems without a FPU. If we had a 68000, 68020, or 68030 (without the 68882 FPU chip), we just selected the non-FPU binary of our software and moved on. But floating-point mathematics is used in a ton of places today, especially in 3D graphics. So if the FPU isn't compatible, does that mean all that software is broken? No, but there's a catch. Just like during the transition that AmigaOS made from Motorola 68K processors to the PowerPC family, emulation is used to allow the A1222 to run software that was compiled for the traditional PowerPC ABI to run on the P1022 processor. In fact there are two emulators: a normal interpretative emulator -- where each instruction is executed one by one on the SPE, copying and restoring register data each time -- and a load-time emulator, where when you launch your program containing floating-point instructions, it does the translation as it's loading the binary into memory. The former is more compatible, the latter is faster. To be sure a real FPU is on the order of 20X faster than even the LTE emulation layer, but it works, and unless you're running something that is doing nothing but floating-point instructions, it will work quite nicely. For older 68K software the emulated 'FPU' is faster than what you'd get on a 060 anyway. This does mean that if you want to do things like 3D rendering or first-person shooter games the A1222 is probably not for you; for everything else using the FPU occasionally, it should be 'good enough' to where you'll be happy. Be sure to peruse the compatibility lists (applications, games) in this guide, though, before you buy.

Is the emulated FPU really "faster than what you'd get on a 060"? Most Quake versions didn't work but Quake II ran with 4-5fps where it is playable on a 68060. The game Freespace reportedly runs at 3fps and the 68060 is perfectly playable. Perhaps the e500v2 is the modern Cyrix 6x86 except A-Eon knew they needed good fp performance for their investment in 3D?

FPU latencies in cycles (single/double)
instruction | e500v2 | e500mc
FABS 1/6 8
FADD 4/6 8/10
FMUL 4/6 8/10
FDIV 29/32 38/68
FSQRT none none

The e500v2 core bastard FPU has good performance and better performance than the e500mc standard FPU replacement. The e500v2 is fully pipelined with most FPU instructions besides division having single cycle throughput where the e500mc FPU instructions usually have a repeat rate of 2 cycles for single precision and 4 cycles for double precision. The ISA and ABI can make a huge different though. I added the Pentium, 6x86, 68060 and now 88110 FPU latencies for easy comparison.

FPU latencies in cycles (single/double/extended if different)
instruction | Pentium | 6x86 | 68060 | 88110
FABS 1 2 1 none
FADD 3 4-9 3 3
FMUL 3 4-9 3 3
FDIV 19/33/39 24-34 37 13/23/26
FSQRT 70 59-60 68 none
FXCH 1 3 unnecessary unnecessary

The 88110 has a high performance extended precision fully pipelined FPU with low latencies. Two FPU instructions can be superscalar issued in parallel if they are to different FPU units. Superscalar FPU scheduling is very easy on the 88110 which improves performance. The 88110 has a minimal ISA and the ISA is important for performance. The 68060 could have borrowed the pipelined FADD and FMUL units and the unpipelined but well optimized FDIV unit to destroy the Pentium for both integer and fp performance but doubling the caches for the 68060+ was a higher priority. Less than optimal FPUs are used sometimes to save silicon space, reduce power and for other reasons. At least the 68060 retained enough of the standard 68k FPU ISA for compatibility and mediocre FPU performance which is more than can be said for the e500v2 core.

Hammer Quote:

Don't focus on IBM. The clones are the real threats. IBM didn't design fast VGA like ET4000.

A 68020+Ranger with VRAM for the price of a Mac would have been demoralizing to all the competition including workstation producers. A low end games machine just wasn't as much of a threat to business computers. Ironically, it was 3D games for high performance hardware which propelled PC clones ahead of the Amiga with a stronger emphasis on FPU performance. Doom, Quake and the first web browser were developed on expensive 68040 NeXT workstations. Jay Miner was on a path to further integrate and upgrade the Amiga which is easy to imagine competing with workstations like NeXT and bringing workstation performance to the PC but CBM put a stop to it. Jay wanted MMU, FPU and memory upgrades already for the original 68000 Amiga. With his vision, a standard FPU was more likely than under CBM or A-Eon leadership and could have allowed the Amiga to better compete with 3D.

@matthey

Quote:

Trevor said he was "recommended" the QorIQ P1022 by a business partner

More like business enemy.

Quote:

http://www.eliyahu.org/tabor/intro.html …But floating-point mathematics is used in a ton of places today, especially in 3D graphics. So if the FPU isn't compatible, does that mean all that software is broken? No, but there's a catch. … This does mean that if you want to do things like 3D rendering or first-person shooter games the A1222 is probably not for you

Yeah, it’s not for a whole bunch of people.

Where I’m scratching my head here, @matthey, is for whom is it?

Last edited by agami on 22-May-2024 at 04:46 AM.

_________________
All the way, with 68k

@matthey

Quote:

The Amiga as a budget gaming PC competed against the consoles but it had to have more graphics memory for higher resolutions. CBM changed Agnus to support larger blits to better support the large flat address space advantage of unified chipset memory but rejected the cost of larger than usual VRAM that Jay Miner wanted with the Ranger chipset.

ET4000AX's DRAMs can go down as low as 60 ns, 32-bit bus, core clock at 65 Mhz, and memory clock at 40 Mhz.

ET4000/W32i has FP (Fast Page) DRAM, 32-bit bus, core clock at 80 Mhz, and memory clock at 50 Mhz.

Amiga 16-bit Chip RAM was designed for 120 ns(?) DRAM timings and effectively frozen in 1985.

VRAM is not a requirement.

Changing chipset timings can cause backward compatibility problems due to interaction between multiple accelerated components.

Amiga's graphics capability (including object manipulators such as the Blitter) is directly linked to video 15kHz timing and the related DMA slots.

VGA monitor's timings have its foundation in IBM Professional Graphics Controller (PGA).
Stock IBM 5175 PGA monitor can display 1987 VGA with small adjustments. https://www.vogons.org/viewtopic.php?t=89084

Quote:

This resulted in consoles in the late 1980s and 1990s often having a graphics performance advantage over the Amiga by using limited amounts of VRAM. Even for AGA where CBM borrowed from AAA which could use VRAM or DRAM, CBM rejected VRAM. The CD32 was the right idea and almost saved CBM but it would have been more competitive with more performance and this required either VRAM or 1MiB of additional fast memory. Interestingly, an earlier ROM based Amiga console would have avoided the contention between the CPU and chipset as the CPU accessing the ROM avoids the bottleneck. CBM gave the world the ROM based C64GS console in late 1990 showing just how out of touch they were.

Slapping VRAM with a 3.5 Mhz graphics/audio chipset wouldn't significantly improve it.
Only Lisa has double-pumped 3.5 Mhz.

AGA has improvements for 2-cycle access Fast Page DRAM.

CD32 units sold was about 100,000 units. At least another CD32 65,000 motherboards for Commodore Canada/AmiTech's A2200 clone in the warehouse.

There are not enough CD32s to save Commodore.

AGA Amiga install base
Germany:
Amiga 1200 = 95,500
Amiga CD32 = 25,000
Amiga 4000/030 = 7,500
Amiga 4000/040 = 3,800
Sub-total: 131,800

https://web.archive.org/web/20230726021525/http://www.bambi-amiga.co.uk/amigahistory/sales.html

UK:
Amiga 1200 (Oct - Dec 1992) = 44,000 (Amiga Format May 1993)
Amiga 1200 (Jan - Aug 1993) = 100,000 (Amiga Format September 1993)
Amiga 1200 (Xmas 1993) = 160,000 (Amiga Format 56 Feb 1994)
Amiga CD32 (Xmas 1993) = 70,000 (Amiga Format 56 Feb 1994)
Sub-total: 374,000

AGA install base from Amiga's strongest markets: 505,800.

Amiga 2200 (based on CD32 board with "Agent 88" board), https://bigbookofamigahardware.com/bboah/product.aspx?id=19
65,000 CD32 motherboards assigned to Amitech's Amiga 2200.

Potential AGA install base: +570,800 units.

CD32 has a lower profit margin compared to A1200's "healthy profit margin".

David Pleasance warned Ali not to sell CD32 in 1993 since it has a lower profit margin when compared to A1200. This could be the difference in narrowing the last quarter's $8 million loss and they pulled the plug. A problem with a single source without insurance.

Quote:

Unified memory became more popular as the early memory bandwidth limitation was reduced by VRAM and then DDR SDRAM. The Amiga had unified memory without VRAM or fast memory meaning performance was reduced.

My comment for the discrete two-bus system is for the quicker fix.

3.5 Mhz Agnus or Alice needs large-scale changes before using something like Rambus RDRAM.

No game console designer has brought over 1985 graphics core into the mid-1990s latest memory tech.

N64's RDRAM used a 500 MT/s (250 Mhz, double rate data) clock on a 9-bit bus, providing up to 562.5 MB/s bandwidth. It has high random-access latencies.

For 1997, NVIDIA RIVA (NV3) used a 128-bit SDRAM @ 100 Mhz memory clock (1600 MB/s).
For 1995, NVIDIA NV1 used a 64-bit EDO @ 60 Mhz memory clock (480.0 MB/s).

Quote:

The price of the 68EC030 was dropping faster than the 68EC020 and availability and discounts were likely to be better at the low end of the clock rating range for the 68EC030. The 68EC030 also has cache burst in addition to the small data cache. Cache line bursts are very effective for the instruction cache as most code is accessed sequentially. Even the tiny 256 byte data cache is effective at reducing memory traffic and long loads and stores to and from memory. Back then, it was common for the execution pipeline to stall on stores due to the lack of a store buffer.

Again, 68EC030 doesn't have arithmetic intensity like DSP3210 or ARM60.

Prove 680EC030 @ 25 Mhz can deliver 25 MIPS on a simple ADD or MUL instruction.

$100 68EC040-25 is useless for the Amiga. Unless you can convince Motorola to release a modified 68EC040 with 68030-style cache handling, Motorola's 68K is not price-competitive with 3D processing.

Like the hardware barrel shifter, 68EC030 needs to be modified with just a few ADD/MUL instructions (a few instructions useful for fixed-point 3D) to be hardware implemented instead of ROM'ed microcode.

68040 approach is most 68K instructions are hardware implemented or bust, take it or leave it. 68040's transistor budget is equivalent to Amiga Hombre.

On the Athlon K8 CPU, X86 instructions are divided by three decoder paths i.e. Fast Single (1 X86 instruction to 1 Macro-Ops), Fath Double (1 X86 instruction broken to two Macro-Ops), and MicroCode(1 X86 instruction broken to many Macro-Ops). As the transistor budget increased, more instructions were moved into the Fast Single decoder. Most common instruction usage goes into Fast Single or Fast Double. Targeting a popular benchmark's instruction usage is a major factor for the instruction's performance category i.e. optimized towards benchmark tactics.

It's Motorola's fault for not developing its "ARM ReadySystem" (with MMU) initiative to be less dependent on a single Apple and many "68000 is good enough" embedded customers.
Do not depend on Apple, they are not loyal.

NXP has joined the "ARM ReadySystem" initiative.

Quote:

A standard FPU is very good and allows FPU code to be several times better performance than calling math libraries. The P1022 e500v2 cores use a different and incompatible PPC ISA and ABI. The ABI is especially important as the standard PPC ABI passes fp function arguments in FPU registers that don't exist with the e500v2 core resulting in corruption and alignment problems. Simply trapping and emulating the FPU instructions does not work as well as missing 68k FPU instructions using hardware FPU registers. The e500v2 GPR registers have to be stored, the FPU operation performed with results stored to FPU registers in memory and then the GPR registers restored for every instruction and this doesn't include trapping overhead. There is a run time patcher but it is very complex, predictably buggy and will likely never be reliable. One of the few things AmigaOS 4 has going for it was 3D support but standard 3D PPC programs run like molasses.

I'm already aware of this.

Quote:

Trevor said he was "recommended" the QorIQ P1022 by a business partner in one of the Amiga show videos. As I recall, the show was Amiga38 and the business partner was likely Ultra Varisys. It looks like a case of blind trust with no tech people at A-Eon to review the hardware choice. The 68k AmigaOS used very slow floating point math libraries but AmigaOS 4 made the PPC FPU standard.

Ultra Varisys has weak experience in desktop computing and has conflicting interests. Trevor has the hot potato.

Last edited by Hammer on 22-May-2024 at 02:07 PM.
Last edited by Hammer on 22-May-2024 at 02:05 PM.
Last edited by Hammer on 22-May-2024 at 02:03 PM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

agami Quote:

More like business enemy.

It is possible, if not likely, that most of the hardware business partners of A-Eon primarily design niche embedded hardware. Non-standard for embedded hardware just means a recompile of the software which is the way to solve the A1222 floating point bottleneck. Embedded systems use a more limited amount of software and the source code of the software is usually available which is one of the reasons why open source software (and open hardware) are more appealing for embedded markets.

agami Quote:

Yeah, it’s not for a whole bunch of people. Where I’m scratching my head here, @matthey, is for whom is it?

The A1222 was supposed to be the low price hardware to expand the AmigaOS4 user base. The original hope was likely higher production volumes to also reduce the production cost. If more affordable hardware was successful, mass production may have been considered. The A1222 has a very narrow appeal though regardless of price. Even at the same price as THEA500 Mini, I believe THEA500 Mini would easily outsell it and THEA500 Mini hardware leaves a lot to be desired too.

The A1222 is for AmigaOS4 zealots. Eliyahu seems optimistic about the hardware. Hans de Ruiter was considering the A1222 for a laptop project. There is a need for affordable, mid performance, low power, standard and compatible Amiga hardware. It is a matter of survival for Amiga hardware but the A1222 is hopeless as a solution as it is a downgrade built on a two decade old broken business model. The $600 Amiga 300 had the same problem that not only was it too expensive and low value but the hardware was wrong by losing compatibility without increasing performance. Most Amiga users already had 68000+OCS/ECS hardware in the same way most AmigaOS4 users already had a X1000 or X5000. The lack of A1222 value and compatibility makes it less appealing to potential new customers as well. New customers would already be giving up 68k Amiga compatibility to get mediocre performance where more compatible and cheaper performance options exist on the 68k now despite the Frankenmiga hardware but those of us left have been dealing with reviving and upgrading the ancient hardware monster forever. The sane Amiga masses already left the Amiga community due to the lack of competitive and compatible 68k Amiga hardware. THEA500 Mini showed it is possible to bring many of them back but the hardware was not good enough to keep them.

Hammer Quote:

ET4000AX's DRAMs can go down as low as 60 ns, 32-bit bus, core clock at 65 Mhz, and memory clock at 40 Mhz. ET4000/W32i has FP (Fast Page) DRAM, 32-bit bus, core clock at 80 Mhz, and memory clock at 50 Mhz. Amiga 16-bit Chip RAM was designed for 120 ns(?) DRAM timings and effectively frozen in 1985.

CBM liked using the cheapest possible memory.

https://retrocomputing.stackexchange.com/questions/5994/amiga-memory-bandwidth Quote:

The Amiga A500 used 41256-15 DRAM chips for its onboard memory. According to the datasheet these had a cycle time of 260ns, so could easily match the 280ns timing constraint with 20ns to spare (although the board had two banks of them, and switching between the banks likely used up most of that 20ns). The 375ns limit really only applied to the 4116 and 4164 type chips that were manufactured in huge bulk and were extremely cheap. Faster chips were available even back to the early 80s, but cost more money so designers of low-end micros tended to avoid using them where possible (cf the 4816A chips in the BBC micro we discussed previously, which were necessary in order to let a 6502 run at 2MHz with capacity left for video access between the CPU's accesses), especially when working with either 1MHz 6502 or 4MHz Z80 designs where faster RAM would have been useless.

The 6502 had more aggressive memory access timing than the 68000 which is partially responsible for the better performance/MHz but it made it more expensive to increase the clock speed as more expensive memory was needed. ARM2 also had more aggressive memory access timing that increased memory cost. CBM liked the 68000 Amiga design because they could use the cheapest possible memory. More performance was possible with better memory but CBM chose the lowest common denominator to turn the Amiga into a C64 replacement.

Hammer Quote:

VRAM is not a requirement. Changing chipset timings can cause backward compatibility problems due to interaction between multiple accelerated components. Amiga's graphics capability (including object manipulators such as the Blitter) is directly linked to video 15kHz timing and the related DMA slots. ... Slapping VRAM with a 3.5 Mhz graphics/audio chipset wouldn't significantly improve it. Only Lisa has double-pumped 3.5 Mhz. AGA has improvements for 2-cycle access Fast Page DRAM.

VRAM was certainly not a requirement. It is just one way to increase chip memory bandwidth but there were not many options to double the bandwidth early in the Amiga life. Dual ported memory is a good fit for the Amiga chipset design as one port can be used for chipset accesses while the other port allows CPU accesses. Timings would have changed but this is no different than adding fast memory which unbottlenecks the CPU allowing it to outpace a chipmem only Amiga. I wonder if the 68000 TAS and later CAS and CAS2 would have operated correctly with VRAM. The Amiga hardware would have been more professional if the 68020+ standard included properly working TAS, CAS and CAS2 instructions. Maybe the Ranger chipset could steal the CPU's VRAM port and half or all the cycles so these instructions would still not be possible. Jay Miner's explanation of the Ranger chipset was brief and I'm not a hardware guy.

Later, page mode DRAM could be used to increase memory bandwidth which the Amiga first used in the Amiga 3000. AGA finally increased chipmem bandwidth in a less compatible way as far as alignment requirements but it allowed cheap DRAM memory and did not require updating the Amiga chipset to a newer chip process that could be clocked higher. This ended up backfiring when CBM needed to clock up the CPU and chipset together to improve Amiga competitiveness with a 68EC030@28MHz Amiga and a single chip chipset. Dave Needle and RJ Mical had integrated the much larger 3DO chipset into one chip in a couple of years which CBM failed to do with the Amiga chipset in about a decade. Except for the Lisa chip, CBM was still using an old NMOS process from the mid 1970s for AGA.

https://en.wikichip.org/wiki/5_µm_lithography_process Quote:

The 5μm lithography process was the semiconductor process technology used by some semiconductor companies during the mid 1970s.

CBM's upgrade strategy was to produce a product until EOL and then replace it. That is what they did with the C64 to Amiga upgrade and that is what they planned to do with the C64 replacement Amiga.

Hammer Quote:

CD32 units sold was about 100,000 units. At least another CD32 65,000 motherboards for Commodore Canada/AmiTech's A2200 clone in the warehouse. There are not enough CD32s to save Commodore.

If CBM had delayed the NA launch which was predictably weak and produced only PAL versions of the CD32 instead, they potentially could have avoided the dead inventory NTSC CD32s. If the CD32 had 68EC030@28MHz with AA+ and 1MiB of fast memory, CD32 sales would have likely been more swift. The CD32 had the right appeal and backward compatibility which allowed to bring the large Amiga game library forward but had trouble playing the new 3D games like Alien Breed 3D that was downright sluggish compared to everything else on the market.

Hammer Quote:

Potential AGA install base: +570,800 units. CD32 has a lower profit margin compared to A1200's "healthy profit margin". David Pleasance warned Ali not to sell CD32 in 1993 since it has a lower profit margin when compared to A1200. This could be the difference in narrowing the last quarter's $8 million loss and they pulled the plug. A problem with a single source without insurance.

The profit margin of the CD32 was healthy so the only concern was if CD32 sales were stealing Amiga 1200 sales. I expect the larger problem was that CD32 prices were reduced and games bundled to improve the value. The CD32 had a successful launch but it was not the hit that flew off shelves that CBM needed because their financial situation did not allow them to update the hardware to be more competitive. The no new chips directive, the Amiga 600 dead inventory and the pivot away from the Amiga to the PC clone market right before the bottom fell out of the market was devastating. CBM made too many big mistakes leading up to the CD32 and Amiga 1200 which were the right products but partially handicapped and lower value hardware as a result.

Hammer Quote:

No game console designer has brought over 1985 graphics core into the mid-1990s latest memory tech.

CBM had Hombre and if that didn't work out they could have licensed a 3rd party 3D chipset for integration into the Amiga chipset. There were already several 3D chipsets available for licensing in the late 1990s as they competed to be used in the Sega Dreamcast.

https://en.wikipedia.org/wiki/3dfx#Dreamcast Quote:

In 1997, 3dfx was working with entertainment company Sega to develop a new video game console hardware platform. Sega solicited two competing designs: a unit code-named "Katana", developed in Japan using NEC and Imagination Technologies (then VideoLogic) technology, and "Blackbelt", a system designed in the United States using 3dfx technology. However, on July 22, 1997, 3dfx announced that Sega was terminating the development contract. Sega chose to use NEC's PowerVR chipset for its game console, though it still planned to purchase the rights to 3dfx's technology in order to prevent competitors from acquiring it.

Could the Amiga be upgraded with 3D? I believe so. AA+ would have added chunky support which is necessary for 3D. The 3dfx chipset was initially 3D only under an existing 2D chipset. The AmigaOS is flexible enough to support hardware upgrades as we found out with Cybergraphics/P96 and Warp3D. This is much more difficult for primitive consoles like the Sega Genesis or SNES without an OS. The AmigaOS is so thin that the CD32 could afford to have an advanced OS and this means 3D could likely be integrated into the Amiga chipset and it would seem surprisingly modern.

The largest obstacle to CBM adding 3D support was CBM. They were terrible at licensing, R&D, integration and marketing. They were good at cost reducing and manufacturing. If CBM was healthy, they could have acquired a business like 3dfx. Actually, a merger between CBM and 3dfx may have been easier as their products have synergies with CBM having a 2D Amiga chipset and 3dfx having a 3D chipset. The Amiga market with standard integrated 3D could have grown and the 3dfx market for PC clones with 2D+3D graphics cards would be interesting. I would hope it wouldn't be managed by CBM as no VRAM and no new chips mandates would have quickly killed any synergies.

Hammer Quote:

Again, 68EC030 doesn't have arithmetic intensity like DSP3210 or ARM60. Prove 680EC030 @ 25 Mhz can deliver 25 MIPS on a simple ADD or MUL instruction. $100 68EC040-25 is useless for the Amiga. Unless you can convince Motorola to release a modified 68EC040 with 68030-style cache handling, Motorola's 68K is not price-competitive with 3D processing. Like the hardware barrel shifter, 68EC030 needs to be modified with just a few ADD/MUL instructions (a few instructions useful for fixed-point 3D) to be hardware implemented instead of ROM'ed microcode. 68040 approach is most 68K instructions are hardware implemented or bust, take it or leave it. 68040's transistor budget is equivalent to Amiga Hombre.

The 68020 and 68030 were still heavily microcoded which reduced performance. The early ARM CPU instructions generally used fewer cycles but the 68020/68030 made up ground with more flexible and fewer instructions to execute while being easier to program. The AT&T DSP3210 has impressive performance and supports single precision floating point. It was a good DSP choice and would have been able to offload the heavy math from the 68k but it is difficult to program and making it standard on the Amiga would have made it a requirement for all future Amiga hardware. It was better to improve CPU performance and incorporate DSP like features in the CPU. The 68040 had much improved instruction timings and FPU support for DSP algorithms making adding a DSP less appealing. A DSP would likely be relegated to I/O processing and audio mixing with a much easier to program 68060 upgrade. Add a SIMD or vector unit to the 68060 and an external DSP is baggage. If the DSP was incorporated into the Amiga chipset it would be fine but an extra chip that can't be incorporated is baggage that needs to be maintained for compatibility. The 68040 was a big jump in price and performance from the 68030 which discouraged upgrading. Part of the problem was that 68040 development had problems and was late but this may be a better topic for the "why Motorola and 68k failed" thread.

Hammer Quote:

On the Athlon K8 CPU, X86 instructions are divided by three decoder paths i.e. Fast Single (1 X86 instruction to 1 Macro-Ops), Fath Double (1 X86 instruction broken to two Macro-Ops), and MicroCode(1 X86 instruction broken to many Macro-Ops). As the transistor budget increased, more instructions were moved into the Fast Single decoder. Most common instruction usage goes into Fast Single or Fast Double. Targeting a popular benchmark's instruction usage is a major factor for the instruction's performance category i.e. optimized towards benchmark tactics.

You are jumping ahead a ways aren't you? One website says the Athlon K8 "Hammer" has 105,900,000 transistors and they were still removing microcode? The 68060 removed all microcode at 2,500,000 transistors. It was incomplete compared to some earlier 68k CPUs but was very compatible. It likely could be complete with missing instructions including FPU instructions in less than 3,000,000 transistors. It's possible the missing FPU instructions would be better off using microcode or 68k code. Without the traps for missing FPU instructions, the 68k support code performance is not too far behind a hardware implementation. There is room for significant improvements to the code and performance of the code with 68k FPU enhancements too.

Last edited by matthey on 23-May-2024 at 10:23 PM.

@matthey

At least in my experience with x1000 and SAM460, its often Hyperion / OS4 not fully supporting and utilizing the board. Linux on both boards shows its full potential. AmigaOS 4 drivers are often half baked or unoptimized, if existing at all and basically OS remains "in beta" for almost any AmigaOne system out there.

Compared to Trevor there is no other AmigaOne hardware choice, due to narrow mindness likely.

Other approach would be like MorphOS - using Apple PPC hardware
OR bold one - taking OS to level it could use e.g. Talos board and IBM POWER series. But that would need serious investment in OS first, making it SMP and 64 bit at least.

I don`t blame Trevor - Varisys so much (except for high end prices) - software side bothers me more. And Trevor tied to work that out with more focus on software recently.

X5000 MorphOS support was nice step forward, but generally Trevor does not much support AROS or OS3 - except when its commercial viable. That said, still waiting for Enhancer for oS 3.2 :D

In general, as seen in other thread we do have some hardware choices, be it Classic accells, FPGA boards, MOS compatible systems. Its bigger oS advancement and productivity modern software we mostly lack on to be at least competitive and viable choice outside retro.

_________________
Future Acube and MOS supporter, fi di good, nothing fi di unprofessionals. Learn it harder way!

@matthey

Quote:

CBM liked using the cheapest possible memory.

The docs for 41256-15 is http://matthieu.benoit.free.fr/cross/data_sheets/HYB41256.pdf

41256-15's random read or write cycle time = 260 ns, RAS access = 150 ns, CAS access = 75 ns.

260 ns is 3.846 Mhz effective.

Quote:

CBM had Hombre and if that didn't work out they could have licensed a 3rd party 3D chipset for integration into the Amiga chipset. There were already several 3D chipsets available for licensing in the late 1990s as they competed to be used in the Sega Dreamcast. .

3DFX Voodoo 1 was released in 1996.

Each of Voodoo 1's SST1 has 1 million transistors and is divided into two ASIC chips i.e. FBI and TMU. Clocked at 50 Mhz.

Amiga Hombre's completion assumes the original Commodore International is still alive.

If Commodore licenses 3DFX GPU IP, what is Commodore's value-added proposition when Commodore does not have a strong 1st party game studio? Just a middleman? "AmigaOS" by itself is not enough. Commodore's Amiga Vision is not strong enough to change the demand factor for the AmigaOS platform.

Sony and Nintendo could survive licensing 3rd party GPU IP since they both have very strong 1st party game studios.

Microsoft could survive on 3rd party PC clones when Microsoft has strong 1st party MS Office developer teams that feeds MS Windows's demand factor. Learn from IBM's OS/2 losing to Windows 3.0. It started from Windows 2.0 and Mac ports of Excel 2.1 in 1988 and Word 2.0 in 1989. "OS/2" by itself is nothing to many people.

For me and many others, IBM OS/2 doesn't have native ports of Doom, enough said. Windows 95 has /DirectX/Doom 95 and WinQuake in Jan 1997.

Apple has Claris (ClarisWorks 1.0 in 1991, AppleWorks 5.0 in 1997), Final Cut Pro (acquired Final Cut Pro development team in 1998), and Logic Pro (acquired Emagic in July 2002).

Microsoft is rumored to be preparing a $16 billion offer to buy Valve Steam.

Quote:

You are jumping ahead a ways aren't you? One website says the Athlon K8 "Hammer" has 105,900,000 transistors and they were still removing microcode? The 68060 removed all microcode at 2,500,000 transistors.

68060 dropped some 68K instructions and the trap emulates the missing instructions with slow performance.

68060 is not a 100 percent 68000's instruction set. Motorola's record for backward compatibility is inferior to AMD's or Intel's.

K8 still has a microcode path that is handled by a microcode engine which is a small processor on its own i.e. using a tiny embedded CPU to translate X86 instructions for the main CPU core. Microcode is also used for applying firmware patches with any CPU bugs post-sale instead of executing a product recall or telling software vendors to change their close source legacy software post-sale.

For Amiga's 68060, the MMULibs software checks the revision of the chip when the computer boots and disables some problematic features. MMULibs is on the OS side instead of being transparent from the OS.

K8's Fast Single and Fast Double decoders are the hardware decoders for most X86 instructions and they are geared for gaming the benchmarks and C++ complier's cooperation. There's a reason for MS's and Intel's investments in the C++ compiler.

Are you aware of the gap between CPU and external memory performance?

Examples
Athlon +1 Ghz against DDR-400 (400MT/s, 200Mhz, 64 bit).
Athlon 600 Mhz against DDR-200 (200MT/s,100Mhz, 64 bit).

Faster memory types have a larger latency even with SDRAM types e.g. DDR4 vs DDR5.


The out-of-order hardware will be significantly large in the attempts to hide or migrate the gap between the CPU core and external memory.

Last edited by Hammer on 25-May-2024 at 06:17 AM.
Last edited by Hammer on 24-May-2024 at 06:39 AM.
Last edited by Hammer on 24-May-2024 at 06:27 AM.
Last edited by Hammer on 24-May-2024 at 06:25 AM.
Last edited by Hammer on 24-May-2024 at 06:21 AM.
Last edited by Hammer on 24-May-2024 at 06:16 AM.
Last edited by Hammer on 24-May-2024 at 05:47 AM.
Last edited by Hammer on 24-May-2024 at 05:45 AM.
Last edited by Hammer on 24-May-2024 at 05:35 AM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB