Unreleased Amiga chipset better than anything today?

[Locutus]

Originally Posted by chb

While it is certainly off topic, this is one of the more interesting information in this thread. Are you sure they were implementations of the 68020 architecture/ISA, and not the 68000? All the information I can find on the web shows that the models with a custom bitslice CPU (DSP160/DN160/460/660) were released in the beginning of 1984 or earlier, before the 68020 was even announced.


It also would make more sense for them to do their own (improved) implementation of the 68000 instead of the 68020, as the 68000 did not fulfill the Popek/Goldberg virtualization requirements; AFAIR Apollo even had two 68000 operating in lockstep on their early machines as a workaround.

I think you are right, they where probably 68000 only with some form of MMU. (ill chalk this up to a brainfart on my part)



Tangentially: Weren't the 68020 bitfield instructions added for Hewlett Packard for use in type setting and printing applications?

[pandy71]

Seem this is more tricky - found information that DN460 used real 68010. Am2900 implementation probably also emulated 68010 instruction list.

https://en.wikipedia.org/w/index.php...0-family_chips

Source:

Originally Posted by https://en.wikipedia.org/wiki/Talk:AMD_Am2900?useskin=vector#Incorrect_information_re_Apollo

Incorrect information re Apollo

The statement that "The Apollo Computer Tern family: DN460, DN660 and DSP160. All used the same system board emulating the Mc68010 instruction set" is at least partially wrong. The Apollo DN460, at least, used a real Motorola 68010 processor because I owned a DN460 and swapped it's 68010 into my Amiga for a time. The 680x0 architecture would've been too complex to emulate practically in any non-ASIC form in the 1980's anyway. My guess is that it was used as part of the massive FPU boards that used some 'SPAD' firmware. Rabit (talk) 14:14, 7 June 2017 (UTC)

[hammer]

Originally Posted by pandy71

68k was on opposite extremely popular in automation https://en.wikipedia.org/wiki/VMEbus?useskin=vector

The automation market is useless for Amiga's general desktop computing market.

With Pi GPIO, the Raspberry Pi Foundation made its fortune in the DYI and small business IoT and automation markets. Raspberry Pi's ARM CPU IP-based SoC is lifted from the very competitive mobile phone market.

Raspberry Pi 4B and 5 Broadcom IGPUs are still rubbish for Android's Genshin Impact benchmark.

Originally Posted by pandy71

Most of those lawsuits was settled out of court - Intel right was possibility to raise case against other vendor and other vendors rights was to not settle case out of court - NEC case was judged on NEC favor so i've wrote nothing wrong. It was common practice to settle two independent teams, one using reverse engineering and second one using white room to create product. NEC made CPU sufficiently different than Intel so.

NEC's V30 8086 clones didn't progress into the 80286 and 80386 competition.

NEC V33 didn't include 286 and 386 MMUs which are used in the "world's best-selling" AT&T licensed Unix distro i.e. Microsoft-SCO Xenix.

Windows 2.x 386 and Xenix 386 were the major issues that culled 16-bit X86 cloners.

[hammer]

Originally Posted by pandy71

LOL - if you need help then don't talk with mirror, go to shrink.

There was no war between 68k and x86 - they made own ecosystems - x86 mostly PC and later embedded, 68k Unix and from beginning serious embedded (VME) - literally 68k dominated automotive, industrial automation, military - many critical applications...

That's a useless argument when Motorola has a thriving desktop and Unix 68K workstation platforms.

Motorola wasn't serious about the Unix market when 68K MMU wasn't integrated until 68030's 1987 release.

For Microsoft's Xenix, Intel integrated X86 MMUs for 286 and 386. Every 286/386 PC has Xenix potential. 80386-based PC laid the groundwork for Linux.

Microsoft's Xenix was "the world's best-selling" AT&T-licensed Unix distribution before Microsoft was distracted by IBM's OS/2 project.


https://techmonitor.ai/technology/mo...0_next_quarter

Date: April 19, 1994.
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.

With 68060's 1994 release, Motorola Inc. made negative remarks against Intel Pentium competition.


For Xmas Q4 1993 time period, Motorola's 68LC040 @ 25Mhz pricing allowed Apple to offer a competitive price vs performance solution against PC clone's 486SX-25/486-33 price range.

Amiga's uncompetitive price vs performance cost issue is Commodore's fault.

Your head is in the sand.

Originally Posted by pandy71

There was no 68k clones - it was officially licensed - x86 is opposite - one of many examples is NEC v20/v30 - better than Intel - with 386 it was even more different implementations...

AMD's X86 clones are licensed from Intel.

AMD's 80386 legal win against Intel is based on IBM's second X86 source contract enforcement. https://www.latimes.com/archives/la-...893-story.html
An arbitrator last year agreed that Intel had violated the agreement

AMD's credible second source insurance worked for X86's 64-bit transition when Intel attempted an anti-X86 Itanium 64-bit transition.

NEC V30 lacks 80286's MMU compatibility, hence no "business" Xenix.
NEC V33 is just a fast 8086 CPU type, hence NEC V30 series is just a toy 16-bit X86 CPU.

Originally Posted by pandy71

Price vs performance... obviously you are ignoring one of the most important factors... Commodore never offered PA or 68060 based machines...

Commodore went bust before offering PA-RISC or 68060-based SKUs.

Commodore offered Pentium-based PC SKUs before its bankruptcy, hence Commodore is aware of Intel's Pentium performance roadmap.

68060 is dependent on 68040 CPU socket infrastructure and Commodore wasn't able to mass produce 68040 CPU socket infrastructure at the same level as Apple's 68LC040/68040-based systems.

This topic is about Commodore NOT being bankrupt.

Originally Posted by pandy71

All above - not sure why are you arguing with me not with Motorola or other sources? Use time machine, go back to 1994 and argue with Motorola, Commodore, IBM, Intel etc...

This topic is about Commodore NOT being bankrupt.

Originally Posted by pandy71

Switch was never targeted to gamers but to people willing to play from time to time in particular types of games.

Switch is a dedicated games console in a handheld form factor. Switch targeted gamers who wanted a handheld form factor and Nintendo's 1st party and exclusive games. Steam Deck is a competitor in gaming handhelds.

Causal handheld gamers are on mobile phones.

Originally Posted by pandy71

Please don't mix various incompatible HW SoC's only because products based on them share some Pi in name...

I'm aware of wannabe Pi clones with incompatible Pi GPIOs.

RPi has many 3rd party add-on support and a large RPi community advantage over wannabe Pi clones.

Originally Posted by pandy71

Are you working in marketing? As a used cars seller...?
btw please provide list of "minor refinments" for 8364R4 vs 8364R7

Paula's 56kHz mode wouldn't work without ECS Denise and Agnus.

Originally Posted by pandy71

LoL - used cars seller you are truly...

You can't handle the truth e.g. Your "There was no war between 68k and x86" is bullshit.

Originally Posted by pandy71

Commodore outsourced Lisa to HP and VLSI - VLSI in 1993 - obviously there must be reason why Lisa from CSG was usually not populated on Amiga boards...

https://bigbookofamigahardware.com/b...t.aspx?id=1493

HP fabricated Lisa chip with 1992 markings.

Originally Posted by pandy71

But with remark that in PC 320x240 is in fact 320x480 in 60Hz...

Don't mix up rendering to display refresh rates.

Originally Posted by pandy71

Once again - Tseng ET4000 is a dumb framebuffer without HW acceleration - this is official Tseng information provided in ET4000 datasheet - if you can please provide alternate datasheet where Tseng provide information about HW acceleration implemented in ET4000 (not VLB ET4000Wxxxx).

Tseng Labs ET4000AX is one of many IBM 8514A clones.

https://ia904501.us.archive.org/28/i...sET4000VGA.pdf
At 16 of 26 pages,ET4000 has drivers for AutoCAD 2.6.
At the 17th page,ET4000 has drivers for AutoCAD R10/R11.
At the 18th page, ET4000 has IBM 8514/A Emulation Driver.
At the 22th page, ET4000 has Nth render drivers for 3D Studio and Auto Shade 2.0. Nth render is a protected mode display/rendering driver. It supports VGA, 8514/A, TIGA 2.0, and Nth Engine 150 adapters.
At the 23th page, the ET4000 Windows 3.0/3.1 driver removes the need for application-specific drivers. Microsoft is the big elephant in the room.

Originally Posted by pandy71

Of course you could create PGA clone - it was based on Intel 8088 CPU so it was not difficult task... - you could even made it on 8086 (NEC v30) so it could be faster than IBM PGA...

CPU by itself wouldn't make PGA. IBM PGA has CGA backward compatibility and it was quickly replaced by 1987 VGA and IBM 8514/A.

Originally Posted by pandy71

True (so no legacy compatibility seem to be not so big issue)

Nintendo usually has several strong 1st party exclusive games and low entry costs, hence breaking legacy compatibility is not a major issue.

Originally Posted by pandy71

But with remark that in PC 320x240 is in fact 320x480 in 60Hz...

December 1993, Doom doesn't have 320x480 resolution.

[pandy71]

Originally Posted by hammer

The automation market is useless for Amiga's general desktop computing market.

Well, this is bit contradictory if consider for example NASA applications for Amiga. But I pointed only that for Motorola embedded market was important at least same as general computing if not more (simply for embedded market you need way more CPU's than for desktop).

Originally Posted by hammer

With Pi GPIO, the Raspberry Pi Foundation made its fortune in the DYI and small business IoT and automation markets. Raspberry Pi's ARM CPU IP-based SoC is lifted from the very competitive mobile phone market.

Raspberry Pi 4B and 5 Broadcom IGPUs are still rubbish for Android's Genshin Impact benchmark.

Gibberish BS - additionally completely irrelevant to this discussion, topic and general Amiga - Boards based on Sitara (TI) are way better suited to Amiga needs than Broadcom SoC

Originally Posted by hammer

NEC's V30 8086 clones didn't progress into the 80286 and 80386 competition.

NEC V33 didn't include 286 and 386 MMUs which are used in the "world's best-selling" AT&T licensed Unix distro i.e. Microsoft-SCO Xenix.

Windows 2.x 386 and Xenix 386 were the major issues that culled 16-bit X86 cloners.

Unix on PC was very niche market from business perspective - it was also not popular in academic and seem also no one tried to create workstations around Intel CPU's (contrary to Motorola CPU's) - feel free to think why Intel was never chosen as good CPU for Unix - i can name at least few reasonable arguments for this.

And i agree - 286 was licensed by other companies as second source...

And Windows no need MMU till 3.0 version - 286 provided simply possibility to address more memory not process/resources protection.

Originally Posted by hammer

That's a useless argument when Motorola has a thriving desktop and Unix 68K workstation platforms.

Motorola wasn't serious about the Unix market when 68K MMU wasn't integrated until 68030's 1987 release.

For Microsoft's Xenix, Intel integrated X86 MMUs for 286 and 386. Every 286/386 PC has Xenix potential. 80386-based PC laid the groundwork for Linux.

Microsoft's Xenix was "the world's best-selling" AT&T-licensed Unix distribution before Microsoft was distracted by IBM's OS/2 project.

Well - many companies historically debunked your argument by real products - guess who is more right on this - many companies was founded around Motorola 68k Unix products, for 68000 Motorola offered 68451MMU so it is also argument to show that you are wrong (and some companies dissatisfied by 68451 added 1 cycle latency created own version of MMU or take another approach... anyway 68010 and 68012 shows that Motorola see Unix as important).

I don't care about Microsoft, Xenix etc - this is gibberish BS, completely irrelevant from this topic perspective, stop spamming thread with such garbage.

Originally Posted by hammer

https://techmonitor.ai/technology/mo...0_next_quarter

Date: April 19, 1994.
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.

With 68060's 1994 release, Motorola Inc. made negative remarks against Intel Pentium competition.


For Xmas Q4 1993 time period, Motorola's 68LC040 @ 25Mhz pricing allowed Apple to offer a competitive price vs performance solution against PC clone's 486SX-25/486-33 price range.

Amiga's uncompetitive price vs performance cost issue is Commodore's fault.

Your head is in the sand.

If my head is in the sand then yours seem to be deeply in the dingo dung.

This shows that APPLE was important customer for Motorola and that APPLE competed with PC's - definitely there is no Commodore there.

Originally Posted by hammer

AMD's X86 clones are licensed from Intel.

AMD's 80386 legal win against Intel is based on IBM's second X86 source contract enforcement. https://www.latimes.com/archives/la-...893-story.html
An arbitrator last year agreed that Intel had violated the agreement

AMD's credible second source insurance worked for X86's 64-bit transition when Intel attempted an anti-X86 Itanium 64-bit transition.

I don't care about Intel/AMD relations - both companies worked closely since 1976 (outcome of the Am9080 and Intel needs for second source supplier) - AMD products was almost never used in Amiga (perhaps with except of the Am95C60 in some very niche graphics cards)

Originally Posted by hammer

NEC V30 lacks 80286's MMU compatibility, hence no "business" Xenix.
NEC V33 is just a fast 8086 CPU type, hence NEC V30 series is just a toy 16-bit X86 CPU.

Users of the IBM PC XT no need MMU - they need cheap and fast CPU and NEC offered V20/V30. It was good for XT users - this is how market on PC works.

Originally Posted by hammer

Commodore went bust before offering PA-RISC or 68060-based SKUs.

Commodore offered Pentium-based PC SKUs before its bankruptcy, hence Commodore is aware of Intel's Pentium performance roadmap.

68060 is dependent on 68040 CPU socket infrastructure and Commodore wasn't able to mass produce 68040 CPU socket infrastructure at the same level as Apple's 68LC040/68040-based systems.

Allow me to once again quote yourself - "No shit Sherlock"...

Originally Posted by hammer

Switch is a dedicated games console in a handheld form factor. Switch targeted gamers who wanted a handheld form factor and Nintendo's 1st party and exclusive games. Steam Deck is a competitor in gaming handhelds.

Causal handheld gamers are on mobile phones.

lol

Originally Posted by hammer

I'm aware of wannabe Pi clones with incompatible Pi GPIOs.

RPi has many 3rd party add-on support and a large RPi community advantage over wannabe Pi clones.

lol - there is no such thing as "incompatible GPIO's" (GPIO = General Purpose Input Output) Sherlock

Originally Posted by hammer

Paula's 56kHz mode wouldn't work without ECS Denise and Agnus.

It can work with CPU and perhaps Copper (but consult this with Meynaf/Ross). And obviously you are wrong providing this as unique feature for "ECS/AGA Paula" - it will work with ICS Paula from Amiga (1000).

Originally Posted by hammer

You can't handle the truth e.g. Your "There was no war between 68k and x86" is bullshit.

yeah, yeah, "punk" (also quoting movie - not trying to insult you)

There was no war between 68k and x86 at least considering obvious 64k segmentation limit for x86 - you trying to create something artificial to justify your point

Originally Posted by hammer

https://bigbookofamigahardware.com/b...t.aspx?id=1493

HP fabricated Lisa chip with 1992 markings.

And VLSI in 1993

Originally Posted by hammer

Don't mix up rendering to display refresh rates.

Being engineer not marketing guy i don't mix things contrary to you.

Originally Posted by hammer

Tseng Labs ET4000AX is one of many IBM 8514A clones.

At the 18th page, ET4000 has IBM 8514/A Emulation Driver.

Don't mix things - going exactly same way you can claim that VGA is IBM8514/A clone as you can emulate it in software...

Check http://www.bitsavers.org/components/...oller_1990.pdf

Where Tseng claim that this is 8514 capable chip - where is bitblit unit?

Originally Posted by hammer

CPU by itself wouldn't make PGA. IBM PGA has CGA backward compatibility and it was quickly replaced by 1987 VGA and IBM 8514/A.

lol mixing things...

PGA offered "HW" accelerated functionality not present on CGA and "HW" functionality is implemented in general computing CPU such as 8088.
So yeah, there can be 6845 (CGA) as CRTC and 8088 on top to perform some acceleration on PGA.

Originally Posted by hammer

Nintendo usually has several strong 1st party exclusive games and low entry costs, hence breaking legacy compatibility is not a major issue.

WTF? How this should bother any Amiga owner/user?

Originally Posted by hammer

December 1993, Doom doesn't have 320x480 resolution.

Do simple math Sherlock - DoubleScan is performed by VGA so physically you have 400/480 lines not 200/240 lines - it doesn't mean that they are software accessible (but you can create such video mode on VGA so there is nothing against to get Doom in 400/480 lines - in fact i don't understand why it was never possible in modernized Doom).

Please focus on main topic, please don't quote irrelevant information - at least try...

[Promilus]

@Pandy71 - I honestly don't know whether should I be impressed or scared that you still have energy to discuss with that troll which uses OT when it suits him and bashes you every time you try to sneak a fun fact semi-related to the discussion. To this date he believes to some magic SIMD in Commodore PA-RISC despite the fact that every SFU related to pixels manipulation is SIMD there and it is already documented (and no, can't be used for general computing). And he still believes Hepler did PA-RISC implementation which works great WITHOUT big cache... so unlike EVERY HP PA-RISC (And yet Hepler in his next workplace never did anything quite like what he promised for Hombre).
Hammer is just someone without solid technical background which was fed with some scraps of information he couldn't understand and believed it to the point of blind fanaticism. You can't reason with him.

AMD products was almost never used in Amiga

Except all those MACH chips in a variety of bus adapters, ata controllers and turbo cards. Mediator is built around AMD MACH chips (and several of them) and so is FastATA.

[pandy71]

Originally Posted by Promilus

@Pandy71 - I honestly don't know whether should I be impressed or scared that you still have energy to discuss.

What i can say - no clue - but my memory strongly reject to accept such BS so... trying to get some things made clear.

Originally Posted by Promilus

Except all those MACH chips in a variety of bus adapters, ata controllers and turbo cards. Mediator is built around AMD MACH chips (and several of them) and so is FastATA.

Doesn't count - CPLD/FPGA can be selected freely and you can replace AMD by Xilinx, Lattice or Altera - in opposite it will be difficult to replace Am85C60 by any other solution - you need significantly alter HW and of course design new software.

[hammer]

Originally Posted by pandy71

Well, this is bit contradictory if consider for example NASA applications for Amiga. But I pointed only that for Motorola embedded market was important at least same as general computing if not more (simply for embedded market you need way more CPU's than for desktop).
.

Gibberish bullshit and it's additionally completely irrelevant to this discussion, topic, and general Amiga.

NASA's use case will not sustain Commodore's core revenue revenue stream.

Originally Posted by pandy71

Gibberish BS - additionally completely irrelevant to this discussion, topic and general Amiga - Boards based on Sitara (TI) are way better suited to Amiga needs than Broadcom SoC
.

Look in the mirror with your embedded market bullshit.

Fact: Commdore's Amiga is a desktop computer.

Originally Posted by pandy71

Unix on PC was very niche market from business perspective - it was also not popular in academic and seem also no one tried to create workstations around Intel CPU's (contrary to Motorola CPU's) - feel free to think why Intel was never chosen as good CPU for Unix - i can name at least few reasonable arguments for this.
.

The best-selling AT&T license Unix is Xenix.

https://en.wikipedia.org/wiki/Xenix

In the mid-to-late 1980s, Xenix was the most common Unix variant, measured according to the number of machines on which it was installed. Microsoft chairman Bill Gates said at Unix Expo in 1996 that, for a long time, Microsoft had the highest-volume AT&T Unix license.


Intel's integrated MMU baseline for 32-bit 386 allowed memory-protected operating system development and large-scale deployment.

Xenix is a niche relative to the very large MS-DOS installations just as 1992-1993 gaming PCs are a minority but still larger than the entire Amiga install base!

Originally Posted by pandy71

And i agree - 286 was licensed by other companies as second source...

And Windows no need MMU till 3.0 version - 286 provided simply possibility to address more memory not process/resources protection.

Windows 2.x 386 requires a 386 CPU. Windows 2.x 386 could take advantage of the Virtual 8086 mode of the 386 processor.

Windows 3.0 was released in 1990. The development work for Windows 3.0 started before 1990.

Originally Posted by pandy71

Well - many companies historically debunked your argument by real products - guess who is more right on this - many companies was founded around Motorola 68k Unix products, for 68000 Motorola offered 68451MMU so it is also argument to show that you are wrong (and some companies dissatisfied by 68451 added 1 cycle latency created own version of MMU or take another approach... anyway 68010 and 68012 shows that Motorola see Unix as important).

https://en.wikipedia.org/wiki/Xenix

In the mid-to-late 1980s, Xenix was the most common Unix variant, measured according to the number of machines on which it was installed. Microsoft chairman Bill Gates said at Unix Expo in 1996 that, for a long time, Microsoft had the highest-volume AT&T Unix license.

You can't handle the truth.

https://en.wikipedia.org/wiki/Motorola_68010
For Unix 68k use cases before 68020, custom MMU was employed.

Custom MMU wreaks mass deployment and makes Unix variants specific vendor-dependent. This caused 68K workstation market fragmentation.

Originally Posted by pandy71

I don't care about Microsoft, Xenix etc - this is gibberish BS, completely irrelevant from this topic perspective, stop spamming thread with such garbage.

Look in the mirror with your embedded market bullshit.

Originally Posted by pandy71

If my head is in the sand then yours seem to be deeply in the dingo dung.

Your embedded market argument is irreverent for desktop platforms such as Amiga, Apple, Sharp 68K, etc'

Originally Posted by pandy71

This shows that APPLE was important customer for Motorola and that APPLE competed with PC's - definitely there is no Commodore there.

You missed Motorola's marketing FUD against Intel. You're a POS.

Again, https://techmonitor.ai/technology/mo...0_next_quarter

Date: April 19, 1994.
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.

With 68060's 1994 release, Motorola Inc. made negative remarks against Intel Pentium competition.

Commodore's Akiko C2P existence shows Commodore is aware of the PC's chunky pixel format advantage and it was too late.


https://www.nytimes.com/1989/03/28/b...ith-intel.html

Motorola Inc. today announced key features of its next-generation microprocessor, the 68040, opening a new round in its battle with the Intel Corporation, which is set to introduce its newest chip on April 10.

Motorola said its 68040, the next generation in its 68000 family, would contain 1.2 million transistors and run much faster than its existing 68030. The new Intel 80486 is expected to be at least twice as fast as the 80386 used in high-end personal computers and is also expected to contain more than a million transistors.

Motorola and Intel have been fierce rivals in the microprocessor business for years. Intel chips, the most powerful of which is the 80386, are used in personal computers made by the International Business Machines Corporation and manufacturers of compatible machines. Motorola's microprocessors are used in the Macintosh computer made by Apple Computer Inc. and in work stations made by Sun Microsystems Inc. and other vendors.

While both companies seem to have well-established customer bases, they are still competing for the business of new companies that enter the personal computer business.



https://techmonitor.ai/technology/mo...versial_claims
MOTOROLA HEATS UP MIPS BATTLE WITH 50MHZ VERSION OF 68030, CONTROVERSIAL CLAIMS

Motorola Inc, clearly making sure that Intel Corp has plenty to think about as it prepares to launch the 80486, and making it clear that despite the badmouthing it received early on, there is plenty more mileage yet in the 68030, this week, as reported briefly (CI No 1,151) came out with a version clocked at a blinding 50MHz – 17MHz faster than its previous fastest. The company claims that the speed means that the thing does 12 MIPS, double that of all conventional processors available today. The 50MHz 68030 is fabricated in 1 micron HCMOS, the first conventional processor to be produced below 1.2 microns – but NEC Corp will have something to say about boths claims: see below. The company reminds us that Apollo Computer, Apple Computer, Hewlett-Packard, NEC, NeXT, Sony Microsystems and Sun Microsystems all use the 68030. The part starts sampling next month at $650 a time, with volume production planned for the third quarter.

[hammer]

Originally Posted by Promilus

@Pandy71 - I honestly don't know whether should I be impressed or scared that you still have energy to discuss with that troll which uses OT when it suits him and bashes you every time you try to sneak a fun fact semi-related to the discussion.

To this date he believes to some magic SIMD in Commodore PA-RISC despite the fact that every SFU related to pixels manipulation is SIMD there and it is already documented (and no, can't be used for general computing). And he still believes Hepler did PA-RISC implementation which works great WITHOUT big cache... so unlike EVERY HP PA-RISC (And yet Hepler in his next workplace never did anything quite like what he promised for Hombre).
Hammer is just someone without solid technical background which was fed with some scraps of information he couldn't understand and believed it to the point of blind fanaticism. You can't reason with him.

You didn't follow this topic's What IF subject. Commodore's discrete PA-RISC CPU solution is Hitachi's PA-RISC implementation and you didn't read the source documents.

Hitachi's PA-RISC implementation has a reasonable-size cache and that's your mistake. Commodore selected the PA-RISC instruction set due to good code density among the RISC instruction set options and it's related to the CPU cache.

Commodore's PA-RISC L1 cache implementation (ref 1)
1. The data cache has 1 KB with a 2 KB goal.
2. The instruction cache has 2 KB with a 4 KB goal.

Hitachi PA/50 has 8 KB L1 instruction cache and 4 KB L1 data cache. PA/50's 1.2 transistors is close to Commodore's 1 million transistor budget.
Hitachi HARP-1 has 8 KB L1 instruction cache and 16 KB L1 data cache. The HARP-1E supposedly included pseudo-vector processing modifications used in Hitachi vector supercomputers.

PowerPC 602 (3DO M2) competition has a 4KB cache. https://www.cpushack.com/CIC/embed/a...owerPC602.html

R3000 MIPS (Sony's PlayStation 1) competition has a 4KB instruction cache. 1KB data cache was repurposed for scratchpad. https://www.copetti.org/writings/consoles/playstation/

Commodore's core revenue stream is from game-centric Amiga SKUs, hence certain CPU instruction needs to be geared towards multimedia games.

Commodore can't survive with just big box A2000/A3000/A4000 Video Toaster sales.

3DFX Voodoo is not "general computing" since its purpose is for certain triangle 3D-based workloads.

Not every MMX instruction is used for integer 3D-related workloads.

Your comments about me are bullshit.


Reference
1. https://ia802805.us.archive.org/Book...ale=4&rotate=0

[pandy71]

Originally Posted by hammer

Gibberish bullshit and it's additionally completely irrelevant to this discussion, topic, and general Amiga.

Obviously you don't understand simple numbers - Motorola sold in 1 year more 68k for embedded market than Commodore aggregate number of Amiga ever sold... This is millions CPU's sold per year.

Originally Posted by hammer

NASA's use case will not sustain Commodore's core revenue revenue stream.

True, Commodore never use this fact in their marketing - why - no clue but i can only guess.

Originally Posted by hammer

Look in the mirror with your embedded market bullshit.

Yep, search for fresh dingo dung to hide your head...
Embedded means huge money for Motorola, general computing is less money than embedded...

Originally Posted by hammer

Fact: Commdore's Amiga is a desktop computer.

Wow... you are true mastermind, thx for shedding light on this interesting topic.

Originally Posted by hammer

The best-selling AT&T license Unix is Xenix.

https://en.wikipedia.org/wiki/Xenix

WTF? Who cares about Xenix and your obsession with Microsoft?

Originally Posted by hammer

Windows 2.x 386 requires a 386 CPU. Windows 2.x 386 could take advantage of the Virtual 8086 mode of the 386 processor.

Windows 3.0 was released in 1990. The development work for Windows 3.0 started before 1990.

Blah blah blah - we talking about MMU in 286 and how accordingly to you it was important for end users.

We NOT talking about Virtual 8086 mode in 386 which is NOT MMU

Originally Posted by hammer

You can't handle the truth.

What truth? Truth about?

Originally Posted by hammer

https://en.wikipedia.org/wiki/Motorola_68010
For Unix 68k use cases before 68020, custom MMU was employed.

Custom MMU wreaks mass deployment and makes Unix variants specific vendor-dependent. This caused 68K workstation market fragmentation.

WTF? 68020 has no embedded MMU and it will also require external MMU.

FYI Motorola offered for MMU's:68451 and 68851. Some vendors implemented custom MMU's but this was vendor choice.

Originally Posted by hammer

Look in the mirror with your embedded market bullshit.

Your embedded market argument is irreverent for desktop platforms such as Amiga, Apple, Sharp 68K, etc'

Obviously you don't understand that Motorola sold more embedded CPU's in 1..3 year period than aggregated number of all computers made by Commodore, Apple, Atari, Sharp etc ever.
This is embedded market power.

Originally Posted by hammer

You missed Motorola's marketing FUD against Intel. You're a POS.

Again, https://techmonitor.ai/technology/mo...0_next_quarter

Date: April 19, 1994.
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.

With 68060's 1994 release, Motorola Inc. made negative remarks against Intel Pentium competition.

Obviously you are confused...

Originally Posted by hammer

Commodore's Akiko C2P existence shows Commodore is aware of the PC's chunky pixel format advantage and it was too late.

For sure Commodore was aware as they offered PC's but chunky mode was implemented very first time in VGA.
And it was rather quick workaround to address software developers requests - AGA could introduce native chunky but Commodore preferred to not, instead they introduced Akkiko on single platform...

Originally Posted by hammer

https://www.nytimes.com/1989/03/28/b...ith-intel.html

Motorola Inc. today announced key features of its next-generation microprocessor, the 68040, opening a new round in its battle with the Intel Corporation, which is set to introduce its newest chip on April 10.

Motorola said its 68040, the next generation in its 68000 family, would contain 1.2 million transistors and run much faster than its existing 68030. The new Intel 80486 is expected to be at least twice as fast as the 80386 used in high-end personal computers and is also expected to contain more than a million transistors.

Motorola and Intel have been fierce rivals in the microprocessor business for years. Intel chips, the most powerful of which is the 80386, are used in personal computers made by the International Business Machines Corporation and manufacturers of compatible machines. Motorola's microprocessors are used in the Macintosh computer made by Apple Computer Inc. and in work stations made by Sun Microsystems Inc. and other vendors.

While both companies seem to have well-established customer bases, they are still competing for the business of new companies that enter the personal computer business.



https://techmonitor.ai/technology/mo...versial_claims
MOTOROLA HEATS UP MIPS BATTLE WITH 50MHZ VERSION OF 68030, CONTROVERSIAL CLAIMS

Motorola Inc, clearly making sure that Intel Corp has plenty to think about as it prepares to launch the 80486, and making it clear that despite the badmouthing it received early on, there is plenty more mileage yet in the 68030, this week, as reported briefly (CI No 1,151) came out with a version clocked at a blinding 50MHz – 17MHz faster than its previous fastest. The company claims that the speed means that the thing does 12 MIPS, double that of all conventional processors available today. The 50MHz 68030 is fabricated in 1 micron HCMOS, the first conventional processor to be produced below 1.2 microns – but NEC Corp will have something to say about boths claims: see below. The company reminds us that Apollo Computer, Apple Computer, Hewlett-Packard, NEC, NeXT, Sony Microsystems and Sun Microsystems all use the 68030. The part starts sampling next month at $650 a time, with volume production planned for the third quarter.

Yep, mass media and their habit to heat up atmosphere - good for housewifes and used cars sellers...

Btw - CF clocked at 33MHz offered over 13.5MIPS - 68030 clocked at 50MHz only 12MIPS - that's why CF costing fraction of 68k price and offering higher computational power than plain 68k was good product for embedded market.
Btw 68030 was never 486 competitor - 68040 was - and MMU was not necessary - same as FPU - that's why you have LC version.
And competition doesn't mean [ Show youtube player ] .

[Promilus]

Originally Posted by hammer

You didn't follow this topic's What IF subject. Commodore's discrete PA-RISC CPU solution is Hitachi's PA-RISC implementation and you didn't read the source documents.

I did and I understood it properly, you did not.
Hitachi PA/50 is microcontroller. Those weren't really meant to be powerful. Graphics controller should be. PA/50 has 16MIPS @20MHz, it all shows how ridiculous amount of cache Commodore tried to put inside their implementation. You are the only one who fails to understand that. PowerPC ain't PA-RISC and neither is MIPS. Why are you trying to force the issue with examples of different architectures requiring different amount of cache to work well? It only shows how limited mentally are you. All fairly powerful PA-RISC had either very large L1 or relatively small L1 but gigantic L2. According to released documentation chip to be used inside Hombre had neither. Let it finally sink in...

Not every MMX instruction is used for integer 3D-related workloads.

Sure, but none of Commodore's SFU did anything relevant to other workloads and packed pixel data type was hardly suitable for anything else.

Your comments about me are bullshit.

And most of your comments about anything are pure crap. You're just silly d2d salesman trying to sell us some shit hoping we'd know nothing about it. You're the one who knows nothing about it and you're typing your spam on a forum full of ppl which knows their shit. So screw you!

[hammer]

Originally Posted by pandy71

Obviously you don't understand simple numbers - Motorola sold in 1 year more 68k for embedded market than Commodore aggregate number of Amiga ever sold... This is millions CPU's sold per year.

1. That's a useless argument when you haven't separated individual 68K models' market share.

http://archive.computerhistory.org/r...-05-01-acc.pdf
Page 86 of 417, DataQuest 1995

1994 Worldwide Microprocessor Market Share Ranking.

For 1994 Market Share
1. Intel, 73.2%
2. AMD, 8.6%
3. Motorola, 5.2%
4. IBM, 2.2%

Page 84 of 417,

Supply Base for 32-Bit Microprocessors—1994,
For Product's Share of Total 32-Bit-and-Up MPU Market 1994

68000, 17%

80386SX/SL, 3%

80386DX, 3%

80486SX, 16%

80486DX, 21%

683XX, 9%

68040, 3%

68030, 1%

68020, 3%

80960, 4%

AM29000, 1%

32X32, 3%

R3000/R4000, 1%

Sparc, 1%

Pentium, 4%

Others, 10%

Motorola wasn't able to convert 68000's success for 68020, 68030 and 68040.


683XX has 68EC000 and CPU32 i.e. similar to the 68020 without bitfield instructions and with a few instructions unique to the CPU32 core. Motorola kitbashing 68K instruction set as product segmentation i.e. Jack Tramiel's Plus 4 (no hardware scroll and sprites) vs C64/ ST vs Mega ST (with custom blitter) mentality. 683XX has 25 to 33 Mhz clock speed. 683XX is a useless main CPU for the Amiga.

You don't know shit.

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

2. Other 68K based platforms do NOT run Amiga software, hence your argument is useless. The X86 camp is largely unified under the X86 PC clone market.

The trinity AMD/Intel/MS governs the PC clones ecosystem and provides reference designs with IBM PC-compatible standards. These days, AMD/Intel provides "Designed for Windows" UEFI/ACPI PC reference designs i.e. go ask Framework Computer.

Motorola's exit from the semiconductor market is real i.e. Motorola divested its semiconductor division as Freescale.

[pandy71]

Originally Posted by hammer

1. That's a useless argument when you haven't separated individual 68K models' market share.

http://archive.computerhistory.org/r...-05-01-acc.pdf
Page 86 of 417, DataQuest 1995

1994 Worldwide Microprocessor Market Share Ranking.


Motorola wasn't able to convert 68000's success for 68020, 68030 and 68040.

Now you proved that you don't understand numbers - since beginning this weird discussion i'm trying to say that Motorola not competed with Intel on general computer market where Intel was most popular due IBM PC architectural domination.
But opposite to your claims Motorola 68k was very popular architecture in various embedded use cases - mostly industrial automation, military ,automotive, networking etc. There was many reasons for this but one of the most important factor was simply neat architecture - many programmers even today say that this is one of most pleasant ISA'a.
So you are true when you wrote that Motorola didn't get market domination in general computing when compared to Intel x86 but it doesn't mean that Motorola doesn't have own market where 68k family was more popular than x86. Those markets provided Motorola more incomes than aggregated general computing.

Originally Posted by hammer

683XX has 68EC000 and CPU32 i.e. similar to the 68020 without bitfield instructions and with a few instructions unique to the CPU32 core. Motorola kitbashing 68K instruction set as product segmentation i.e. Jack Tramiel's Plus 4 (no hardware scroll and sprites) vs C64/ ST vs Mega ST (with custom blitter) mentality. 683XX has 25 to 33 Mhz clock speed. 683XX is a useless main CPU for the Amiga.

No clue what are talking about - seem you are creatin
g problem then you bravely arguing and fighting with this artificial problem... good for you - i can say only good luck...

Originally Posted by hammer

You don't know shit.

True, as i've wrote earlier - i have no clue (opposite to you) with dingo dung - good luck...

Originally Posted by hammer

Motorola's exit from the semiconductor market is real i.e. Motorola divested its semiconductor division as Freescale.

True - semiconductor division in Motorola was split to: Freescale later acquired by Philips and combined to form NXP and second On Semi (mostly discrete electronics - transistors, diodes etc).
But this process was common for many big semiconductor brands.

[hammer]

Originally Posted by pandy71

Now you proved that you don't understand numbers - since beginning this weird discussion i'm trying to say that Motorola not competed with Intel on general computer market where Intel was most popular due IBM PC architectural domination.

FALSE.

Motorola tried to compete against Intel with price matching their 68030-25 against Intel's 386DX-25 and releasing 68040 against 80486.

Motorola's 68030-25 didn't factor in AMD 386DX-40's price.

Again, Motorola didn't convert 68000's success for 68020, 68030 and 68040.

Originally Posted by pandy71

But opposite to your claims Motorola 68k was very popular architecture in various embedded use cases - mostly industrial automation, military ,automotive, networking etc. There was many reasons for this but one of the most important factor was simply neat architecture - many programmers even today say that this is one of most pleasant ISA'a.

Your argument doesn't negate DataQuest 1995's worldwide CPU market share research.

http://archive.computerhistory.org/r...-05-01-acc.pdf
Page 86 of 417, DataQuest 1995

1994 Worldwide Microprocessor Market Share Ranking.

For 1994 Market Share
1. Intel, 73.2%
2. AMD, 8.6%
3. Motorola, 5.2%
4. IBM, 2.2%

In 1994, Motorola's market share was already smaller than AMD's, let alone Intel's.

Originally Posted by pandy71

So you are true when you wrote that Motorola didn't get market domination in general computing when compared to Intel x86 but it doesn't mean that Motorola doesn't have own market where 68k family was more popular than x86. Those markets provided Motorola more incomes than aggregated general computing.

1. My cited DataQuest 1995's worldwide CPU market share research was for 32-bit CPU market share that included Motorola's 68000. Many 68K consumer and workstation platforms were dropping like flies during the 1990s.

68020 and 68LC040 weren't 100 percent drop-in 68000 replacements due to the instruction set incompatibility. When Motorola didn't respect 100 percent 68000 backward compatibility, they couldn't convert 68000's success for 68020, 68030 and 68040.

2. Motorola tried to compete against Intel with price matching their 68030-25 against Intel's 386DX-25

https://archive.computerhistory.org/...-05-01-acc.pdf
Page 119 of 981

For 1992
68000-12 = $5.5
68EC020-16 PQFP = $16.06,
68EC020-25 PQFP = $19.99,

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


For 68030-25, Motorola nearly price matched Intel's 386DX-25. Motorola didn't factor in AMD's 386-40 clone. "Always two, there are. No more. No less. A Master and an apprentice.".

https://websrv.cecs.uci.edu/~papers/...LES/080502.pdf
The 68000 line has had an intense rivalry with Intel’s x86 ever since IBM spurned Motorola for its first PC. Each processor generation has been lined up against its rival in performance. The 68040 delivers performance similar to the 486 running at the same clock speed. The weakness of the ’040 is that Motorola struggled to ship 33-MHz parts even as Intel was boosting the 486 to 50 Mhz and beyond

Down The Garden Path

Although Motorola has yet to begin shipments of the 50-MHz 68060, the company is already claiming that the chip will reach 66 MHz by the end of this year, just one quarter after 50-MHz shipments begin. Unfortunately, the company has a poor track record of meeting such claims, as shown by this brief history of the 68040.

4/89—68040 preannounced two weeks before Intel announces 486. Formal ’040 introduction set for 3Q89.

10/89—Company discloses ’040 architecture.

12/89—Motorola admits “No ’040 in ’89.”

1/90—’040 formally announced, general sampling slated for end of 1Q90.

4/90—Company claims ’040 on schedule for volume production midyear. Sample date slips.

9/90—Motorola announces it has begun general sampling of ’040 with volume shipments to begin in October.

11/90—Company claims 1,000 ’040s shipped this month and that volume production has begun.

1/91—HP begins shipping 25-MHz ’040 systems, promises 33 MHz in 2Q91. Benchmarks of new systems fall short of Motorola’s original claims.

6/91—Motorola says 33-MHz parts in September.

6/92—Volume shipments of 33-MHz parts begin. Motorola promises 40-MHz 68040 for September.

12/92—40-MHz parts begin to ship in volume.




You're wrong with Motorola not competing against Intel. Your embedded market defense is a cop-out.

[hammer]

Originally Posted by Promilus

I did and I understood it properly, you did not.
Hitachi PA/50 is microcontroller. Those weren't really meant to be powerful. Graphics controller should be. PA/50 has 16MIPS @20MHz, it all shows how ridiculous amount of cache Commodore tried to put inside their implementation.

https://www.openpa.net/pa-risc_processor_other.html
1993
Hitachi PA/50L has 33 Mhz with 12 KB cache.
Hitachi PA/50M has 60MHz with 12 KB cache.

1994
Hitachi HARP-1 has 150 Mhz with 24 KB and 1 MB L2 cache.

Originally Posted by Promilus

You are the only one who fails to understand that. PowerPC ain't PA-RISC and neither is MIPS. Why are you trying to force the issue with examples of different architectures requiring different amount of cache to work well? It only shows how limited mentally are you. All fairly powerful PA-RISC had either very large L1 or relatively small L1 but gigantic L2. According to released documentation chip to be used inside Hombre had neither. Let it finally sink in...

Factor in the code density difference between PA-RISC vs PowerPC. Let that finally sink in...

[hammer]

Originally Posted by Promilus

I did and I understood it properly, you did not.
Hitachi PA/50 is microcontroller. Those weren't really meant to be powerful. Graphics controller should be. PA/50 has 16MIPS @20MHz, it all shows how ridiculous amount of cache Commodore tried to put inside their implementation.

https://www.openpa.net/pa-risc_processor_other.html
1993
Hitachi PA/50L has 33 Mhz with 12 KB cache.
Hitachi PA/50M has 60MHz with 12 KB cache.

1994
Hitachi HARP-1 has 150 Mhz with 24 KB and 1 MB L2 cache.

Originally Posted by Promilus

You are the only one who fails to understand that. PowerPC ain't PA-RISC and neither is MIPS. Why are you trying to force the issue with examples of different architectures requiring different amount of cache to work well? It only shows how limited mentally are you. All fairly powerful PA-RISC had either very large L1 or relatively small L1 but gigantic L2. According to released documentation chip to be used inside Hombre had neither. Let it finally sink in...

Reminder, PowerPC 602 (3DO M2) competition has a 4KB cache.
https://www.cpushack.com/CIC/embed/a...owerPC602.html

Factor in the code density difference between PA-RISC vs PowerPC. Let that finally sink in...

[hammer]

Originally Posted by pandy71

Obviously you don't understand simple numbers - Motorola sold in 1 year more 68k for embedded market than Commodore aggregate number of Amiga ever sold... This is millions CPU's sold per year.

Your "millions CPU's sold per year" argument is crushed by Data Quest 1995 market share research.

Data Quest 1995 market research made the situation clear on the 32-bit CPU market share for 1994. This includes the market share breakdown for each 68K model.

Your argument is misinformation.

Originally Posted by pandy71

Yep, search for fresh dingo dung to hide your head...
Embedded means huge money for Motorola, general computing is less money than embedded...

A typical Motorola drone's embedded market defense while forgotten Motorola's attempt to compete against Intel with price matching e.g. 386DX-25 vs 68030-25 in 1992.

Motorola's 32-bit CPU market share was beaten by AMD in 1994.

Originally Posted by pandy71

Blah blah blah - we talking about MMU in 286 and how accordingly to you it was important for end users.

Important enough for Xenix 286 and OS/2 286.

Originally Posted by pandy71

We NOT talking about Virtual 8086 mode in 386 which is NOT MMU

VM86 uses MMU, idiot.


"The Memory Management Unit (MMU) provides the support for both the segmentation of main memory for both protected mode and real mode"

WTF? 68020 has no embedded MMU and it will also require external MMU.

An extra cost factor.

FYI Motorola offered for MMU's:68451 and 68851. Some vendors implemented custom MMU's but this was vendor choice.

An extra cost factor.

Obviously you don't understand that Motorola sold more embedded CPU's in 1..3 year period than aggregated number of all computers made by Commodore, Apple, Atari, Sharp etc ever.
This is embedded market power.

For 1994, your argument is misinformation when Motorola's 32-bit CPU market share is below AMD's.


Data Quest 1995 market research made the situation clear on the 32-bit CPU market share for 1994. This includes the market share breakdown for each 68K model.

Obviously you are confused...

You can't handle the truth when Motorola makes competitive remarks against Intel.

For sure Commodore was aware as they offered PC's but chunky mode was implemented very first time in VGA.

Nope. With EGA cards, a programmer's program can write pixels to video memory in a chunky way using Write Mode 2, but it is slow.

Faster clones have these functions to be faster than IBM's EGA.

And it was rather quick workaround to address software developers requests - AGA could introduce native chunky but Commodore preferred to not, instead they introduced Akkiko on single platform...

AAA had the chunky pixel mode, but Commodore was shooting for the moon i.e. 24-bit display capability with 24-bit color palette.

https://en.wikipedia.org/wiki/Advanc...ecture_chipset
Direct Chunky 16-bit pixels

24-bit hybrid mode (with chunky/planar properties) consisted of 3 byte-planes of 8-bit chunks each.

New packed (compressed) pixels (2-bit PACKLUT and 4-bit PACKHY) decompressed by Linda to 8-bit half-chunky or 24-bit Hybrid pixels respectively, used for speeding up animations.

New Hold-and-Modify modes (HAM-8 chunky and HAM-10 for 24bit / 16.8 million colors).

Sprites size can go up to 128 pixels in width at any height.

Dual 8-bit playfields.

12× to 20× memory bandwidth of Chip RAM access of ECS.

8× Blitter speed increase of AGA/ECS blitter.

64-bit pixel bus with 114 MHz pixel clock in dual systems which makes 1280×1024 @ 72Hz screens possible.

--------------
From 1989, only 1 year of serious R&D was committed to AAA. A lesson for not kicking out the original Amiga team.

[pandy71]

Originally Posted by hammer

Your "millions CPU's sold per year" argument is crushed by Data Quest 1995 market share research.

Data Quest 1995 market research made the situation clear on the 32-bit CPU market share for 1994. This includes the market share breakdown for each 68K model.

Your argument is misinformation.


A typical Motorola drone's embedded market defense while forgotten Motorola's attempt to compete against Intel with price matching e.g. 386DX-25 vs 68030-25 in 1992.

Motorola's 32-bit CPU market share was beaten by AMD in 1994.

How conveniently you "forget" about Motorola departing from own designs since at least 1991 when AIM alliance was formally established.

Originally Posted by hammer

Important enough for Xenix 286 and OS/2 286.

Guess how this is important for Amiga - ok, can't demand too much from you but at least try to guess (you can use coin as help).

Originally Posted by hammer

VM86 uses MMU, idiot.

Yep but to simulate 86 in 386 - virtual 86 has no direct benefits from MMU itself moron.

Originally Posted by hammer

"The Memory Management Unit (MMU) provides the support for both the segmentation of main memory for both protected mode and real mode"

obviously they use same HW to cover more functionality - right choice but 8086 has no benefits form MMU as it is not present in original 8086.

Originally Posted by hammer

An extra cost factor.
An extra cost factor.

Perhaps justified by usage case?

Originally Posted by hammer

For 1994, your argument is misinformation when Motorola's 32-bit CPU market share is below AMD's.
Data Quest 1995 market research made the situation clear on the 32-bit CPU market share for 1994. This includes the market share breakdown for each 68K model.
You can't handle the truth when Motorola makes competitive remarks against Intel.

You can't handle fact that Motorola since at least 1991 give-up with 68k future development choosing Power PC than own designs.

Originally Posted by hammer

Nope. With EGA cards, a programmer's program can write pixels to video memory in a chunky way using Write Mode 2, but it is slow.

Faster clones have these functions to be faster than IBM's EGA.

Hope you aware that EGA can provide maximum 16 colors i.e. 4 bit per pixel and 86 has limit of 64KiB per page so using chunky means that they need to use two 64KiB pages - that's why planar was selected over chunky.
Write mode register only help to alter planar data more efficiently but using it is affected by I/O latency.
Famous Mode X was planar not chunky.

Originally Posted by hammer

AAA had the chunky pixel mode, but Commodore was shooting for the moon i.e. 24-bit display capability with 24-bit color palette.
.
.
.
From 1989, only 1 year of serious R&D was committed to AAA. A lesson for not kicking out the original Amiga team.

AAA was crappy design - from Amiga future perspective it means high cost, practically no benefits i.e. dead end.