What spec would the A1200 had launched with if it was up to YOU

[pandy71]

Originally Posted by Thomas Richter

That rendering is slow on the A3000 is not because the code runs from chip. It does not. It is slow because display DMA on ECS occupies almost all cycles. 16 colors for the web is also quite limiting, but this is already the maximum. That the frame buffer for VGA is linear and mapped in total into the CPU address space is a (relatively) recent addition. However, unlike Amiga, the PC architecture evolved, and chip vendors added linear addressing. If PC vendors would have followed the same stupid management principles of "read my lips - no new chips", then we still had four planar VGA planes and segmented memory today. Thus, the PC architecture was surely awkward, but unlike CBM, the open market and open architecture helped to evolve it into something useful.

Once again try to run code from space 0x000A0000 - 0x000BFFFF in PC and compare speed.
It is obvious that video bandwidth requirements limiting memory bandwidth - this could be workarounded but still limitations will apply - solution is local graphic memory not shared with CPU.

I hear you but still think that we need compare apples with apples.
CBM could create legacy compatible single chip with Agnus, Denise, Paula redirect video to combine it with video from some RTG board or even using sufficiently fast interface (PCI?) blit (as overlay) in HW video buffer so virtualize graphic output from legacy Amiga into frame buffer in RTG card.
This will solve most if not all problems with custom chipset and provide Amiga possibility to reuse PC solutions.

[Thomas Richter]

Originally Posted by pandy71

Once again try to run code from space 0x000A0000 - 0x000BFFFF in PC and compare speed.

Once again, this is irrelevant. An A3000 does not run code from ChipMem, neither does the PC.

Originally Posted by pandy71

I hear you but still think that we need compare apples with apples.

Then please do.

Originally Posted by pandy71

CBM could create legacy compatible single chip with Agnus, Denise, Paula redirect video to combine it with video from some RTG board or even using sufficiently fast interface (PCI?) blit (as overlay) in HW video buffer so virtualize graphic output from legacy Amiga into frame buffer in RTG card.

What they /could/ is not quite relevant, isn't it? What they *did* is relevant, and management blocked such plans in early stages and changed plans too late.

[Gorf]

Originally Posted by Thomas Richter

What they *did* is relevant, and management blocked such plans in early stages and changed plans too late.

Well ... I guess in this thread I is actually "could" and "would" ... around 91/92 to be specific.

[pandy71]

Originally Posted by Thomas Richter

Once again, this is irrelevant. An A3000 does not run code from ChipMem, neither does the PC.

Well... Updating video RAM content on PC without interfering with video refresh is slow similarly to Amiga.

Originally Posted by Thomas Richter

Then please do.

Trying.

Originally Posted by Thomas Richter

What they /could/ is not quite relevant, isn't it? What they *did* is relevant, and management blocked such plans in early stages and changed plans too late.

Whole thread is irrelevant from this perspective but despite this there is many participants in discussion. I've heard once CTO in company i was worked - guy was skilled, smart etc - he was fully aware that product have poor quality, software is crap (his own words "spaghetti code") yet he push to deploy product - why? - because product was already advertised - after all customers was seriously disappointed, government authorities involved etc and guess what - he was promoted to become CEO.
This is standard in management - only some companies are too small to survive and CBM was too small to survive shitty management.
On PC market there was many CBM's - they not survived but market on overall was way stronger and PC survived - this ends everything.

[Bruce Abbott]

Originally Posted by Thomas Richter

That the frame buffer for VGA is linear and mapped in total into the CPU address space is a (relatively) recent addition. However, unlike Amiga, the PC architecture evolved, and chip vendors added linear addressing.

The Amiga's architecture evolved too, but it didn't need to evolve as much because it started from a better place. The Amiga always had linear addressing. No stupid 640k RAM limit with tiny banked areas above it for video and memory expansion. No brain-dead A20 gate mechanism to switch from 286 'protected' to 'real' mode. No 'DOS extender' needed to support games using 32 bit memory.

The PC architecture evolved because it needed to, but it was done in a cack-handed way. It wasn't anything to be proud of. And they were still carrying that baggage around well after Commodore folded.

The PC industry rejected attempts to shake off that crap and truly evolve. They rejected IBM's 1987 Micro Channel architecture, which had 32 bit operation at 10MHz and automatic configuration. The industry's first attempt at a standard 32 bit bus, EISA, introduced in 1988, didn't fair well either (only being used in servers and high-end workstations). It wasn't until the VESA Local bus was introduced in 1992 that a 32 bit bus became popular on mainstream PCs, and mostly only on 486s. VL bus was designed as a low cost stopgap solution to the limited bandwidth of ISA. It simply extended the CPU's own bus signals to the connector with minimal intermediary logic.

So the statement 'unlike Amiga, the PC architecture evolved' is a lie. In fact the Amiga 'evolved' a 32 bit expansion bus 2 years before the PC industry as a whole did, with Zorro III in 1990. The A1200 had the Amiga equivalent of VL bus with its 150 pin 32 bit expansion bus in 1992, the same year that VLB was adopted on PCs.

If PC vendors would have followed the same stupid management principles of "read my lips - no new chips",

This quote from Dave Haynie regarding his disappointment with the A3000 ("It wasn't the machine we wanted. We asked for better graphics, we were told by management: "Read my lips - no new chips.") is regularly trotted out to prove that Commodore management was totally incompetent. But Brian Bagnall's book Commodore the Final Years tells a different story.

Finishing the A3000

Almost three years had passed since Commodore had released a new computer. With executives starting to wonder if their engineers were still capable of releasing new machines, it would be up to Hedley Davis and his team to finish... "I identified extremely strongly with the A3000. I worried about that machine in a way that my coworkers didn't really appreciate".

Unlike the C65, the A3000 had been remarkably well managed with realistic timelines and a minimum of feature creep. "This was push the product out the door and absolutely make it happen", says Davis. The first sample of A3000s had been sent to international countries in December 1989. The production schedule called for all casework, manuals, final software, and ROMs by March 15, 1990, and wide release planned for June...

The A3000 was a unique project at Commodore because it was a collaboration among many of its top engineers... "The A3000 was my favorite project because we had a real team", says Dave Haynie. "That made for a better result than those systems I did mostly alone".

Meanwhile, most PC manufacturers weren't designing any chips - they were just using what had been developed by others who specialized in chipsets for motherboards, video cards etc. The majority of vendors merely assembled machines using boards made by 3rd parties (I was one of them!).

Originally Posted by =Thomas Richter

...then we still had four planar VGA planes and segmented memory today. Thus, the PC architecture was surely awkward, but unlike CBM, the open market and open architecture helped to evolve it into something useful.

VGA wasn't 'open architecture'. It was reverse-engineered. The first outfit to do it was Chips and Technologies, who had previously reverse-engineered EGA.

The CS8240 chipset was C&T's crowning glory in late 1985. C&T openly stated that it reverse-engineered the [IBM] EGA by removing the silicon chips from their DIP packages and making photographic enlargements of the circuits, then working backwards from the photos to logic diagrams. C&T patterned its own copies with a few changes - all perfectly legal before the Silicon Copyright Act of 1984. One penalty for such slavish copying is that all of IBM's bugs came as part of the package.

These guys could have reverse-engineered the Amiga chipset if they wanted. But they didn't - I wonder why? Perhaps because when IBM released the PC it instantly became the industry standard and nobody was interested in anything else? Commodore could have truly opened the Amiga chipset design in 1987 and it wouldn't make any difference. The big money was in ripping off IBM, not helping Commodore become more successful (as if cloned Amigas would do that).

[Promilus]

That's fine BS here...
MCA - fee for IBM for both mobo vendors and card vendors... yeah. Cost of connector was twice as high as ISA iirc. Yeah. Also support for new cards had to be done through BIOS update. That's the reason PC world didn't really adopt that.

Amiga1200 32b VLB-style trapdoor connector - damn, you have 16bit ISA style connector on A500 as well. Why? Because both ISA and VLB, and trapdoor, and egde connector - those are basically CPU bus signals! But... only 68k inherent bus arbitration. The problem is... there's only one such slot on A500/1200 and there aren't all that many devices made on that in comparison to VLB or ISA.
Also Zorro III is basically EISA at that point.

VGA wasn't 'open architecture'. It was reverse-engineered

And with that particular quotation you see "open architecture" relates to PC not VGA...

Cloned amigas most likely would be great past 93, but the problem is ... Commodore was too weak to be a leader of standards for own computers. That's why they did fail with RTG and did offload it to 3rd parties. With basically everything. They didn't make single turbo card for Amiga home computers and the one for A3/4k were a bit quirky. And again 3rd parties did introduce own roms, fastkick etc. And many tools back then created some incompatibilities. Sure, in PC world it was the same but it lead to creation of associations working towards standards for all. Like VESA.

[Thomas Richter]

Originally Posted by pandy71

Well... Updating video RAM content on PC without interfering with video refresh is slow similarly to Amiga.

That is not quite the case. VGA chipsets typically have a write queue that accepts writes from the CPU, then release the bus and perform the write whenever there is time. VGA chipsets also have a higher bandwidth on their memory than what CBM was able to offer.

[Gorf]

Originally Posted by Bruce Abbott

Perhaps because when IBM released the PC it instantly became the industry standard and nobody was interested in anything else?

Except for all the customers of DEC, HP, SUN, SGI, Apple, Atari ... and even Commodore.

The PC was not an instant victory, but a slow march to power.
In the second half of the 80s the workstations market was growing faster than the PC market.

Survey: computer workstation market up 53 percent in 1988
...
Sales of workstations -- which are more powerful than personal computers but not as powerful as minicomputers -- rose to $4.1 billion in 1988 from $2.7 billion in 1987, according to San Jose-based Dataquest.

Sun Microsystems Inc. of Mountain View, Calif., retained its dominance over the industry, capturing 28.3 percent of the market with sales of $1.2 billion.

[AestheticDebris]

Originally Posted by Bruce Abbott

The PC architecture evolved because it needed to, but it was done in a cack-handed way. It wasn't anything to be proud of. And they were still carrying that baggage around well after Commodore folded.

The lesson from history on the PC architecture is that maintaining compatibility was far more important than architectural purity. Coming up with a nice "clean" design like MCA just wasn't as useful from a practical standpoint as coming up with something that "just works" with existing software.

If there had been a bit more of that attitude in Commodore, then maybe there could have been a migration to off the shelf VGA chips, even if they had to be hacked into the architecture somehow.

[grond]

Originally Posted by AestheticDebris

The lesson from history on the PC architecture is that maintaining compatibility was far more important than architectural purity. Coming up with a nice "clean" design like MCA just wasn't as useful from a practical standpoint as coming up with something that "just works" with existing software.

If there had been a bit more of that attitude in Commodore, then maybe there could have been a migration to off the shelf VGA chips, even if they had to be hacked into the architecture somehow.

The problem here is that the most valuable part of the Amiga software body was software accessing the hardware interface directly (games and demos) which made the step to something completely different while maintaining compatibility close to impossible.

I still believe that AGA could have been technically "good enough" until the Playstation hit the market, but for Amiga to survive as a platform it should have been clear at that point already that it was only used to buy time until the platform could move entirely over to standard PC hardware plus some legacy hardware add-on like some sort of AGA-on-a-plugin-card. The Amiga lacked the critical mass for such a transition as there was too little software out there that represented invested capital that ought not have been thrown out of the window when moving to a more modern hardware. It was just too easy to simply replace the Amiga altogether with either a more capable games or productivity machine.

[pandy71]

Originally Posted by grond

The problem here is that the most valuable part of the Amiga software body was software accessing the hardware interface directly (games and demos) which made the step to something completely different while maintaining compatibility close to impossible.

But providing legacy compatibility was relatively simple - same as it was done in PS2 and in PS3 - just provide dedicated legacy HW and combine legacy video/audio with new functionality - this will provide especially at the beginning access to legacy software with 100% compatibility.

Sony introduced HW compatibility with PS1 in PS2 and later PS2 compatibility in early versions of PS3.

[pandy71]

Originally Posted by Thomas Richter

That is not quite the case. VGA chipsets typically have a write queue that accepts writes from the CPU, then release the bus and perform the write whenever there is time. VGA chipsets also have a higher bandwidth on their memory than what CBM was able to offer.

True - so at a cost of latency CPU may have access to local video RAM - frequently mentioned ET4000 was in fact large FIFO/cache to deal efficiently with this - (IIRC similarly AAA chip Andrea should act as buffer to hide RAM bandwidth limitations).

But when you place code on VGA video RAM and try to run it or even use VGA RAM as RAM disk then you will face obvious bandwidth limitations - same as faced by Amiga - to compare apples with apples - in basic configuration Amiga is like PC with RAM located on VGA board.
This was not the case obviously in real life so 640KB in PC was "fast RAM" and PC "CHIP RAM" was rarely used to act in a different way - it was just frame buffer.

[pandy71]

Originally Posted by Promilus

That's fine BS here...
MCA - fee for IBM for both mobo vendors and card vendors... yeah. Cost of connector was twice as high as ISA iirc. Yeah. Also support for new cards had to be done through BIOS update. That's the reason PC world didn't really adopt that.

Amiga1200 32b VLB-style trapdoor connector - damn, you have 16bit ISA style connector on A500 as well. Why? Because both ISA and VLB, and trapdoor, and egde connector - those are basically CPU bus signals! But... only 68k inherent bus arbitration. The problem is... there's only one such slot on A500/1200 and there aren't all that many devices made on that in comparison to VLB or ISA.
Also Zorro III is basically EISA at that point.

Cost was higher by many reasons - VLB, EISA, PCI all was more expensive than ISA - EISA and PCI was covered also by royalties (and not so convenient as Zorro) - not sure if anyone added MCA or EISA card to system but it was pain - not possible without dedicated floppy disk - technically autoconfiguration was done at SETUP ("BIOS") level.

[grond]

Originally Posted by pandy71

But providing legacy compatibility was relatively simple - same as it was done in PS2 and in PS3 - just provide dedicated legacy HW and combine legacy video/audio with new functionality - this will provide especially at the beginning access to legacy software with 100% compatibility.

Yes, that is what I was referring to as a path towards the future after (a better version of) AGA. At the time actual AGA came out, it was certainly too expensive to put a whole alternate computer system (likely choice would have been OCS 0.5M+0.5M) into a new Amiga in parallel with the new gen hardware.

[Gorf]

Originally Posted by grond

Yes, that is what I was referring to as a path towards the future after (a better version of) AGA. At the time actual AGA came out, it was certainly too expensive to put a whole alternate computer system (likely choice would have been OCS 0.5M+0.5M) into a new Amiga in parallel with the new gen hardware.

Take VRAM as "normal" FastRAM and add a RAM-DAC with CLUT and you can use that portion of RAM as a dumb framebuffer ... no bus needed.
CPU has to do all the timings and blitting and so on..

Or make use of the UHRES features of AGNUS/ALICE - that would be VRAM as ChipRAM (+CLUT+DAC)
Slower because of the ChipRAM-bottleneck but otherwise "carefree" for the CPU.

Both solutions are comparably cheap.

(Legacy screen modes and new VRAM screen modes could be mixed together via simple digital genlock, as long as signals are synchronized -> effectively a 3rd playfield)

[grond]

Originally Posted by Gorf

(Legacy screen modes and new VRAM screen modes could be mixed together via simple digital genlock, as long as signals are synchronized -> effectively a 3rd playfield)

Then you'd be tying the new hardware again to 50Hz screens.

[Gorf]

Originally Posted by grond

Then you'd be tying the new hardware again to 50Hz screens.

Which might be quite useful in case of the CD32, or for games in general.
This would be an option and not tied in general.

[Thomas Richter]

Originally Posted by grond

The problem here is that the most valuable part of the Amiga software body was software accessing the hardware interface directly (games and demos) which made the step to something completely different while maintaining compatibility close to impossible.

Depending on your definition of "valuable", but unfortunately yes, the software basis was mostly that. Which brings me back to "wrong userbase" and "wrong marketing".

Originally Posted by grond

I still believe that AGA could have been technically "good enough" until the Playstation hit the market, but for Amiga to survive as a platform it should have been clear at that point already that it was only used to buy time until the platform could move entirely over to standard PC hardware plus some legacy hardware add-on like some sort of AGA-on-a-plugin-card.

There was no sufficiently valuable software left that made the Amiga hardware worth preserving, but that mistake was made long before AGA came to be. Games, unlike produtive software, fall out of fashion, and you'd rather buy a new hardware to play the latest game instead of preserving the old hardware to continue playing the old games.

[pandy71]

Originally Posted by Thomas Richter

Games, unlike produtive software, fall out of fashion, and you'd rather buy a new hardware to play the latest game instead of preserving the old hardware to continue playing the old games.

I disagree - now playing PS3 - skipped PS4, for a moment seriously considered PS5 but now i don't know - perhaps PS6? New games are better technically but from my perspective this progress is not enough.
So technically now i'm preserving old hardware to play old games - i have too many games - and guess what returning to many of them and still have fun.
Keeping legacy software compatibility is important (i would be happy to migrate my game library to modern console but this is not possible).

[PortuguesePilot]

Originally Posted by pandy71

I disagree - now playing PS3 - skipped PS4, for a moment seriously considered PS5 but now i don't know - perhaps PS6? New games are better technically but from my perspective this progress is not enough.
So technically now i'm preserving old hardware to play old games - i have too many games - and guess what returning to many of them and still have fun.
Keeping legacy software compatibility is important (i would be happy to migrate my game library to modern console but this is not possible).

I understand you and I also get a bigger and better kick with old games than most newer ones (games cater to dumb players so no real challenge, it's almost as if the games play themselves and all you have to do is watch it; you have to be permanently logged-in; you buy an incomplete beta and have to playtest the thing and then buy all the DLCs via micro-transactions; pay to win mentality; almost an infinite amount of time to upgrade; etc)...

But when it comes to old games, most people are going the "emulation" way instead of preserving the hardware. I preserve the hardware because of itself, not necessarily because of the software. Heck, I even own some system that I have no software for. And even I, who am a hardwarephiliac, still resort to emulation quite often (I have a PSP littered with emulators in my bedside table and another one on my most-used toilet) for my gaming needs. Yes, granted, I often boot my A1200 and play for a bit but mostly so that electricity flows through it and heat it up a bit.

A little more on-topic: when the Macs jumped from MacOS 8.x to MacOS X they broke retro-compatibility with the older software and the way they resolved that issue was through a very useful - if clunky - "software emulation" that solved most compatibility problems. Granted, most of the Mac software didn't "bang the metal" so the "emulation" route was simpler there than a putative use of the same strategy on the Amiga side of things, but the ReloKick programs and similar software showed that even on the messy, over-complicated and highly-specific Amiga side of things, "software emulation" was perfectly possible. As a matter of fact, the ShapeShifter was precisely that: a "software emulator" that took advantage of the hardware similarities between the two systems to make unnative software run.