Is the Amiga architecture still relevant today?

[Lockon_15]

@ejstans
 
A500 is a great example of technology compromise Bloodline mentioned.
To be aware of what that actually means you have to know A1000 well in detail.
A500 - It wasn't designed for major upgrade path in mind, since for that purpose there was A2000 launched side by side. That crappy (as you said) trapdoor expansion was not part of Agnus upgrade either, but wise engineering made i.e. slow>chip ram hacks feasible at field service level. If you take a good look on various A500 revisions, you'll see the impact of cost economy on motherboard design - first revisions (Rev3 & 5) had then expensive DIL DRAM covering the whole lower left area for just a mere 512Kb of ChipRAM. In spite of hacks, I don't really belive CBM left trapdoor bus just to allow ChipRAM increase somewhere in future, when RAM price drop occur. It was a best-fitting solution for that time - it wouldn't compromise FastRAM bus; it won't help you with ChipRAM, but it may persuade Amiga development to start using. And it did in great exent, just after Kickstart v1.3 debut. Later revisons introduced cheaper DIL RAM with more capacity, which made platform even more attractive for modding. A500+ went to the edge, preparing for full 2Mb ChipRAM at the cost of trapdoor slot. So, in some way, CBM closed the circle.
 
If the crowd was left with true FastRAM only, platform wouldn't attract more audience than A1000 and there would be no A1200 nor else.
AFAIK, A500 trapdoor 512Kb expansion might be the most selling Amiga peripheral.

[ejstans]

@ejstans
 
A500 is a great example of technology compromise Bloodline mentioned.
To be aware of what that actually means you have to know A1000 well in detail.
A500 - It wasn't designed for major upgrade path in mind, since for that purpose there was A2000 launched side by side. That crappy (as you said) trapdoor expansion was not part of Agnus upgrade either, but wise engineering made i.e. slow>chip ram hacks feasible at field service level. If you take a good look on various A500 revisions, you'll see the impact of cost economy on motherboard design - first revisions (Rev3 & 5) had then expensive DIL DRAM covering the whole lower left area for just a mere 512Kb of ChipRAM. In spite of hacks, I don't really belive CBM left trapdoor bus just to allow ChipRAM increase somewhere in future, when RAM price drop occur. It was a best-fitting solution for that time - it wouldn't compromise FastRAM bus; it won't help you with ChipRAM, but it may persuade Amiga development to start using. And it did in great exent, just after Kickstart v1.3 debut.
 
If the crowd was left with true FastRAM only, platform wouldn't attract more audience than A1000 and there would be no A1200 nor else.
AFAIK, A500 trapdoor 512Kb expansion might be the most selling Amiga peripheral.

In what way is slowmem best fitting?! Your explanation just leaves me puzzled. Your last paragraph doesn't make much sense to me. Yes, the 512KiB expansion is probably the most sold peripheral. That's because it dropped in price extremely significantly. After awhile it could be bought for some $20 or so. Sideslot fastram cost like five to ten times this! Which expansion are people going to buy?
 
That's why I said: "imagine if the trapdoor ram had been fastram". There needed be no difference in the actual ram expansion; the only difference is which of the buses it would be on.
 
If the trapdoor had been on the fastram bus than the chipbus, how on earth would that be a bad thing and "not attract more audience than A1000"? Doesn't make any sense! And certainly doesn't seem like a reasonable compromise...

[blakespot]

Au contraire mon ami...



OS9 onwards had Preemptive multitasking.

Now, if you'd said Amiga was the first to have preemptive multitasking, you'd have been correct.

I'm afraid you're incorrect again.  MultiFinder, released in 1988, allowed cooperative multitasking.  Also, no version of Apple's "classic" OS (all Mac OS's prior to OS X) allowed preemptive multitasking.  A sloppy threading capability was tacked onto, I think it was, Mac OS 8 - but it was not an ideal model.

Although, and most are unaware of this, the operating system of the Apple Lisa (which predated the Macintosh) offered cooperative multitasking and protected memory.  It was based on a 5MHz 68000 -- all before the Mac existed on the market.




blakespot

[blakespot]

There were plans in the early 1990's for hardware and software that would have extended Commodore's technological advantage and made your P100 CPU with 16 meg running Win 95 every bit the boat anchor that it was.  No amount of bug-fixes for that set up would have turned it from the horse-drawn carriage that it was, to the modern motor car that the Amiga still was.  But some illegal business practices from MS, stupidity from IBM to let the x86  patent lapse,  plus total business incompetence from Commodore, along with some smart business practices like selling cheap to the business world and subsidizing workers home computer if they ran MS crap, results in inferior technology eventually winning out.  Apple was on its knees for the same reason, and was saved by a portable music player.

Not to lay down a tangent to the discussion, but Apple was saved by the iMac and Jobs' return.  A portable music player turned the company into a massive force, though.  The local retail stores helped significantly, as well.



blakespot

[Lockon_15]

OK.
I'm not 100% sure, but getting trapdoor to Zorro bus while leaving sidecar open for expansion would involve Buster which is no-go for entry level A500. Moreover, as you already know, all trapdoor expansions above 512Kb were built with so called "Gary adapters" which provided missing address signals needed for bank switching. The most advanced of those were designed to host 4Mb RAM which was used as part-Chip, part-Slow combo, never true Fast. I found this case most convincing on the issue of trapdoor mapped to Zorro bus.
 
Next, price difference - trapdoor vs sideslot (sidecar, whatever). Of course they were expensive - in 80% of cases they came as a secondary or tertiary feature, alongsinde SCSI/IDE controller or more potent CPU (020/030). A500 made it's fame through gaming which requirred stock 68000 and 1Mb RAM, untill 1991. Who needs 68020 and harddrive, out of productivity purposes ? For that matters, you go for big box Amigas.
 
When I said "... was left with true FastRAM only...not attract more audience... " I meant that trapdoor FastRAM was not feasible and that if there were no other alternative, all those pricey sideslot expansions would make A500 just another A1000. Does this makes any more sense at all ?

[ejstans]

OK.
I'm not 100% sure, but getting trapdoor to Zorro bus while leaving sidecar open for expansion would involve Buster which is no-go for entry level A500. Moreover, as you already know, all trapdoor expansions above 512Kb were built with so called "Gary adapters" which provided missing address signals needed for bank switching. The most advanced of those were designed to host 4Mb RAM which was used as part-Chip, part-Slow combo, never true Fast. I found this case most convincing on the issue of trapdoor mapped to Zorro bus.

I don't want it to be a Zorro device, just out of the chipmem bus onto the CPU bus so that it can be used in parallel with the custom chips. Making it a Zorro peripheral wouldn't make sense since the point is to make it cheap and I'm guessing the glue logic to support the Zorro protocol is at least partly responsible for making the sideport expansion so expensive.
 

Next, price difference - trapdoor vs sideslot (sidecar, whatever). Of course they were expensive - in 80% of cases they came as a secondary or tertiary feature, alongsinde SCSI/IDE controller or more potent CPU (020/030). A500 made it's fame through gaming which requirred stock 68000 and 1Mb RAM, untill 1991. Who needs 68020 and harddrive, out of productivity purposes ? For that matters, you go for big box Amigas.

This is a silly argument. Yes, gaming on Amiga 500 was restricted to 68k and 1MiB (slow)mem. But had a cheap way to add fastram been available as opposed to slowmem, of course the gaming companies would have taken advantage of it! Fastram means that the Amiga can actually operate the way it was intended: the custom chips can run in parallel with the CPU and provide extra processing capabilities. If there is no fastram, there is not enough available memory bandwidth and the copper and blitter will instead stall the processor! Not too mention the bitplanes...

 

When I said "... was left with true FastRAM only...not attract more audience... " I meant that trapdoor FastRAM was not feasible and that if there were no other alternative, all those pricey sideslot expansions would make A500 just another A1000. Does this makes any more sense at all ?

Okay, so you meant that it was either slowmem trapdoor expansion or no trapdoor expansion...

But I will question that, and maintain that it would make much more sense to have a cheap way of adding fastram so that the CPU would have enough memory bandwidth. Really, lack of bandwidth is probably the biggest short-coming of the architecture!

[bloodline]

@ejstans

The sideport is basicly a direct bus to the 68k (which is all zorro really was), if you want to have both the side port and fastram, you need to have some bus arbitration hardware (ie buster)... The A500 was a cheap device, so opted to have a single zorro slot and a single memory expansion (that was arbitrated by agnus to save costs).

[ejstans]

@ejstans

The sideport is basicly a direct bus to the 68k (which is all zorro really was), if you want to have both the side port and fastram, you need to have some bus arbitration hardware (ie buster)... The A500 was a cheap device, so opted to have a single zorro slot and a single memory expansion (that was arbitrated by agnus to save costs).

And the (fast) ROM ?

[bloodline]

And the (fast) ROM ?

The Kickstart ROM is addressed via gary, IIRC.

[ejstans]

Ah, ok, I looked at schematics. I basically forgot about DRAM...Agnus had to do the DRAM address translation (+refresh). That does explain it...Still, I think it's a bad compromise. I think it would have been worth it to add DRAM support to eg Gary in order to have cheap fastram...

But of course, that's in hindsight. Maybe Commodore thought cheap fastram sideport expansions would flouris?

[amiga92570]

Amiga did a good job in it's day of making computing accessable to folk. However, it has to be said that things have moved on a very long way from Windows 3.11.

You mean to tell me windows 3.11 is not current. I must run out and get the new version.:roflmao:

[bloodline]

Ah, ok, I looked at schematics. I basically forgot about DRAM...Agnus had to do the DRAM address translation (+refresh). That does explain it...Still, I think it's a bad compromise. I think it would have been worth it to add DRAM support to eg Gary in order to have cheap fastram...

But of course, that's in hindsight. Maybe Commodore thought cheap fastram sideport expansions would flouris?

Well, by the time Commodore took control, it was all about saving money! Agnus has a DRAM controler, so they just used that.

[ejstans]

Well, by the time Commodore took control, it was all about saving money! Agnus has a DRAM controler, so they just used that.

Yeah, it's not such a completely bad idea as I thought, and it probably made the Amiga 500 (a little) cheaper. The Amiga 1000 shipped with only 256KiB but an option to upgrade to 512KiB. I guess maybe that's where the idea of the trapdoor expansion came from, even though in the Amiga 500 the extra memory couldn't be used by the custom chips.
 
Anyway, I'm glad the reason for slowmem was sorted out, because when I wasn't considering the DRAM addressing issue, I just thought it insane to tie the trapdoor expansion to the chipbus when it could as easily have (in my mistaken mind) been sitting directly on the CPU bus and been providing a great memory bandwidth injection. But if in penny-saving mode, I guess it made sense to do it that way, although it is a real shame...I mean, it probably wouldn't have costed so much to just clone the existing DRAM controller...
 
Ahhh, anyway, I do maintain that this was not a good compromise! Had they chosen my way, the Amiga of 1987 would have been just about perfect, considering the constraints of the time! :lol:

[persia]

I think you're right, Amiga was put on the road to irrelevancy when Commodore toke over, it just took a decade or so for it to happen.

Well, by the time Commodore took control, it was all about saving money! Agnus has a DRAM controler, so they just used that.

[bloodline]

The penny pinching you have identified became a cancer in the Amiga Development... That is why the outdated AGA wasn't released until 92... When it needed to be out in 88...

[bloodline]

The penny pinching you have identified became a cancer in the Amiga Development... That is why the outdated AGA wasn't released until 92... When it needed to be out in 88...

Also communication between the dev teams in commodre was nonexistent!!! The CDTV team had no idea that the ECS chipset was even being worked on let alone ready to ship...