Re: Kickstart 3.1 versions -- What's the difference
@Andy
Lets Talk Amiga Roms
Amiga Roms, are pretty much like Firmware, however as opposed to just storing memory ranges (like that in a PC) they actualy store small programs that are run on boot.
Most notibly the firswt thing to run on boot is Exec this is a preemptive multitasking kernel and everything else runs on top of that. Kickroms also include an abstraction of the Amiga's unique hardware, a disk operating system called AmigaDOS, a windowing system API called Intuition which provides a *base* graphical user interface (windows) that is called by Workbench.
Amiga A500, A600 and A2000: Rom unifictation
Now before 3.0 and 3.1 there were many differing versions of 2.x, some had native IDE support, some only partial and some none at all. Ssome 2x ROMS had PCMCIA mapping and some didn't (and this even includes some that were shipped with the A600!)
Taking the A600 aside for the moment, the A500 and A2000 are almost identical in design, with the exception of a zorro II bus, the chipset and electrical layout is in fact identical. This is one of the reasons why Jay Minor went on record (and against commodore) by saying "The A2000 does not represet a significant enough of an upgrade to invest in" not that it stopped anyone lol. Within 8 Years A2000's equipped with 040's became the work horse for Raytracing / rendering and Video Toasting.
Now lets get back the A600
The A600 unlike the A500/A2000 has a PCMCIA port, this is a mapped area of memory space that can be ignored and over written, so in the ROM unification of the A500,A600 and A2000 under the official Kickstart 3.1 there exists memory mapping for the PCMCIA port even though the A500 and A2000 dont have this circuitry.
Getting to grips with the A1200 / A3000 and A4000
Surprising as it might sound None of these machines share a direct architecture, sure the A4000 uses AGA as does the A1200, but after that they have very little in common. The A1200 architectually speaking is A lot like an A2000 with AGA and a base 020EC processor.
The A1200 has essentially a Zorro 2 bus, however although its a 32bit machine it only has a 24bit encoder/decoder of memory addressing. This limmits the memory map drastically comparied to a full core 020 or higher.
Whoms the close relative?
The A3000, although uses an ECS chipset is a full 32bit design, having Zorro 3 and DMA transfers (very important btw), in a lot of ways the A3000 is more like an A4000 both have Zorro 3, DMA and 32bit wide architecture from the offest, and thats why an A3000 is more like an A4000 compared to the A1200 (irrespective of AGA chipset).
A little bit about A3000 and Amber
The A3000 also has an Amber Circuit that was an optional card on the A2000, implemented on the A3000 and then removed and placed an an optional component on the A4000 video port. Due to the extra Bitfiled of AGA it was cost effective from Commodores standpoint to allow the user to *upgrade* to use an industry standard monitor was quite a cheek at the time compared to the price of an A4000!!! (there method was to bundle multisync monitors)
So why two chips of ROM instead of 1?
As you have touched on, its all a matter of bits and bytes =D, the A1200/A3000 and A4000 are infact 32bit machines and as such process their data in 32 bits wide, as such they access (address and decode) both roms (16bit each) simultaneously for a (32bit wide data)
It can be a trick to get your head round the 24bit nature of the A1200 ec020, but this 24bit limmitation is its amount of memory to address not how it addresses it.
there.... that should help.