All about the (original KS1.4/ROM tower) A3000 (many questions, some answers)

[rhester72]

Long-time readers may be aware of my peculiar obsession with the original Amiga 3000...if not, prepare yourself.

If you owned one of these machines ("back in the day" or more recently) and have actual experience with it, please chime in!

I first saw the machine upon release in 1990 at a local computer store across the street from my university - it was sufficiently advanced as to cause me to spill a small amount of soda onto the floor, and sufficiently expensive as to be out of my budget for many, MANY years.

I've since owned a number of (used) A3000s, but none with tower ROMs, so I always hoped to experience the "original" under emulation...which appears to be very much a reality now.

Operating system releases that were actually distributed to customers include:

- AmigaDOS 2.0 (Workbench 36.69, Kickstart 36.141), shipped with the first machines?
A3000 Install Disk V2.6C [335603-02, 335603-03]
SuperKickstart 2.0 [367245-02]
Workbench 2.0 [317235-01]
Extras 2.0 [317954-01]
Workbench 2.0 (DE) [317954-03] * Was there no Extras disk?
Workbench 2.0 (International) [317954-02]
Extras 2.0 (International) [317235-02]

- AmigaDOS 2.01 (Workbench 36.69, Kickstart 36.143) - were the Install/Workbench/Extras disks identical to 2.0?
SuperKickstart 2.01 [367245-03]

- AmigaDOS 2.02 (Workbench 36.70 [nee 36.77?], Kickstart 36.207) - was there a DE or Int release, and does anyone know the part #s of the disks?
A3000 Install Disk V?.? [??????-??]
SuperKickstart 2.02 [367245-04]
Workbench 2.02 [??????-??]
Extras 2.02 [??????-??]

- AmigaDOS 2.03 (Workbench 36.102, Kickstart 36.209) - was there a DE or Int release?
A3000 Install Disk V2.8 [335603-04] *** I have this, but I believe it is modified, as A3000Install:Enforcer is not present but the InstallHD script attempts to use it
SuperKickstart 2.03 [367245-05]
Workbench 2.03 [317954-04]
Extras 2.03 [317235-04]

I've checked all the usual suspects, and I'm having a very difficult time locating pristine (or otherwise) copies of pretty much all of these disks - did anyone bother to preserve them for historical archival/research?

Also, I know the machines shipped with a set of Workbench 1.3 disks for "production use" until 2.0 was fully baked - they are installed automatically during hard drive preparation from the A3000 Install disks up to 2.03. I've long-known that:

- SetClock was "different" owing to the different clock chip/NVRAM in the A3000
- AmigaBASIC was removed entirely because it wasn't 68030 cache-friendly

It seems that the SetClock "fix" is actually handled by the A3000 Install disk - it copies over the complete contents of Workbench 1.3, then replaces C/SetClock with its own.

However, what about AmigaBASIC? The Extras 1.3.3 disk I have is mysteriously missing the binary entirely...was this a 'special' version of 1.3 that excluded it specifically for the A3000? (The A3000 Install script doesn't do any 'special handling' of AmigaBASIC at all, so if it were on the disk, it would copy it...the only way to ensure it's not there is to not have it on the disk in the first place.) So...

- Assuming there was a "special" run of at least Extras 1.3 (and perhaps Workbench 1.3, see below) for the A3000, was it labeled as such? Is the copy that's missing AmigaBASIC entirely the 'right' one for the A3000, or was my copy modified? Was the 1.3.3 release A3000-specific (it does not seem so, see next question), and 1.3.2 (and below) was the "final" release for the A500/A2000?

- Was there a 'special' Workbench 1.3 for the A3000? (Functionally, there wouldn't need to be, see the part about SetClock above...but it identifies itself on boot as "Workbench 1.3.3 for the A500/A2000", and Commodore was picky enough about presentation during this era that it seems they wouldn't let something like that slide.)

- Since AmigaBASIC was (deliberately) excluded for the A3000 AmigaDOS 1.3 installation, why are the AmigaBASIC FD 1.3 files deliberately _included_ (not only are they copied from the Extras disk, IconPos is used by the installer to 'fix up' the drawer icon for FD1.3, so it wasn't simple oversight)? Aren't the FD files useless without AmigaBASIC?
A quick question re: WinUAE 68030 MMU handling - if you power on a SKS ROM-tower A3000, you see it iterate through the screens of (by default) loading Kickstart 2.0, rebooting, then booting Workbench 2.0 via the WB_2.x partition. If you were to reboot the Amiga via Ctrl-Amiga-Amiga then, and held down both mouse buttons, would it ignore that and boot from the already-loaded SKS 2.0, or did it reload Kickstart every time and never cache it in memory (thus the double-button boot screen would appear every time you held them down)? I've heard it both ways, but I'm not sure which is correct.

Any other insight/clues people may have specific to the original A3000 would be most welcomed - I far and away consider it to be Commodore's crowning technical achievement and am *very* interested in properly and thoroughly preserving the experience for the future.

Rodney

[rhester72]

I don't have confirmation on this, but it appears that the A3000 actually shipped with Workbench 1.3.2 (not 1.3.3), with AmigaBASIC removed (as it was not compatible with the 68020+). It also seems possible that 1.3.3 actually came out _after_ the original Workbench 2.0 (36.6x) release, and that the International release of 1.3.3 also did not include AmigaBASIC (for licensing reasons?).

If correct, the Amiga 3000 Workbench 1.3.2 part numbers were:

- AmigaDOS 1.3.2 (Workbench 34.32, Kickstart 34.5), shipped with the first machines
Workbench 1.3 [367270-01], dated 1990-02-03
Extras 1.3 [367272-01], dated 1988-07-06, AmigaBASIC removed for incompatibility

The "A3000 Extras 1.3" disk does indeed have the FD1.3 directory intact.

Another question today's research raised:

- Cloanto's "Workbench 2.0" historical ADF is Workbench 36.68, but the earliest fully-documented customer-shipped release was 36.69. What was 36.68?

Rodney

[Arnie]

From what I have learnt through setting up am A3000 using WinUAE, is that you set which partition you want as default. Most people would use the default setting and have WB2 set as default.

If you hold both mouse buttons down on booting then you are presented with the dual boot screen where you can choose to boot from WB2 or WB1.3.

Also I believe holding down the esc key (may be another key) during boot brings up another menu where you can choose between 68000 & 68030 cpus, although, obviously, I can't get that to work in WinUAE.

[rhester72]

There's no way to switch between 68000 and 68030 as the A3000 does not have a 68000. Perhaps you're thinking of clicking the invisible close gadget in the upper-left corner on the boot screen to get directly into Kickstart 36.016 (i.e. the "tower ROMs")?

I know how to use both mouse buttons in 36.016 to select the Kickstart to load (either from HD or floppy), but my question is around how the MMU maps the Kickstart image it loads from disk - once mapped, is it persistent (i.e. survives reboots) or can you "re-select" each boot?

I was once told by an original ROM tower owner that, for instance, if you load SuperKickstart 2.04, the only way to get back to the menu screen and choose SuperKickstart 1.3 was to power off and back on before holding down both mouse buttons, but if that's the case, WinUAE isn't behaving that way.

Rodney

[Toni Wilen]

Originally Posted by Arnie

Also I believe holding down the esc key (may be another key) during boot brings up another menu where you can choose between 68000 & 68030 cpus, although, obviously, I can't get that to work in WinUAE.

Perhaps you meant A2620/A2630? It has 680[2/3]0/68000 boot menu.

I was once told by an original ROM tower owner that, for instance, if you load SuperKickstart 2.04, the only way to get back to the menu screen and choose SuperKickstart 1.3 was to power off and back on before holding down both mouse buttons, but if that's the case, WinUAE isn't behaving that way.

I have A3000 and I probably have "1.4" roms somewhere too (and if not, I can burn eproms) but as usual, it may take long before I bother with it

I know that A3000 boot code does check if MMU setup looks valid after reset. How do you reset emulation? Only keyboard reset works, reset button in GUI is hard reset.

[Arnie]

Originally Posted by rhester72

There's no way to switch between 68000 and 68030 as the A3000 does not have a 68000. Perhaps you're thinking of clicking the invisible close gadget in the upper-left corner on the boot screen to get directly into Kickstart 36.016 (i.e. the "tower ROMs")?
Rodney

I may have been confusing it with the A2000

[NMI]

The kickstart is persistent until you turn off the power.

I went through my disk-drawer and can conclude that my original A3000 was shipped with disks marked:
A3000 Kickstart V2.03 [367245-05 International A3000]
A3000 Install Disk V2.03 Rev2.8 [335603-04 International A3000]
Workbench Version 2.03 [317954-04 International A3000]
Extras Version 2.0 [317235-02 International A3000] - Yes 2.0, no mistake.

Workbench 1.3.2 [367270-01 International A3000]
The Extras 1.3 disk has unfortunately been replaced and I can't find the original.

The A3000 I bought second hand was delivered with:
A3000 Kickstart [367245-02]
(Install disk is missing. )
Workbench Version 2.0 [317945-02]
Extras Version 2.0 [317235-02]

A3000 Workbench 1.3.2 [367270-01]
Extras Version 1.3 [367272-01 A3000 English]

I don't have an Amiga available at the moment so I can't verify what is actually on the disk (if they are readable).

[r.cade]

I think this is answered, but all my 3000's I've owned have been tower ROM 1.4's.

Yes, the kickstart survives reset. You have to power off to choose again. If you have the right kickstart files, you can use other versions (3.0, 3.1, etc.) or even AMIX.

No, there is no 68000 menu. That is the 2620/2630 card that does that.

1.4 ROMs also won't work with the A3640 card- you have to have at least 2.04 ROMs for it to boot.

[rhester72]

Toni, et. al - thanks for the clarification on reset, I have been using the GUI. Q: If I Ctrl-Home-Insert (I don't have both Win keys), it looks like a keyboard reset, but I guru with 8000 0004 from within AmigaDOS 2.04 in WinUAE. After that, it looks like the disk-loaded Kickstart *is* maintained. Keyboard soft reset from 1.3 results in a dark gray screen and a lock-up. Enforcer is _not_ installed (as it's missing from my Install disk). Any clue?

NMI - dumps of all of those disks would be *VERY* useful if there's anyone around you that could do so (ADF is fine, there's nothing really 'special' about them except they are painfully rare . The Extras 2.0 (not 2.03) disk is...odd. 317235-04 is known to exist, but it's not like Commodore never made a mistake before. *LOL*

All - Per http://www.gregdonner.org/workbench/wb_133.html, 1.3.3 Extras did indeed have AmigaBASIC removed, as did (presumably) 1.3.2 Extras _as shipped with the A3000 only_, which means 1.3.3 did indeed come after 36.69. I can't find any obvious differences at all between the Workbench 1.3.2 disk as shipped with the A3000 and the same disk supplied with every other Amiga up to that point (as I noted earlier, SetClock is replaced with an A3000-compatible version at install time).

Rodney

[rhester72]

Splitting this into a separate message so it's seen.

Toni, the guru is coming from having MMU explicitly enabled...which seems like it ought to be necessary for the SuperKickstart mapping to work in the first place. Am I missing something obvious? =)

Also, it looks like without having MMU ticked, the Kickstart persists through the _first_ keyboard soft reset only - a second reloads the Kickstart from disk.

Rodney

[Toni Wilen]

Originally Posted by rhester72

Toni, the guru is coming from having MMU explicitly enabled...which seems like it ought to be necessary for the SuperKickstart mapping to work in the first place. Am I missing something obvious? =)

Guru was caused by stupid design flaw in MMU prefetch hack (That prevents programs to crash if they switch MMU on and then execute JMP (an) or similar from memory that disappeared when MMU was enabled)

Also, it looks like without having MMU ticked, the Kickstart persists through the _first_ keyboard soft reset only - a second reloads the Kickstart from disk.

This is a hack. Originally A3000 + 1.4 ROM support was a hack, was done long before MMU emulation existed. Real 1.4 ROM requires 68030 with MMU.

[idrougge]

http://eab.abime.net/showpost.php?p=983063&postcount=36

[rhester72]

Originally Posted by idrougge

http://eab.abime.net/showpost.php?p=983063&postcount=36

Good to know - were these ever Zoned? (And any idea why the Install disk was modified or how?)

Rodney