A3000 ROM labelling

[Richard42]

I ran across a mystery this weekend while working on a couple of A3000s that I am repairing, and I was hoping that someone here could shed some light.

The A3000 has a 32-bit bus to the ROMs, of which there are 2, and they are 16-bits each. I verified that data line D0 goes to the ROM socket which is labelled ROM0, so presumably we have D0-D15 going to ROM0 and D16-D31 going to ROM1. So far, so good.

However, in order for the computer to work properly, you must install the chip which labelled as ROM1 (390630-03 for v2.04) into the socket labelled ROM0, and vice versa. Now that's just messed up. Also, by reading the data off the ROM, it can be seen that the high 16 bits are on the chip labelled ROM0 (which goes into the socket labelled ROM1).

It seems to me that the chips are just mis-labelled. Is this something to do with big-endianness? Can anyone explain?  Thanks!

[Matt_H]

I think the motherboard revision may also be a factor - what are you working with?

[Richard42]

I think the motherboard revision may also be a factor - what are you working with?

My observations above are based on the work I did this weekend on 2 A3000s, with motherboard revs 8.9 and 9.3. They are both labelled the same. But from what I observed, it seems to me that the chips themselves are mislabelled, not the motherboards. The high 16-bits of each word are in the chip labelled ROM0.

[Richard42]

There's no hardware gurus checking the threads up here in amiga.org?

[mechy]

I suspect the numbering matches the rom tower used in the rev7 boards which were the first production boards?  just a guess though, i dont have a rev7 3000 handy to verify this.

[thebajaguy]

http://www.amigawiki.de/dnl/schematics/A3000.pdf

Page 4 of the PDF...

[Richard42]

http://www.amigawiki.de/dnl/schematics/A3000.pdf

Page 4 of the PDF...

Yes exactly, this is the schematic that I was using while working on my A3000s. Note that the low 16 data lines (D0-D15) go to U180, while the high 16 data lines (D16-D31) go to U181. Then look at this picture of a rev 9.3 motherboard:

http://www.fascinationsoftware.com/media/A3000_rev9.3_ROM_sockets.jpg

On this motherboard silkscreen, U180 is labeled ROM0 and U181 is labeled ROM1. This appears to make sense. (However as a side note, and to further confuse things, it appears that the unpopulated ROM locations for the 27C220-style chips U182 and U183 are labeled wrong).

Here comes the strange part. If you want your computer to actually work when to power it up, the ROMs must be installed like this:

http://www.fascinationsoftware.com/media/A3000_rev8.9_ROM_installed.jpg

(Note this is a rev 8.9 motherboard but the silkscreen and component locations are the same). The chip marked as ROM1 must go in the socket marked for ROM0, and vice versa. If you read the ROM files with a burner, you will find that the low 16-bits are in the chip labeled ROM1, and the high 16-bits are in the chip labeled ROM0 (at least I did for Kickstart 1.4 and 2.04). Also, all of the TOSEC files that I found for Kickstart ROMs also named exactly like the chips are marked, ie backwards.

[thebajaguy]

When you considered that they printed the PCB motherboard with two different ROM socket pin outs (the space wasted!) to cover their bases, and also made it possible to run a single A2000-style 16-bit ROM (note the pins for the jumper option below the lower socket), a screw-up in labeling the code and having the ROMs burned / labeled wrong in production was probably thought of.  Remember the 2.04 ROM with the wire for the early A500 models?