AddMem

[Zac67]

I would've thought this is possible...
$80000 is just behind the 512 KB chip mem and the A590's RAM should start at $200000 - could you verify this?
If this is true, it must be some kind of ghosting problem: data written to $80000 gets also written to another location, e.g. $280000. To get rid of that you'd have to analyse (and modify) the hardware hack. Testing should be easy: don't Addmem the RAM and try writing to a memory location at its end. If that changes memory at a similar address on the 590, you're bust. Might even be a bug in the pre-Gary address decoder...

[Zac67]

... and without actually making it chip mem. The old (8367?) Agnus can't use more than 512 KB as chip RAM, so Amy would crash very soon...

[Zac67]

@Doobrey: it's a hack, so it uses Agnus as RAM controller (RAS/CAS generation & refresh), but augments the address decoding logic in such a way that a 2nd 512KB bank becomes accessible for the CPU. How exactly this is done might be the cause of the problem, e.g. when just A19 is used for bank switching.

[Zac67]

Probably there is some ghosting going on. Could you try:
- disconnect A590
- AddMem $180000-$1fffff
- see if it actually works / RAM test

Did you AddMem $300000-$3fffff with the A590 connected? (I don't think this is possible because of the way free mem is organized, but am not at all sure.)

Try this e.g. in Basic:
- connect A590
- don't Addmem
- peek($3fffff), note value
- poke something else to $0fffff
- peek again and compare value, if it's changed there is ghosting
- repeat with pokeing $0effff, $0dffff, ..., $08ffff

The best way to find out fore sure would be to open up amy and look what cable connections were added inside. To eliminate ghosting you'd have to modify the hack anyway...

[Zac67]

Yes, that's the RAS/CAS distribution to the RAM chips - but where do they come from?