Please do, I _have_ a working News-connection but temporarily
I cannot post anything and my System Administrator is too busy
to repair it. Maybe tomorrow...
(got another message from You while writing this, so comments to that too)
|1) The RAM chips are too slow, e.g. 120ns (not true, mine are 80ns)
I added wait-states and reduced clock speed, did not help.
|2) Remove or reseat the SIMMs (I tried to reseat them, but didn't work)
I moved the memory banks around, did not help (new memory works perfectly
when it is used as a low part of 4 megs)
I Have ideas of two possible reasons:
1. Linux is broken, it thinks that if you have more than 4 Megs, You must have
8 megs. Has anybody reported that He/She has Linux 0.98 working in 5 or 6 M ?
2. The address lines are broken; the same ramchip is active in two different
addresses (I had such a problem once in another machine in 1986 I think.
It was a difficult job to find that bug !!!)
So when CPU tries to write to an address above 4 Megs, it is succesfull
and data can be read from that same address. BUT !!! the same data is
visible in some address below 4 megs and causes "some confusion" there.
This could be tested by a program which:
- fills the memory below 4 megs with 0ffh
- fills the memory above 4 Megs with 0h
- checks that nothing has changed below 4 megs
I don't have tools to test this. pity
BUT: I have tested my RAM with CheckIt and no errors were found. I am not
sure if this "stuck address-line" error is checked in that program.