> On May 11, 2023, at 10:41 PM, Cameron Kaiser via cctalk
> <cctalk@classiccmp.org> wrote:
>
>> I thought about this, but the KIM is a pretty simple system. The only memory
>> mapped device in that range (really, on the entire unit) are the RIOTs, and
>> their RAM at $1780 is fine and does not echo.
>>
>> The KIM only does address decoding for 8K and echoes the rest, so the same
>> fault is mapped at $2280, $4280, etc. I would think this would still suggest
>> data is the problem.
>>
>> I suppose I could randomly replace the RAM and see what changes but again it
>> seems weird to have a fault so neatly aligned and only in a specific range.
>
> ...
> and the high nybble is OK). Now that sounds more like a bad RAM chip, but why
> would it be *just* those addresses? Does that sound like a plausible failure
> mode?
Yes, if they are small RAM chips. For example, in a 1k DRAM chip (and perhaps
in slightly larger sizes too) the row size could be 32, which means that the
failure of one row on one RAM chip would cause the failure of an aligned 32
byte range.
paul
