[rescue] Sun 3/110 Memory Issue

[Ahmed Ewing]

On 8/20/07, earl at baugh.org <earl at baugh.org> wrote:

> However, I've got a memory issue with the 3/110... I get the
> following output on the console with diag enabled:
>
> PROM Checksum Test
> DVMA Reg Test
> Context Reg Test
> Segment Map Wr/Rd Test
> Segment Map Address Test
> Page Map Test
> Memory Path Data Test
> Err 2: Addr 0x00000400, Exp 0x00000001, Obs 0x00010001, Xor 0x00010000
>
> Can anybody tell me how to determine what bank this is on the 3/110
> board, so I can replace the chip?  They all look socketed
> and I've got a drawer full of RAM that I can swap out, if I can
> figure out the chip...
>
Yes, they are socketed... this may have occurred to you already, but
as long as this is a fully recreateable error which never allows
successful POST completion in diag mode, you can figure it out. It
stands to reason that if you do some trial-and-error swapping of one
good "control" chip to other socket locations in succession,
eventually thru process of elimination you'll see the "Addr" value
change in the POST error, at which point you've isolated the physical
location of the culprit. Presumably this could be completed in a
single (albeit tedious) sitting, assuming the chips aren't too hard to
pull and POST doesn't have to get too far before choking on the bad
chip.

Plus you could then use the known bad chip in the same way to create
and publish a definitive address-to-socket map of the entire layout
for posterity... I know I'd be interested ;-)

Hope that helps,

-A