7 Oct
2007
7 Oct
'07
10:38 a.m.
On Saturday 06 October 2007 17:30, Tony Duell wrote:
I don;'t see how software could do anything more. The purpose of memory
is to store data, all the program can do is say that it's not doing so
correctly.
To go further, I normally look at the timing of signals going to the RAM
chips (marginal timing causes all sorts of itneresting prolems), check
supply rails, etc. And then fidn the problem is due to ground bounce.
Seriously, the worst memoery prolems I've had to trace were due to poor
PCB layout.
Again, how on earth cna you know what the fault
is if you can't diagnos=
e
it. You mentioned swapping a DIMM. How can you
know if an intermittant
memory problem is a fault in the DIMM or in the memory controller on th=
e
motherboard?
I've wondered about that for a while now. I've used various memory=20
diagnostics on different machines over the years, and basically they all=
=20
seem to have in common that they write different patterns to memory and t=
hen=20
read it back for comparison. While this is okay as far as it goes, how =
DO=20
you deal with the situation you mention there? PGC =3D
Professional Graphics Controller. An IBM board set (3 boards
fitting into 2 adjacent ISA slots (there's a memory PCB sandwiched
between them) that form an intellegent-ish graphics card with an 8088 t=
o
control it (!). I have the techref, I'd love
to play with the boards.
Ah. I was wondering about that too. If you'd said "PGA" I would've
had =
a=20
clue right off.