"flipper" wrote in message
...
Then there was the time memory write/read diagnostics passed with no
RAM installed. LOL

Ha! I seem to recall hearing about that in IBM-PC compatibles. There are
supposed to be pullups on the data bus, so it open-circuits to FFh when
nothing's addressed. But if data were present sufficiently recently
(either from the instruction being processed, or a previous I/O / DMA /
bus mastering operation in the case of an I/O bridge*), and the charge
wasn't bled out by pullups or TTL inputs, it would happily remain floating
at whatever level by the time it's read. The bus charge will act as one
byte of DRAM, accessible from all unaddressed locations!

*AFAIK, an I/O bridge is standard equipment, at least on the 16 bit
models. My Amstrad 1640 has a 16-bit "FSB" so to speak, coupled to the
8-bit I/O bus through some gate array chip. I'm guessing this bus is only
active when addressed, since the manufacturer would know exactly where all
that "FSB" memory is going (00000-9FFFF), easy to minimise bus operations
that way.

As I recall, all empty locations in said computer read FFh as they're
supposed to. I don't recall if I tested it at speed; maybe it's possible
to observe the effect in consecutive operations.

Also about that computer, the documentation (which I managed to find
online) is quite technical. It tells a lot about what the BIOS does, how
the hardware works, where all the registers are and what they do, and so
on. With that information, you could write your own ROM, boot loader,
build your own hardware, etc. It also mentions how the memory is scanned
during POST -- combinations of bit and address patterns, scan up, scan
down, etc.

Heh.. I never did see the value in checking memory. In the old days it
was supposed to be unreliable, they even added parity (I've got a whopping
80kiB wasted on parity RAMs!). I've *never* seen a memory error on an old
machine, and as far as I know, the RAM chips never fail, at least not
after possible infant mortality (which I can't possibly know about, being
that I am only 1 year older than this machine, which is a pretty new
machine for an 8086!).

Ironically, new memory seems a lot less reliable, though I am of course
hard pressed to do a proper statistical comparison between 2000 IBM-PCs
that I've never seen, vs. a single 2GB DIMM running at clock speeds a few
orders of magnitude faster...

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms