6 Sep
2005
6 Sep
'05
6:20 p.m.
Date: Mon, 5 Sep 2005 21:11:02 +0100 (BST)
From: ard at p850ug1.demon.co.uk (Tony Duell)
> >Date: Sat, 3 Sep 2005 00:17:50 +0100 (BST)
> >From: ard at p850ug1.demon.co.uk (Tony Duell)
>
> >Most, if not all, of the single-bit-wide DRAM chips (4164, 41256, etc)
> >had separate DIN and DOUT pins.
> What is the point in using them separately? Was
it for the case
> where the destination of a read is different from the source of a
> write?
YEs, that was the idea. It's no use if you want to
link the data pins to
a bidirectional bus.
I suspect the _real_ reason for it was that there was a
spare pin on the
package and it didn't ever make life more difficult to have separate DIN
and DOUT pins (oyu could just strap them together externally if you
wanted a bidirectional data pin).
I'll be getting a IIfx on Thursday and I can confirm (I hope) that
the DIN and DOUT are common on the mother board at that time. Your
advice has given me hope that this is much more doable than I feared.
If they're tied, then I should be able to build a 16MB SIMM using two
16M X 4 chips. My concern was that the X4 chips do not have separate
DIN and DOUT and I wasn't certain if the IIfx actually needed the
separation. I think I have some 16M X 1 chips which would work in
either case, but my supply of 16M X 1 chips is much smaller.
If the DIN and DOUT are common on the MB, and I use 16M X 4 chips, I
assume that I should either use the DIN pin or the DOUT pin on the
SIMM but not both? Using both could result in different length
datapaths which remerge, possibly causing interference from slightly
out of phase signals. The difference in length probably wouldn't
affect the timing much, but still...
It would still be handy if someone has a mechanical drawing of the 64
pin SIMMs. I can derive it from a physical example, but I'm less
likely to get the actual center measurement and tolerances correct
that way. Hmmm. I'll check around and see if any of the socket
manufacturers are still selling a 64 pin SIMM socket. If so, I
should be able to derive the SIMM dimensions from the socket
datasheet.
Jeff Walther