[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: 66Mhz 64bit motherboard



...
> clearly shows two 64-bit,
>_5.5V_ connectors ... yet the text directly below it says "Two 64-bit
>(66MHz) 3.3V PCI slots".
>
>Unless the photo is wrong, they can't be 3.3V slots so they can't be 66MHz
>either.
 
This tickled my memory, so I checked my old mail.  A few months ago, there
was some mention in this forum that Intel L440GX motherboards did
effectively the same thing: 66MHz slots with connectors keyed for 5V!  I
found (and still find) this rather disturbing.

Someone implied that these slots might have had 5V signaling if you plugged
in a 5V card, or 3.3V signaling if you plugged in a universal card.  You
couldn't plug in a 3.3V card, of course, so some cards that are truly
designed for 66MHz might not be usable in those 66MHz slots!

But that explanation doesn't make sense, because there is no way to tell
whether you have plugged in a 5V card or a universal card, except by
"looking" at the keying.

If Intel is using 3.3V signaling on a connector keyed for 5V, as they
apparently are doing (in order to support 66MHz), it is a violation of the
PCI Spec.  Why they (apparently) knowingly violate the PCI spec, I don't
understand.

I think, perhaps, they are capitalizing on the fact that 5V and 3.3V
signaling can be nearly the same.  It's easy to make the output drive levels
and input thresholds satisfy both simultaneously.  The only necessary
differences are: (1) the presence/absence of overshoot clamps just above
3.3V on all I/Os; (2) the keying of add-in cards and connectors; and (3)
whether +3.3V or +5V is applied to the Vi/o connector pins.

My guess is that Intel violated (2), and Tyan followed suit.  How they
handle (1) and (3), I don't know.

Or maybe, as someone suggested, Tyan used the wrong connector for the photo
because the right one was in short supply when they needed to get a picture.
(They are 64-bit connectors, so the 3.3V version really IS different.)

The Tyan board does have a hardware jumper to select whether the bus runs at
33MHz or 66MHz.  This shouldn't be necessary because the M66EN pin on the
connector is supposed to take care of that automatically.  So maybe this
jumper REALLY selects whether to apply +5V or +3.3V to the Vi/o pins, and
whether to clamp above +5V or +3.3V on the bus.  Set it wrong, and you risk
letting out the smoke!

If system designers followed the rules, this couldn't happen.

Looking at some other web pictures, there are systems out there that have
the right (3.3V) connector for the 66MHz slots, so (fortunately) not
everyone is doing it this way.

Andy Ingraham