RE: What is real lowest speed of PCI?

["Ingraham, Andrew" <Andrew.Ingraham@compaq.com>]

Henry,

The "DC to 33MHz" requirement has been around for a long time, not just the
latest spec version.  You must have a very old one (possibly before the idea
of PCI plug-in cards was added) that has a 16MHz lower limit.  (Or maybe you
are thinking about the provision, even in rev. 2.2, that operation at
frequencies below 16MHz may be "guaranteed by design rather than by
testing."  That is just a way to make it easier to test chips faster.)

Several motherboards have been made that did not run the PCI bus at 33MHz.
Many older Pentium boards, still in use today, had 30MHz, 25MHz, even 20MHz,
to get the clock ratios to come out right.  This is not an issue with newer
corelogic chipsets.

Section 4.3.5 says that one way to build systems with more PCI slots, is to
use a slower clock.

No PCI card can operate fully (i.e., do anything useful) while the clock is
0Hz.  Obviously no data can be transferred at 0Hz.  But PCI doesn't
guarantee a minimum data transfer at any speed!  Even at 33MHz, if your card
needs to compete with others, it may not get the bandwidth it wants.  But it
must not crash and burn and require a reboot, just because the clock
stopped.

It is possible for a system to pulse the PCI clock at varying rates.
Nothing in the PCI spec (for <33MHz) says the clock is continuous and
uniform, and you should assume that it looks more like data than clock, with
periods that might vary from cycle to cycle.  Portable systems may go into a
"sleep" mode where the clock stops or becomes very slow.  Your PCI adapter
card should remain connected even though it isn't passing data anymore, and
shouldn't require a system reboot when the clock resumes.

Finally, keep in mind that PCI is not tied to any particular CPU or
architecture.  Even though the majority of P3-based desktops made in 2000
may run the clock continuously at 33MHz, others may not.


> Nobody seems to be interested in extremely low speed. 
 
You might be if you ran on batteries.

> People simply piss
> off the spec and still claim PCI compliant, or my understanding in
> incorrect?
 
I hope it is incorrect!