I'm hoping for some words of wisdom.  This is all NetBSD 1.6.2.

I have a long standing mystery - when I put a faster CPU in my AGP G4,
(7455 CPU 1Ghz) the machine got silly.  In fact, when I brought up my
new G4/AGP server (450 Mhz 7400 CPU) I found that these symptoms were not
limited to the 1Ghz CPU.

The easiest and most visible symptom
is that the ping times from this machine are 10.003 ms consistently.
The first ping is shorter (1-10 ms), but all subsequent pings
are 10.00x ms.

Tim Kelly kindly pointed me to a patch in arch/powerpc/mpc6xx/cpu_subr.c
where I simply set "powersave = 0;"  This "fixes" my problem.

I have ported patches for PR 29559, and a patch from last Dec. in
macppc/extintr.c to handle interrupts better, and these do not
affect this behavior.  It is somehow related to powersave mode.

Because the "other way" I "fix" it is by running a spin-loop
program in the background, I've concluded that there is something
not right in NetBSD 1.6.2 CPU handling when it goes into
powersave mode.  It acts like it misses interrupts (??)

Thoughts?  Advice?

These problems do not occur with NetBSD 2.0, so something has
changed there that makes it better.  I want the fixes for my
(production) server, so I don't want to run -current.

Many thanks in advance,

-dgl-