The pmax (at least the 3maxplus) seems to massively lose
clock interrupts as soon there is some system (in particular,
network) activity.
A diskless (root on fta0 - FDDI) box which keeps a quite
accurate time normally was 3 seconds behind after a single
compile of "bash":

ntpdc> peers
     remote           local      st poll reach  delay   offset    disp
=======================================================================
*zelf01          134.94.233.21    1   64  377 0.00183 -0.001696 0.00093
ntpdc> kerninfo
pll offset:           -0.001586 s
pll frequency:        6.655 ppm
maximum error:        0.023256 s
estimated error:      4.3e-05 s
[...]

[compile "bash"]

ntpdc> peers
     remote           local      st poll reach  delay   offset    disp
=======================================================================
*zelf01          134.94.233.21    1   64  377 0.00214  3.086226 0.00099
ntpdc> kerninfo
pll offset:           0 s
pll frequency:        512.000 ppm
maximum error:        0.556237 s
estimated error:      0.343926 s


ntpd is not happy, obviously...

Anyone an idea what caused it and how to fix this?

best regards
Matthias