Ok, I have taken a more "scientific" approach.

I determined the theoretical maximum speed of my CPU as a function of core
voltage, and waterblock temperature, i.e. Speed (max at full load) =
f(vcore, wbtemp).

There is also a simple equation for determining the max power of the CPU as
a function of speed and core voltage.

Now using this formula I can understand why on a hot day (water block temp
approx 33C) the CPU would get unstable.

Also I can plot the ideal vcore and speed for the hottest day, which
incidently is 239 MHz at vcore of 1,64V. This applies approx 138 Watt power
to the CPU. Anything more will activate CPU thermal throttling.