Intel Corp (INTC)

[chipguy]

Climate models show

I design chips by doing circuit simulations using extremely complex
models of transistor operation. These transistor models have close
to 100 parameters most of which are curve fit from the extensive
measurement of countless test chips and process monitor inserts.
Circuit design itself must account for both systematic processing
variation and also random device to device variation using corner
simulations across hundreds of parameter combinations as well as
Monte Carlo simulations across hundreds of trials.

For all this thorough and detailed computer simulation every new
chip looked at in the lab has some aspects of operation that doesn't
closely match computer simulation. Sometimes this requires a circuit
change, sometimes it can be accommodated by built-in trim.

This shows the practical limitations of computer modelling even in
an extremely limited and well understood regime of physical laws.

In contrast climate modelling is an extremely crude approximation of
many poorly understood physical, chemical, and biological processes
whose interactions are largely guesswork and hypothesis.

The potential effects of climate change are a serious business but
expecting reasonable predictive answers from computer models is a
tiny step removed from inspecting the entrails of a sacrificial goat.