Why does an ill-conditioned model give different results in ANSYS 9.0 vs. 10.0?
This is due to the new PCG "difficulty settting," of PCGOPT,2, which is the default.
Setting pcgopt,1 gives results that match 9.0 exactly. A sparse solution also
matches 9.0 exactly. What is unusual about this model is the RBE3 condition,
which is connected to a short beam that is fixed at the other end. This situation
will be handled correctly by PCGOPT,2 at 11.0.