Precision problem with matrix inversion

[Maro]

Thank you Road, you are right. When using approx(...) it works in textbook mode.

But I'm still surprised, since it should work also without using approx(...) in the Home environment, or what do I miss? Is there any reason why the Prime fails without approx(...) in this case? I'm working in RPN mode, putting the two matrices on the stack and divide them ... always resulting in zero-vector. How can I use approx(...) in RPN mode ??