It sounds like one of two things is going on:
1) The part is broke. I would just the supply current and the input bias levels. There should be about 20 mA of supply current and 100mV below the positive supply bias on the inputs.
2) You don't have VMAG shorted to MSET and VPHS short to PSET. If the feedback loop isn't closed, then it will act sort of like a comparator and rail. I would measure the resistance between VMAG and MSET; same for VPHS and PSET. It should either be close to zero or in the mega Ohm range.