The top end of both 1k resistors MUST be tied to the same voltage as the Source of the PFETs. If they are not, then the PFETs will stay on!
It matters not if my R1 (3.3K) is returned to 5V or 12V. At 12V it dissipates a max of E^2/R = 12*12/3300 = 44mW.
If you are messing with surface mount components, I'll bet you just mixed up the resistors.
When you simulated it, did you use some "real" parts (ones that have similar On resistances, threshold voltage, gate capacitance), or where you just using the default models. The default models are "tiny" devices, and will give unrealistic results.