My HP GPS/PDA (sirf star 3) does NOT correct for the variations in earth's density/gravity across the globe using the EGM96 geodetic model.
Garmins DO correct. I couldn't work out why there was consistently a approx 100 ft difference between them, until I read about this somewhere.
As it happens, NW Europe sits on an extreme of variance from the WGS84 ellipsoid datum. (I looked for the map of variance, but cant find it). I seem to remember that in the UK, the GPS will over-read by about 140' without correction, and India is an area that under-reads significantly.
In some gliding comps, due to various GPSs used, (some which compensate and some which don't), a small latitude is allowed where the GPS shows a climb into airspace, because the barometric altitude may have been OK.