This was posted on a US site
Up to 15kt of error (against a known-accurate airdata computer) and with varying error directions.
Wind is calculated from GS, TRK, HDG and TAS. The former two are directly determined by the GTN, the latter two must be fed to it. The error must therefore stem from the air data. The actual calculation is trivial.
My ADC is part of the Aspen which has all sensors builtin. It should be very accurate.
Dont forget the OAT/TAT. My guess is that they are not getting the same inputs or the heading input is the analog x, y, z and they are resolving slightly different headings.