Hi @OTsyganov, this is actually timed well as I have done alot of work recently validating drag results, here is a good write up for a NACA 0012, indeed you are correct, default numerics are not sufficient for drag:
'NACA 0012 Verification' simulation project by dlynch
There is a heading in the write up about ‘Numerics’ read that first, also I have found best results for K-omega SST when Y+ approaches 1. I reckon both of these will solve your problem.
Best,
Darren
Edit: Also see how it is effected by mesh finness, in the independence study.