components of velocity used by fluent and cfx
 

Which components of velocity are used by fluent and cfx?
curvilinear (covariant) or cartesian?
Which one is more preferable?
For an external turbulent flow over a blunt body which one do u suggest?

I don't know anything about CFX, but Fluent, in a sense, uses both. The momentum components are formulated in normal cartesian frame. I suspect the same to be true for all unstructured codes...only meshes that are mapped really have a uniquely defined covariant frame. Having said that, Fluent (and OpenFOAM) defines its mass fluxes as scalar values oriented normal to the boundary faces. In a sense, these fluxes are covariant.

I can't really answer your question about which is preferable. If you have a finite difference solver and an analytical mapping for the grid, the covariant frame may give you slightly better precision in resolving flows near the surface, as the surface-normal component will be small or zero and the tangent components may be orders of magnitude larger. In 64-bit floating point, I doubt that will make much of a difference though.