The martensitic transformation is described in geometric terms, i.e. referring to the cystalline structure, without considering the diffusion or grain growth. Of course it's about a new packing of atoms which propagates at the sound speed, being  favored in the ordered structure.

Description of the non-martensitic(diffusion based) transformation does not use crystallographic information but rather the free energy F versus some parameter.

However, martensitic transformations (as any phase transformation) always have a thermodynamic description, which however is not dominant in explaining the optical microscope image.

The numbers you provided are self-explanatory.

The diffusion based transformations do not change the pole figures(of rolled material) according to some crystallographic relationship applied massively inside the grain but reflect all growth mechanisms.

I do not know if this is a general rule. It depends of course on what you mean with ordered structure. I guess you refer to chemical ordering, i.e. either short-range order or long-range order where certain atoms prefer certain lattice sites. The important point is then rather the thermodynamic driving force and the possibility of transforming one lattice into another by migration of transformation dislocations. In the classical case of steels the austenite is usually disordered. 

Dear Mohsen,

Martensitic transformation in steels is, in most of cases, taking place in disordered structures (form an atomic point of view), so there is not a general rule.

However, I agree with you that in many cases, for instance in most of shape memory alloys (SMA), the high temperature phase is an ordered one. Typical cases are the TiNi (NiTiNOL), which has a B2 structure and Cu-Al-Ni having a L21 structure. And in these alloys martensitic transformation is taking place very easily and in a reproducible and reversible way. Indeed, you indicate a case in which both situations (ordered and disordered can be studied). To analyse that, let me move to two concepts ordering and nucleation.

On one hand, when an alloy undergoes an ordering phase transition, this happens just to minimize the Gibbs free energy of the lattice. Then, for the same high temperature phase (for instance in Fe-Pt), the ordered lattice has much lower energy than the disordered one and consequently the jump of energy to the martensite phase (which still has lower energy) must be smaller when transformning from the ordered high temperature phase than when doing from the disordered one.

On other hand, the nucleation of the martensite phase involves short distance atomic motions in the lattice (not diffusional) that are typically classified by shears and suffles. Both kind of atomic motions are cooperative, being stimulated by the vibration modes of the lattice. In many cases (particularly in SMA) there are soft modes for which some particular shearing elastic constants practically are vanishing in proximity of the transition, promoting the transition from the high temperature lattice to the "frozen" lattice of the martensite. It is the case that in ordered structures (superlattices) the apparition and propagation of these cooperative motions and the corresponding soft modes are easier than in disordered structures, promoting the nucleation of the martensite.

I hope that this concepts could be useful for answering your question.

Kind regards