XeNTaX

Data is packed mostly in ".bnk" archives, which is extractable using TDU tools (TDU and VR3 were both made by Eden Games)
I've attached what pops out after the extraction, a ".g2d" file which I guess is texture related, and a ".g3d" which should be the 3D part of the package.
A guy said (here >>> viewtopic.php?p=97370#p97370) these can be already imported, but I don't know how, and which tools are needed.
I've attached both files, original ".bnk" package included, just in case.

http://www12.zippyshare.com/v/qu1KJPX5/file.html

well, you'll have to collect the faces of the submeshes somehow:

using triangle strips it looks kinda better: I'd think about creating a Make_H2O project but it would take too much time, I guess.
(As you may know in most cases I go for the first submesh only.)

Here the chassis could be done easily provided all models share the same format, FVF= 12 bytes, tristripped face indices (do they?).

The mesh is marked by 'GEOM' so its start is easy to be found.
But I don't know where to find the vertex count(s). For the first 6 (or 7) submeshes it should be 2802 or 2844 (0xB1C) in sum.

Also didn't check for uvs so far.

just another lifetime wasting project?
There are about 77 VFX objects with 147 submeshes (in sum).

I won't bother finding out the hierarchy ('HIER' sections), position offset can be found here, maybe, or in 'OBJ' sections.

I'll upload a tiny exporter sooner or later which will not care for positions, I guess: (upps, forgot about uvs, should be in UVF sections, nice format, though )

(Did you try out a 3D ripper? From what I saw with 'Ride' (PC) it does not make sense
to dig in the hex data if you can get the model with a ripper.)

Yes, I've tried and the rip was screwed, something between the first pic and the second pic you posted

Automotive Gaming wrote:Yes, I've tried and the rip was screwed,

well, no surprise.
There's some problems with analysing, too many missing objects, some missing faces though I've made some progress: (I know there's people on Xentax who could handle this better but I remember only one of them who cared for cars.)

I managed to include the missing objects: (The uvs are a little bit refractory though they're floats.
When entering the parameters into hex2obj and clicking the 'UVs' button
after having loaded Cordoba.g3d you'll see what I mean.)

Well, I could tell you that the uvs for this submesh probably start at 0x65818 with an UVB size of 8.

But you won't believe me that I found it by trial&error, will you?
It's an offset of vertexcount*16 to the supposed original start address 0x634A8, ok.

finally I got the uvs working somehow: (No time to check the first texture set.)