well, 40 fps with 14k polys on screen doesn't sound that good to me, with a good system like yours.
Maybe it is due the performance loss of some of the recent FSO builds, or maybe it is your model that isn't very optimized for HT&L or it isn't triangulized so the true pcount is far higher.
To have the max performance gains possible with HT&L, you have to optimize the models as much as possible:
1-you have to use as less subobjects as possible. Only turretts, rotating objects, destroyable objects, "special" subobjects (like shield, lods, debris..) should be in a different layer than the main geometry
2-you have to use as less textures as possible on the same object, but of any size you want (as long as your card support em).