Are HOS slow ?

I’m using a GF3 and soon will switch to a GF4.
I have a static mesh made up of 10,000 and a single triangle patch divided in 10,000 triangle usig nv_evaluators, with static control point. For the mesh i’m using vertex_array_range. Does it make sense that the mesh gets a bigger frame rate?
Shouldn’t hos be usefull to save bandwidth a get higher frame rates with the same number of triangles?

In general, HOS will only improve your speed over a staic mesh if you are limited by vertex bandwidth. The evaluation of the surface takes time and resources on the chip. There is a reasonably good chance that these are not completely independent from the rest of the T&L. This means that you are likely trading off some of your peak transform to do the HOS. In consumer apps (games) this rarely matters as most apps are fill limited rather than geometry limited anyway, essentially making the HOS free.