Smoke Particle Effect

The current particle smoke calculations are performed on a single processor and limited to 2000 particles per smoke instance.  After 10,000 particles are on screen the performance starts to decrease much below 30 fps. I'm currently working on implementing a 3D CUDA function that will perform the smoke step operation, but I'm not getting proper data out of it yet:

pos.x += speed.x * lifetime/interval
pos.y += speed.y * interval
pos.z += speed.z * lifetime/interval

and then update the lifetime based on the fade/mass value.

I currently have a few GLUI controls set up for changing parameters of the smoke. They are fade/mass, size and colour. Below are some screenshots with different settings:

I'm hoping for the ability to render up to 30,000 particles (or more) with CUDA step operation compared to the current frame-rate using single instructions. There are also many more GLUI controls that could be added to change the affect of gravity on the smoke and its speed.