What your results?

My ray-tracing algorithm works:
750 000 ray-triangle intersections per second on the Celeron A 366 Mhz with MMX and 128 Kb cache memory.
Can I get higher number?
How much fast works your algorithm?
Thank you for attention.

I think I was getting about 3 000 000 tests per sec. I dont really remember anymore and should review that stuff.