Parallel execution strategies

Dear all,

I am using siesta 4.0 for calculating zeolites properties with pbc. Zeolite unit-cell consists of 288 atoms.
I tried to run in parallel upto 32 cores. Speed of calculation dis not improve considerably.
By using K point parallelization along with the number of ProcessorY increased the speeds considerably.
However, the observed increased in speed plateaus for 4 ProcessorY and 16 cores. with further increase in cores or process there is no improvement but marginal decrease.

I believe there would be a better parallel strategies for this problem.
Kindly help me to effectively use the parallel strategies of siesta.

Thanks and regards
Janardhan