Hey! which processor do you have ? parallel processing only helps while solving (iterations). In my case, the parallel processing (upto 8) saves a lot of time while solving. Its not the i7, its the core. first of all find out how many cores does your processor has, then find out how many threads are there. set the parallel processing to 8 if you have 8 threads. But in order to use parallel processing , you first need to setup mpi. 

It is recommended  use  partitions if any partition has around 150-200K nodes. It is because  the data transfer between scores can spend more time to calculate the subsystem. If is a local run, with a good L2 Cache you can be not affected by data transfer. A good FSB velocity and RAM memory is good also.  If is a distributed run, your infiniband must be adequate to data transfer, if not your scalability will decrease.

A good setup with a good convergence rate is mandatory to speed up your run. 

I never used all threads because is good to have at least one core to other functions and avoid your computer to freeze. But it is case dependent. There is no a general recipe. 

Some suggestions..

Use Parallel processing up to 24 parallel processor..

Increase RAM up to 32G or more..

Use PC cluster

Use Linux system

Try to reduce number of mesh cell size as low as possible.

Use appropriate time step size, early convergence leads your simulation more quick.

Good Luck !

M. F. Ali