Setting stepspercycle to 1 would result in big-time performance loss at no benefit to the stability of the simulation.  Put simply, this setting indicates the neighbor searching interval.  See, for instance:

http://www.ks.uiuc.edu/Research/namd/wiki/?NamdPerformanceTuning

and

http://www.ks.uiuc.edu/Research/namd/2.6/olddocs/ug/node24.html#section:electdesc

The first link provides good advice; leave the "typical" values of stepspercycle = 20 and parlistspercycle = 2, you get a neighbor list update every 10 steps, which is pretty normal.

Hmm,

Thank You for your reply! but if it's possible could you give me a primitive anwser, what is the difference between the timesteps in one cycle? if the system doesn't move than everything should more or less still be the same, until the next cycle, or am i much wrong?

Well, everything moves - it's a dynamics simulation.  The point is that atoms don't move especially far over very short time periods (e.g. 10-20 fs), so there is no reason to update the neighbor list every step, because it's wasted time - the neighbors don't change.  Over longer times, molecules move in and out of the neighbor list range.  So the settings in NAMD basically say "how many steps can elapse before the neighbor list is re-populated."  So within one cycle (i.e. before the neighbor list is repopulated), the lists of atoms experiencing short-range interactions don't change.  That doesn't mean the forces don't change, of course, because atoms move around.

In NAMD one cuts the whole space into small cubes to speed up the calculation.  For atoms in the same cube (patch) and the nearest cubes, they are considered as "neighbors" and the non-bounded force for each atom is calculated with its neighbors.  Like Justin said, the molecules can move outside the cubes after some time and the neighbor list needs to be updated.