DFT simulation slowdown after 10k steps

[SAcompuser]

Hi all! So I'm running a water dimer DFT simulation and I'm trying to run a long simulation so I can get sufficient sampling for some properties that I'm trying to calculate, however I have noticed that once I surpass 10 thousand steps (with a time step of 1fs) the speed of the simulation seems to slow down considerably. At the HPC centre where I run the simulation, the first 10 thousand steps take 10 hours, but the 2nd 10 thousand steps takes way longer and my job runs out of allowed time. I have attached my job script. Any advice to fix this issue would be greatly appreciated.

Regards,

SAcompuser 

[Edoardo Aprà]

If you provide more details about this simulation, it is more likely you might be helped.

For example, how many processors and/or nodes have you been using (e.g. provide slurm script). How was NWChem compiled?

How was memory usage during the run? Did it remain stable or was it constantly growing (e.g. evidence of a memory leak)?