How to bring wall time closer to cpu time


Just Got Here
I'm a novice user of NWChem and noticed significant difference between the cpu and wall time of my calculations (e.g., the factor of 2-2.5 for TD DFT calculations with frozen core electrons) even after using scratch directory (which helped a lot). Are there any general tips for settings in NWChem input file to bring the wall time closer to the cpu time?
Thanks!

Forum Vet
The key to bringing cpu and wall time close is to reduce the amount of data written to disk. In the DFT module and input you can reduce disk usage by defining the keywords "direct", "noio" and "grid nodisk", especially if you have a functional with exact exchange. For TDDFT, the calculation tries to store the data in memory, and if it runs out of memory it stores to disk. Here, increasing the global memory per core might help.

Just Got Here
similar way for pspw module?
I have similar situation but I'm using PSPW module. My wall time is about 3.4x logner than CPU time. I tried dft; direct; noio; end in nwpw module but it did not work with error message. Is there any way to direct; noio like effect for PSPW module? Thank you.


Forum >> NWChem's corner >> Running NWChem