topic Nastran WALLCLOCK TIME less than TOTAL CPU TIME? in Simulation Mechanical Forums (Read-Only)

Nastran WALLCLOCK TIME less than TOTAL CPU TIME?

bryan_mcnatt — Fri, 10 Feb 2017 17:02:40 GMT

How can the Nastran reported WALLCLOCK TIME ever be greater than the reported TOTAL CPU TIME? My understanding was that wallclock time includes the CPU time plus other time spent reading, writing, waiting, etc. Are the two times distinct and should be added for the total time to run an analysis? Or, is there some other more complicated explanation of what CPU time and wallclock time are?

I'm running a static stress, linear model and my CPU time is reported to be 723.8 sec and my wallclock time is reported to be 538.3 sec.

???

Thanks!

Re: Nastran WALLCLOCK TIME less than TOTAL CPU TIME?

John_Holtz — Fri, 10 Feb 2017 22:13:38 GMT

Hi Bryan,

The clock on the wall has no idea what the CPUs in your computer are doing, so the WALLCLOCKTIME is just the elapsed time that it takes to run the analysis.

Of course, your computer is not connected to the clock on the wall, but they should agree pretty close to each other unless the computer clock is affected by the heavy computations.

I would guess that the CPU time and Wall time can be either greater than or less than the other. For example,

If the analysis takes 5 minutes to run (wall time = 5 min), and 6 processors are used at 100% capacity, the CPU time would be 30 minutes. (CPU time > WALL time)
If the analysis takes 5 minutes, and 2 processors are used at 25% capacity, the CPU time would be 2.5 minutes. (CPU time < WALL time).

Re: Nastran WALLCLOCK TIME less than TOTAL CPU TIME?

bryan_mcnatt — Fri, 10 Feb 2017 22:59:48 GMT

That was all I could come up with myself, that the two were distinct and that CPU time must be the sum of all my CPUs.

Thanks John.