EnergyPlus 8.6 taking 9 times longer than EnergyPlus 8.5

asked 2017-01-12 12:14:20 -0600

updated 2017-01-12 12:14:58 -0600

Having updated some of my osm and idf files from OS 1.11/1.13 and EP 8.5 to 8.6, I have observed with dismay that with updated files simulation time is considerably longer. E.g.: a model that used to take appr. 12 minutes for a 1-year simulation under EP 8.5 now after simple update to EP 8.6 needs 1 h 51 minutes (!) . The simulation results are similar but the time consumption has exploded. This problem also occurs when using Openstudio 1.14.

Are there other users who have observed similar worrying deterioration with EP 8.6 ?
Is there anything that can be done - keeping the same accuracy levels - to achieve similar simulation times as with EP 8.5. ?
What can developers say to explain this strange behavior ? Are there plans to investigate possible weaknesses in new algorithms ?

Technical details of my transition:

EP 8.5 under Windows 7 (2 CPUs), 32 bit;
EP 8.6 (64 bit stand-alone, 32 bit bundled with OpenStudio) under Windows 10 (4 CPUs) 64 bit;

edit retag flag offensive close merge delete

Comments

I'll let one of the developers chime in, but E+ releases are extensively regression tested to make sure results and run times don't change appreciably. I'd be very surprised if that kind of performance degradation was in any way typical.

ljbrackney ( 2017-01-12 12:26:39 -0600 )edit

Please provide the input file so we may test to confirm. You can find my email in my user profile.

rraustad ( 2017-01-12 12:41:23 -0600 )edit

I'll send you both the 8.5 and 8.6 version files.

OS-user-AT ( 2017-01-12 12:49:51 -0600 )edit

Probably not the cause, but you may want to make sure your project or IDF file is not on a network drive (separate from the application). In the past that would have a big impact on EnergyPlus runs, but not sure if that is still valid.

David Goldwasser ( 2017-01-12 13:42:18 -0600 )edit

there wasn't any network connection, everything only on local harddisks, computer operated offline. The input files have already been sent to R. Raustad

OS-user-AT ( 2017-01-12 13:55:02 -0600 )edit

add a comment

Program Version,EnergyPlus, Version 8.5.0-c87e61b44b, YMD=2017.01.12 15:11,IDD_Version 8.5.0 ************* EnergyPlus Completed Successfully-- 626 Warning; 486 Severe Errors; Elapsed Time=00hr 08min 45.14sec

Comments

Could you perhaps offer a quickfix for download (?), because waiting for the next official release could take months (E.g.: Due to another issue with crashing simulations I had to stop work for 3 months in 2016)

OS-user-AT ( 2017-01-13 01:22:16 -0600 )edit

Others are looking into this now. Let's see if they can track it down.

rraustad ( 2017-01-13 09:49:23 -0600 )edit

It appears this slow down has to do with the use of dynamic shading surfaces. An issue has been filed for the development team to look into it (https://github.com/NREL/EnergyPlus/is...). But it may turn out that this slow down is expected after a bug fix that corrected diffuse sky modeling with varying transmittance shading surfaces (and now that the onerous calculations are being done right they take longer).