Optimizing solver-opt files for stochastic runs

[LucasRM]

Dear forum members,

I am currently running a few stochastic models, and I was wondering if there is any optimal/good solver configuration for stochastic models in Veda? I have so far been using the CPLEX solver with solver-option BA121-7. 

I am using a laptop with an AMD Ryzen 7 5800U processor with 8-core/16 thread, and I can see in my SolverOptFiles that I have the thread count set to 7. I assume setting it to -1, leaving only one core free, would make the program run faster.

Does anyone have any input on this or any suggestions for other things I could change?
If I need to give more information let me know. Thank you!

[MakRacz]

This is a very nice presentation about cplex options in TIMES https://iea-etsap.org/webinar/CPLEX%20op...models.pdf

Multithreading lowers cpu clocks speed, so using more threads doesn't necessarily improve solution time. In my personal experience, one should never exceed the number of physical cores.

[LucasRM]

This is a very nice presentation about cplex options in TIMES https://iea-etsap.org/webinar/CPLEX%20op...models.pdf

Just popping in here again to say thank you, this presentation was super helpful, and helped in reducing my runtimes 

[VictorG]

This is a very nice presentation about cplex options in TIMES https://iea-etsap.org/webinar/CPLEX%20op...models.pdf

Just popping in here again to say thank you, this presentation was super helpful, and helped in reducing my runtimes 

Could you please briefly explain which specific actions reduced computation time ?

[LucasRM]

Could you please briefly explain which specific actions reduced computation time ?

For me setting predual = -1, in the cplex.opt file cut my runtime down. In one of my model cases it went from 2 hours to 45 min. This is likely due to the fact that my model contains a lot of variables compared to constraints. 
Apart from that my relevant settings are: 
barorder 0 (also recommended)
baralg 1 (The fastest would likely be 3, but I have kept 1 for the solution quality)

I have not used barcolnz yet, and haven't had time to go into depth with the other options either.