Rack Workstation slower than Laptop ??
Awaiting user input
hello all,
I have a question regarding the solution speed. My normal laptop (fairly new) with 12 cores solves the problem twice as fast as a rack workstation (3 years old) with 64 cores.
I would expect the workstation with a lot of processing power to be at least on par, if not much faster.
I called the programme with Python from the command line.
What could be the reason for this? Can something be set wrong?
Best regrads,
Florian
-
Hi Florian,
Could you please share the logs of both runs?
Best regards,
Jaromił0 -
Hi Jaromil,
sure.
Here is the worksttation:
And this is the laptop
The last row is the time that was needed for a one year simulation in seconds. The log shows only one day-
Best regards,
Florian0 -
Hi Florian,
There are a few things which may explain the difference in performance.
- On your laptop, you are using v9.5.2 while on the workstation you are using v10.0.2.
This results in a different presolved model and a different solution path.
- The model is solved very quickly in the root node and Barrier was not necessary for the root node relaxation. Since Barrier and the B&B algorithm profit the most from many cores, this advantage of the workstation is strongly diminished.
- You are saying that your laptop is fairly new. Do you know whether the processor in your laptop are stronger (better performant) than the ones built in the workstation?
- You can see that at least for this quickly solved model, the ratio of work units to seconds is better for your laptop (0.23/0.28 > 0.28/0.4) which is an indicator that indeed your laptop is performing better than the old workstation.
Are all models solved that quickly or do you have a log of a model that shows a significant solution time difference? It might be possible that the time spent between the solves of models also plays a role.
0 - On your laptop, you are using v9.5.2 while on the workstation you are using v10.0.2.
Please sign in to leave a comment.
Comments
3 comments