I am using a server unit with 2 physical processors running BOINC Manager 5.4.11 to run a CPDN climate model. My settings in the ‘General preference’ is currently set to allow up to 4 processors and to use at most 100% of CPU time. But according to windows ‘Task Manager’ the model is only utilizing 50% of the CPUs, in other words probably only using one processor.

How can I set BOINC and/or the CPDN preference so that I can utilize BOTH processors to run at 100% for one model so that I can better utilize my computer’s resources to speed up the run time? Can I allocate multiple processors to work on one task/model? I\'ve checked the CPUs\' affinity setting for the application in question in \'Task Manager\' and both CPUs are checked and allocated for the task, but again, performance is only at 50%.

---

You CANNOT run one model on multiple processors.

At the moment there is a problem with the server software, so that generel preference changes arn\'t being passed back to the user.
If you want to utliise the other 50%, then you need to either attach to another project and set the preferences there, (not SETI which has the same problem), or create a preferences override file, and set the number of processors in that. This explains the file.
Then you can run several models at the same time.

You CANNOT run one model on multiple processors.

At the moment there is a problem with the server software, so that generel preference changes arn\'t being passed back to the user.
If you want to utliise the other 50%, then you need to either attach to another project and set the preferences there, (not SETI which has the same problem), or create a preferences override file, and set the number of processors in that. This explains the file.
Then you can run several models at the same time.

Les, thank you for your prompt reply to my question. I will take your advice and attach to another project.

Just to be sure that I understand the latter part of your advice regarding the creation of a preferences override file; that the script/instruction at the link will allow me to run multiple models simultaneously, thus utilizing more of the CPUs.

I believe that that option is readily available by changing the preference on ‘Your Account’ over the web at climateprediction.net and then clicking ‘Update’ on the Command tab of BOINC.

Perhaps it is probably better for me to attach another project that has shorter time slices while running a single model for climate prediction. Lots of worthy causes out there.

I dread being responsible for running multiple climate models and possibly losing them due to any unforeseen malfunction of my machine (regardless loss of points or credits).
That is why I was hoping to be able to allocate both processors to run 1 model.

I wanted to run a single model as quickly as my machine is able and upload it for analysis incase of any mishaps.

I should put this feature in the ‘Wish List’ if its not already on it.
Thanks again for your help.

I believe that that option is readily available by changing the preference on ‘Your Account’ over the web at climateprediction.net and then clicking ‘Update’ on the Command tab of BOINC.

It is. But it\'s broken.
Which is why I said:

At the moment there is a problem with the server software, so that general preference changes aren\'t being passed back to the user.

With spelling mistakes, I see. :(

Eventually, the server software will be updated, but in the meantime there are two ways around it:

1) Attaching to another project, changing your prefs there, then detaching.
This works because the prefs settings on one project affect that on ALL projects.

2) Creating, with a text editor, the local override file.

Worried about losing your models?
Make regular backups.

---

Backups: Here

Update.

A change was made to the server software a couple of hours ago, which is supposed to have fixed to \'fail to pass on updates\' problem.

My machines have been fine on updating for a while with CPDN. It would not update at a \"update\" request, but when a trickle happened it would update. So it was partially working, just not as clean as it could have been.

Hopefully this is now fixed, and will update when people manually request an update.

---