LHC@home

What does the max # CPUs mean in the preferences? Is it the max number of CPUs used per vbox session (per task) or max number of CPUs for whole virtual box environment?

---

This means max CPU's for the virtualbox instance. The max number of jobs is how many physical cores will be used

What does the max # CPUs mean in the preferences? Is it the max number of CPUs used per vbox session (per task) or max number of CPUs for whole virtual box environment?

The "Max # CPUs" setting controls <avg_ncpus>x</avg_ncpus>.
If you don´t use an app_config.xml the value x tells BOINC how many CPU cores the WU needs and the wrapper starts a VBox VM with x CPU cores.

You can overwrite both settings if you use an app_config.xml.
<avg_ncpus>y</avg_ncpus> is used by BOINC
<cmdline>--nthreads z</cmdline> is used to configure the VBox VM

The values for y and z can be different but not every combination makes sense.

Thank you for your replies. So because I want each vbox task to use just one CPU core I should set max # CPUs to 1 and set the max # jobs to a value how many task I want to run concurrently (at the maximum).

Is the max # jobs also the limit of how many tasks are downloaded from the server? This would disable the possibility to have any meaningful cache for the sixtrack application. The vbox tasks seem to take much longer than sixtrack tasks.

I am also seeing that the CPU load is fluctuating a lot on BoincTasks. I have 4 vbox tasks running (3 Theory + 1 CMS) and only two VBoxHeadless.exe programs are using any CPU time in Windows Task Manager (there are totally 12 VBoxHeadless.exe programs running). All 4 tasks show progress in BoincTasks (slow but steady). I have to wait and see if they all finish.

---

Yes that how it works.

No that's control by the boinc setting for cache, under "other" in settings

CMS and LHCb taks can be a bit light on work depending on hw queues are filled,