Check your global preferences on web-page, make sure "On multiprocessors, at most use: N processors" has N => 4, and make sure "On multiprocessors, use: 100% of processors" is 100% and not something lower.

After saving this on web-page, hit "update" in the BOINC Manager.

It's also possible you're using local preferences instead, if so either remove these after making sure web-preferences are correct, or edit the local preferences, the Processor-tab, and similar to web-page make sure it's set to 100%. Local preferences is accessible in Advanced View by selecting menu-option Advanced - Preferences.

I am now a bit confused, because on

WCG i get a version 6.2.28 which uses only 2 cores

on

BOINC i get version 6.6.38 which uses all 4 cores !?!

I reinstalled the BOINC version without changing any setting and it works again.

Could it be that WCG suggests a very old version ?

Regards,

Martin

PS: I have a icon which looks like a heart. Is this the latest icon, previous icon was orange, circle or so...

PS, the Blueish V shape Icon is the custom icon of the WCG skinned 6.2.28 version. The orangy wheel icon is the BOINC original

Hello Capiman,
WCG performs some debugging, consolidation and improvement works on boinc before releasing it. Berkeley releases much more boinc versions than WCG is able to "finalize". Because grid computing projects should not disturb volunteers, WCG is really careful by recommending a new boinc version.
For this reason, WCG members should use - if possible - the boinc version released by WCG.
For my-self, I did not upgrade since version 5.10.45 and I am quiet happy with this boinc version, not experiencing any significant problems and in particular without any crash.
Cheers,
Yves

The number of cores is quasi unlimited Sunnyboy. 32-96 cores have been spotted at BOINCstats.

WCG does more rigorous testing, 6.2.28 mostly focused on getting the proxy for all, a major issue for the partners and since there are those with thousands of devices, WCG makes very sure the standard and mass-rollout package does it right for them in particular... thus the automatic reluctance for the bi-fortnightly release policy.

In 6.10 there should be more proxy automation and some improved scheduling... well not really, I reconfirmed a 6.6 bug yesterday, an ugly bug of which finally was said "I'll pass it to the developers". Just follow out 6.10 alpha test thread that the techs are following too.

This 2 core / 50% conflict is due poor implementation for the upgraders hence the text in the website device profile [nobody reads] and the poorness of the local prefs having removed the option hard core number option, many baffled for the rule works 'lowest of both and rounded down' i.e. 99% on a quad is dealt with as 75%, but at the same time 99% on a 96 core device works as 95 cores, which is meaningful.

The WCG package has a function to only notify if WCG upgrades, not Berkeley. Sadly that part only works if there was no pre-existing cc_config.xml on first 6.2.28 install as the WCG kit puts in it's own with special lines. Hope that works more transparent when 6.10 is adopted.