Please read this thread starting from an explanatory post by Rick Alther and concurrence that HPF2 on UD is not getting it's fair share when running together with BOINC. The same applied to HCMD.

http://www.worldcommunitygrid.org/forums/wcg/viewthread?thread=10210#79444

There are a few things one can do to speed the HPF2-UD combo up somewhat, but 50/50 will only be achieved playing the multi thread affinity settings or some rigorous 3rd party throttle tool registry tweaks.

Personally, the UD-BOINC concurrent running was never meant and given the former is planned to be phased out, not expecting any developers effort to remove the problem.

Thank you all for trying to clarify these differences. With your contribution and hints I think I begin to see an explanation.

@Didactylos
I have a P4 HT, and I think that is the biggest difference with cases where UD cannot find room to work.
Because XP is not able to report processor activity correctly with these processors (wrongly reports 50 % inactive if running a single task) UD "thinks" it may run. So it does it...

@Sekerob
Thank you for this pointer. It helped me to think about what the throttle effect is. However I am sorry for Rick but I think facts show he is wrong on this particular point:

I said that "my" UD is usually throttled at 25 %. If Rick is right it should never run since Windows says Boinc is already using 50 %. But if the UD throttle effect is actually meaning "do not use more than 25 %" it's all fine, and Taskmanager says UD takes 25 %... of the inactive 50 %!
So it seems that the UD throttle means "Do not use more than xx % of the available processor resource". And remembering cases where I had to boost UD to match the stats schedule UD has worked as expected whether I boosted at 50% or 100 %.
Obviously again all this is working for me only because of the wrong processor activity info delivered by Windows for HT processors.

Which leads me to my last comment, to retsof this time.

That would be too good to be true, unfortunately. All of us know that an HT proc is not more than a single processor. So I have not really produced the 0:001:18:38:35 per day that my stats show. I have only produced 75 % more statistical units (time or points) than I should, and research has not progressed any faster during that time.

Thanks again to all of you. Jean.

Both BOINC and UD are incrementing wall clock seconds correctly in real time, even though each only has 50% of the CPU. That could be where the points inflation is coming from.

Never mind. Today is an even day. Yesterday was an odd day. Huh!

Yes. Mine is doing something similar this morning.

Hi Troy Boy,
The BOINC throttle is a joke, admittedly. The solution is to leave BOINC at 100% and use ThreadMaster to control the applications.

The general problem is that throttling should be an OS level routine. Programming it into an application is difficult and wrong-headed. IBM developed a good throttling system to use with UD applications, but it is difficult to implement. I do not know when BOINC will improve its throttle.

You have hit on my pet peeve. I originally argued against putting a throttle into the UD applications, because it approached the problem from the wrong end. Throttles should work with the OS rather than with a specific application. I have been surprised at how well the UD throttle has worked out, with only a few problems.

Lawrence