Hello JmBoullier,
My understanding is that the throttle on the UD applications works differently than it appears to. If it is set to 60%, then it computes that the System Idle process should be running 40% of the time and keeps cutting until that happens. The throttle is meant to keep a CPU running cool rather than to allocate CPU time.

I may be wrong about how the UD throttle works.

Lawrence

It's called the 'HPF2' factor and is explained by Viktors somewhere. Use Neither UD nor BOINC throttle i.e. both on 100% for a clean view to see the difference with FA@H.

There are discussions about modifying the throttle on BOINC to behave more as users expect (i.e. a finer resolution then 1 second for the suspend/resume logic to give the throttle %) as well as possibly being smarter to subtract its actual processor utilization from what the OS reports as total processor utilization and then adjust so that it doesn't cause the OS to exceed the throttle % unless some other process uses more then that %.

However, this is all discussion at this point. Planning is starting for the BOINC 6.0 client and this is one of many things being discussed. I refer you to the boinc-dev mailing lists if you wish to keep up on changes planned.

retsof,
I have often read you or other CAs mentioning what you describe today more precisely, and I am amazed, particularly today when I see how disastrous your measurements are. You can take my amazement the other way if you prefer, i.e. why is it not that at all in my machine?

Since the beginning (end of January) I am running both, simply because I was quickly convinced by the advantages of Boinc, but being in team Decrypthon it was the least that I did some computing for HCMD.

Since HCMD is finished I am running HPF2 in UD and FAAH (after GC) in Boinc, and UD is producing what I expect from it. It is throttled for crunching one WU per day in order to score something even when the validation of Boinc WUs is going very bad. To achieve that UD is throttled at 25 % and I rarely need to give it a one-hour boost to help it finish before the night stats. Boinc itself is throttled at 100 % usually, sometimes at 90 % if I get fed up by the noise of the fan.

If that can be of any interest for the case, UD behaved the same while Boinc was crunching a few HPF2 WUs these last days because of the FAAH transition.

Obviously I trust you and the figures you give, and I hope you will trust me the same (you can probaly check it in my stats from your side anyway). So what? There is something very strange here...

Cheers. Jean.

My findings are exactly the same as Retsof. It will share 50/50 with both on FAAH but if the UD one is HPF then it only gets 1 or 2%....