What the hack is wrong with the BOINC manager. Just before the recent server shutdown I reattached to the project after a few weeks away. I manager downloaded 2 Hadam3p WU to my single core machine. One is now running, the other is suspended.

I tried to change my preferences so that it would not do the same thing to on my duel core machine. I set it to download Hadsm3mh WU’s and also set it to use only one core, to use only 50% of the processors and to use only 50% of the CPU time. I then clicked update in “Projects”. I figured that this way it wouldn’t send me more than 1 WU. It ignored all the changes and downloaded 4(!) Hadam3p WU’s. Two WU’s are running (26% complete) and 2 are suspended.

The single core machine is running XP, has 1 GB of RAM and is running the BOINC 6.4.7. The duel core machine is running Windows 7 RC with 2 GB of RAM and in running BOINC 6.6.35.

What gives with the extra WU’s? I can run all of them in about 16 days, but, it is a good thing that my preferences weren’t set for Hadcm3s. It could have sent me 2 years work.

P.S. the Hadsm3p’s seem to be running well together. At 7.6 s/TS they don’t seem to be fighting each for memory or slowed down to much.

---

I\'ve found that it\'s best to set the project to No new work before changing prefs.
That way, you can do an Update after the changes, and watch the message tab to check that it does what it should.

One problem at the moment, is that the project is still recovering from the hammering of the 4,000 new members from Facebook.
If you look at the Server status page, you\'ll see that the hadam3p\'s are the only models available.
So it may depend on how you excluded \"other\" models.
The best way is to tick just the one type of model, and have all the others \'unticked\'.
It\'s also necessary to be very careful about venues. If you set up things on one venue, but the computer is set to a different venue, then you WILL have problems.

Then there\'s the matter of a model coming to an end, and being in the connected to the Internet about every, or, Maintain enough work for an additional, range. This can result in a \"spare\" data set being download ready to start when the \"near\" model finishes.
This is made worse when you run several projects, as only BOINC will know which project it\'s decided should have a turn on a processor.

Did you change your preferences from within BOINC Mgr or on the website? If you used the website, then you have to click the update button first before fetching more work. Otherwise, when it connects, it still requests as much work as the previous preferences thinks it can handle.

FYI, setting 50% processors and 50% time equals 25% of your machine. 50% processors and 100% cpu time means it will only run one core (on a dual core machine).

Edit: Yes, previous poster is correct. You should have settings such as this in your Computing Preferences:
Computer is connected to the Internet about every = 1 days (or 2 days)
Maintain enough work for an additional = 0 days

When you change your preferences in your CPDN account (on its web pages) and then click Update in Boinc Manager, the update only works if the server receives this command from you before it times out. In my experience if your update command times out you receive no indication/message to tell you. But if you check your account you see that your preferences have reverted to what they were before.

Before the server upgrade the old server was running so slowly that we know some members\' preference changes did not take effect. The same thing may have happened on 11 August while the new server was struggling with the massive demand for models from the influx of new Progress Thru Processors members.

In addition, I believe that by default, if your chosen model type isn\'t available the server sends you a model type it has got.

I\'m going to post about this in the News thread now. Thanks for your report, Jim.

You\'ll probably find that if you change your preferences again, the update will now take effect.

---

Cpdn news

Hi, everyone:

Thanks for the advise. I think that I have found the problem. The “maintain enough work for” box was set at 10 days. Ten days is longer than one of the Hadam3p models is likely to take to finish, so it sent doubles for each core. With the longer models it didn’t matter. I’ve now set it for one day. The “Computer is connected to the internet about every” is set at 0 because it is always connected.

I did change the preferences from within the manager.

---