Hello gdlxn,
This is not a problem unless your Results Status page starts showing errors / invalids. If you are interested in research on this problem, start by searching for finished in Start Here.

Lawrence

Hi gdlxn,
Here is the FAQ: https://secure.worldcommunitygrid.org/forums/...ead,16784_offset,0#133103

What you need is to keep the work unit running continuously until it finishes. The first thing to do is to work on the profile. Select My Grid - Device Manager - (Selected Profile). If it is default, changing that could be confusing. I use 'home' as my profile. First I checked 'Maximum Output', then I selected 'Custom Profile' to allow me to change those settings. I checked 'Set as Default'. I set everything to run full blast, no GPU, only 1 CEP project at a time, no HCC since the GPU computers are much more efficient than my CPU and under the Memory Usage section I have 'Leave applications in memory while suspended? YES'. Hit SAVE at the bottom. There are various ways to make BOINC load your profile, but reboot always works.

Once you are running on your new profile, check Task Manager to make sure that your BOINC projects are using nearly 100% of your CPU without anything interfering. This should do the trick.

Lawrence

Hi gdlxn,
Glad to hear that you are making progress in tracking down the problem. CPU Usage is in your profile. You get to choose a limit past which BOINC shuts down. Naturally, my limit is 100.0%. Why would I want my CPU to ever go idle? CEP2 is a very demanding program that can can fill up a long queue of I/O operations for the hard disk. From caution, I only run one CEP2 work unit at a time even though any fast computer should be able to run 2 at a time and some people boast of running 4 or even 8 on their speed demons. But we added the control statement to the profile when we first started Clean Energy and quickly ran into trouble on normal computers.

The Leave Application in Memory is vital since anything that stops the project, like a cpu limit in the profile, will otherwise force CEP to start over at a checkpoint and lose as many as 3 hours computing since there are only 15 check points, some of them hours apart.

Lawrence

Summarizing previous advice in this thread and expanding:

1) Revert to setting 1 CEP2 task at the time. With clients prior to 7.0.39 that can only be done through the web-device profile. With 7.0.39 and up, it can be done by setting a <max_concurrent> in a user creatable app_config.xml project file.
2) Set Leave application in memory when suspended to On / tick mark.
3) Set in Local/Computing Preferences the option "While processor usage is less than 25%" (Experiment the percent up or down to see when heartbeat/zero status issues stops).
4) Set in BOINC Manager > Activity menu the CPU to Run based on Preferences.

These settings work for me, on all my devices.

CEP2 is *not* recommended to run concurrent for more than half of the processors, meaning if you have a quad, not run more than 2 simultaneous. Individual devices can be tuned up or down based on their typical use. One where your working on can run less without you noticing or getting problems, than devices that are not used for anything but crunching, or a little file serving.

Since I can run 8 concurrent on my octo-hyperthreaded with 98% efficiency or better, but only when absolutely left alone, nothing else running, no user input, there's little doubt it's the individual device setup and utilization and capability [for instance slow CPU, shortage of RAM, too much memory ot disk swapping]. For sure, since we know CEP2 a tough cooky to run, the project is explicitly Opt-in. Since we do not want any member to be impaired by this/or any WCG project running, or frequently producing invalid/error results or results that restart many times, thus run very inefficient, the advice is indeed to opt-out of this science.

Let us know if the points 1-4 do not resolve the matter, even for running 1 at the time.