Messages

Just a few hours later Windows installed a patch for 1809 update and now everything seems to work OK. I did install the lastest beta version of BoincTasks before rebooting the computer after Windows update. So if new problems arise I'm ready for it.

OK, spring is here and Windows decided that now is the time to install the 1809 update. And again BoincTasks (1.79) is unable to start and crashes 10 times before giving up. Running as an administrator did not help.

I installed the 1.79 version, but it failed like 1.78. No user interface becomes visible (and no icon on taskbar's notification area) and after 10 crash dumps the program terminates itself.

I believe that I ended up accidentally installing win10 updates that are still in testing phase. So I decided to return to the previous version of windows 10 and now 1.79 started without a problem (after I had removed the 10 crash dump files). Now windows is again downloading new updates but this time I try not to do anything to install "extra" updates like I did yesterday. It seems that if you go to windows update and manually check for updates, you may end up with updates that haven't had final acceptance yet. This of course is not mentioned on the update window  .

If the failure happens again I will report it here.

Hi,
My Windows 10 host just installed the feared 1809 update. Now BoincTasks 1.78 can not start but it creates 10 crash dump files before giving up. The dump files vary in size between 140 to 155 kbytes. I tried to install the 1.78 version again but to no avail. Should I try to install the 1.79 beta version next or what?

Just to verify my finding: I changed the CPU usage in app_config.xml for Seti and Einstein to 1 (from 1.1 or 0.99) for GPU tasks and I get the long version of the Use column although "condensed" is selected in preferences.

I was using the 1.1 value, or sometimes the 0.99 value, just to get the short version text on the Use column. I have set the Boinc preferences to 75% of CPUs (so 6 out of 8 cores), but sometimes Boinc doesn't follow this setting. I believe that happens when Boinc is switching between Seti and Einstein tasks, I run 1 Seti task per GPU (app_config= 1 & 1) and 2 Einstein tasks per GPU (app_config= 1 & 0.5). I will change now the 1.1 and 0.99 to 1 in my app_config.xml files to see if this makes things better.

Thanks for your reply, I had completely missed that option. That was already selected, but it doesn't seem to work correctly (BoincTasks 1.73).

I originally had cpu set to 1.1 and gpu to 0.5 in app_config.xml (Einstein@home) and the use column displayed short values. Then I edited my app_config.xml and set <avg_ncpus>1</avg_ncpus> and <ngpus>0.5</ngpus> and restarted Boinc. Now use column is displayed as long although the option your mentioned was selected in both cases.

Another oddity that is related to the CPU usage values: The CPU usage % number is calculated in BoincTasks differently if your CPU setting is 1 or 1.1 for a GPU task. The Einstein tasks require a complete CPU to support one Nvidia GPU task. If I set CPU usage to 1 or 0.99 the % value shows about 98.5% which is correct. If I set it to 1.1 the % value shows about 48.5% which is wrong and means that the 1.1 is rounded up to 2 and used in the calculation of the % value.

On a CPU task (LHC Atlas VirtualBox task) that is set to use 3 CPU cores always shows 100% usage after the initialization of the task is over.

The default way of showing a long string when use CPU/GPU is defined as 1 is annoyingly long [1 CPU + 1 Nvidia GPU (device 1)]. If your app_config.xml defines a decimal number (like 0.9) the string is shortened to nice value [0.9C + 0.9NV (d1)]. I think this should be an user selectable formatting (short/long form).

Now the way the Use column displays it's values makes the BoincTasks window unnecessary wide and I am running out of space on my display although I have 2560x1440 display.

The removing of the 'smart mode' seems to have done the trick. I haven't seen big numbers of short tasks until today: 9 pieces of 4 second tasks one after another and all were catched by BoincTasks.

The short SETI tasks probably don't overload the BOINC client buffer, normally there are only a few of them.
It may be a LHC bug, it makes no sense to produce tasks so small.

I think that LHC would like also not to produce such a short tasks, but as the nature of the tasks is what it is, it may be unavoidable. The sixtrack application is simulating the control parameters for the superconductor magnets that control the particle beam in the collider. The idea is to weed out those parameter sets that will make the beam collide to the wall of LHC. So in a weird kind of way a short task is also a success.

I will try that. Currently the short tasks are over at LHC but we'll see if it helps.

There are sometimes short tasks at Seti too, about 12...20 seconds on the GPU. They are actually so short that they probably doesn't get to the GPU but fail before that phase. At least BoincTasks doesn't tell me the GPU number on those tasks (I have two GTX970s on one host). But Seti hasn't had so many short ones in a row as LHC had. I haven't noticed Seti tasks being missing, but I'll keep my eye on them as well.

Hi
Some of my finished tasks do not register to the history tab. LHC@home sixtrack applications are currently having quite a lot of tasks that are finished in 6...9 seconds. Task do not have errors, they finish succesfully. Today I noticed that that I had 29 such tasks finished in succession (visible in project website) but only 7 of them are visible in History. I tried to shorten the min and max update times to min=1 and max=6 seconds but that did not help.
Boinctasks version is 1.73, Boinc is 7.6.33 on windows 7 64 bit. So is this a bug or what gives?