correlate wus on server status page to GPU types

Forums › Cruncher's Corner

Anonymous
24 Sep 2018 14:01:48 UTC
Topic 216440
(moderation:
    )

    when I look at the WUs available on the server status page how do I tell which WUs are available for AMD, NVIDIA GPUs etc.  right now I have suddenly stopped processing AMD GPU WUs but my interpretation of the server status page indicates that there are WUs available for AMD GPUs.  

    solling2
    solling2
    Joined: 20 Nov 14
    Posts: 219
    Credit: 1577671300
    RAC: 21403

    AFAIK, the server doesn't

    24 Sep 2018 16:54:06 UTC
    Message 166957
    (moderation:
      )

      AFAIK, the server doesn't even have to differentiate between different GPU brands as it will send text files with data (.dat) only. An operating system specific application for how the GPU has to handle that already sits on your system since project installation (or publication of a new one). Then the driver arranges the run. So the server stats page just lists overall numbers. If my assessment is right, you may possibly have changed your settings accidentally? 

      Holmis
      Joined: 4 Jan 05
      Posts: 1118
      Credit: 1055935564
      RAC: 0

      robl wrote:when I look at the

      24 Sep 2018 17:33:18 UTC
      Message 166958
      (moderation:
        )

        robl wrote:
        when I look at the WUs available on the server status page how do I tell which WUs are available for AMD, NVIDIA GPUs etc.  right now I have suddenly stopped processing AMD GPU WUs but my interpretation of the server status page indicates that there are WUs available for AMD GPUs.  

        The direct answer would be something like 'Compare the server status page and the Applications page' and that's best done if one have the translation key between the application name and it's abbreviation shown on the server status page.

        A better way of going about finding out the problem would be to ask oneself what's changed?
        A good starting point would be the last server contact log for the computer in question, in this case host #12619116.
        The the log can be viewed by going to one's list of computers and clicking on the last contact time for the host.

        From the above linked log one can ascertain the reason to why no work for the GPU is sent: (trimmed a bit for readability.)

        2018-09-24 16:26:43.1224 [version] parsed project prefs setting 'gpu_util_fgrp': 0.330000
2018-09-24 16:26:43.1224 [version] OpenCL GPU RAM required min: 803209216.000000, supplied: 51097600
2018-09-24 16:26:43.1225 [version] Checking plan class 'FGRPopencl1K-ati'
2018-09-24 16:26:43.1225 [version] parsed project prefs setting 'gpu_util_fgrp': 0.330000
2018-09-24 16:26:43.1225 [version] OpenCL GPU RAM required min: 1048576000.000000, supplied: 51097600

        This part of the log indicates that Boinc reports the GPU of having 48 MB of RAM. (That's not a typo!)
        The list of computers claims the same.
        I would start with either updating or reinstalling the GPU driver to see if this corrects the memory detection fault.

        Gary Roberts
        Gary Roberts
        Moderator
        Joined: 9 Feb 05
        Posts: 5872
        Credit: 117785938597
        RAC: 34670864

        The last time your computer

        24 Sep 2018 21:57:52 UTC
        Message 166963
        (moderation:
          )

          The last time your computer fetched GPU work was on 13th September.  Perhaps there was an update for your computer just after that time that has caused problems.  Definitely looks like a driver/OpenCL libs issue.  The BOINC startup messages might contain a clue, particularly if you note any difference compared to previous startup messages.  You may be able to find an older message in stdoutdae.old if there isn't one in stdoutdae.txt.

           

          Cheers,
          Gary.

          Anonymous

          Gary Roberts wrote:The last

          25 Sep 2018 9:38:00 UTC
          Message 166966 in response to message 166963
          (moderation:
            )

            Gary Roberts wrote:

            The last time your computer fetched GPU work was on 13th September.  Perhaps there was an update for your computer just after that time that has caused problems.  Definitely looks like a driver/OpenCL libs issue.  The BOINC startup messages might contain a clue, particularly if you note any difference compared to previous startup messages.  You may be able to find an older message in stdoutdae.old if there isn't one in stdoutdae.txt.

             

            Late today I thought that maybe the problem was driver related (I had not noted your driver comment above Gary) so I rebooted the PC and in came the GPU jobs.  I am not sure why this happened.  I am running ubuntu 16 and have not done any updates lately.  Anyway it is working once again after 6 lost days of crunching.

            Thanks to all who replied.  

            Comment viewing options

            Select your preferred way to display the comments and click "Save settings" to activate your changes.

            Forums › Cruncher's Corner