Discussion Thread for the Continuous GW Search known as O2MD1 (now O2MDF - GPUs only)

Forums › Cruncher's Corner

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 4965
Credit: 18756604525
RAC: 7160205

mikey wrote:Keith Myers

23 Apr 2020 21:49:09 UTC
Message 177051 in response to message 177048
(moderation:
    )

    mikey wrote:
    Keith Myers wrote:
    Yes my pending list is predominantly those Velajr1 task awaiting a wingman.

    So is that these tasks?

    Gravitational Wave search O2 Multi-Directional GPU v2.07 () x86_64-pc-linux-gnu

    Yes, one host is getting those mostly along with the FGRPB1G tasks.  The other host is getting mostly the FGRPB1G tasks and a smattering of GW 2.07 gpu task.

    This host has a lot of the Velajr1 tasks in Pending and awaiting a wingman with a 1 Jan 1970 sent date.

    https://einsteinathome.org/host/12444941/tasks/2/54

    Or it did earlier.  Looks like I have a wingmen for those tasks finally.  The scheduler must have crawled through the pendings and sent out for wingmen since my last post.

     

    Mr P Hucker
    Mr P Hucker
    Joined: 12 Aug 06
    Posts: 838
    Credit: 519325953
    RAC: 14237

    mikey wrote:Peter Hucker

    23 Apr 2020 22:05:59 UTC
    Message 177052 in response to message 177047
    (moderation:
      )

      mikey wrote:
      Peter Hucker wrote:
      mikey wrote:
      !! I'm at the point now where I'm trying to figure out which Project to run but have too many choices to make!! I have 15pc's running now and some can run this  Project but not that Project while others can't  run either of those. I'm also on a Team and trying not to pass my Teammates, just because I can, so that adds into things too. I will pass them if they quit a Project but not if  they are actively crunching it, I believe in the axiom 'just because you can doesn't mean you should'.

      Agreed, apart from the teammate thing, I don't get why you wouldn't want to pass them.  A team in formula 1 could have one driver passing another.  If one can win the race and beat other teams, then they have to pass the rest of their own team first.

      Oh I could easily pass any of them but that's not being a good TEAMmate, they earned their spot at that Project and why should I knock them off it just because I can. One teammate had over 20 million at one project while I had 3.5 million, 3 weeks later I had over 10 million and stopped. I did pass another teammate at another project during a challenge but then stopped and let him pas me back again after it was over, I passed him in under 4 days. At Collatz if I put every gpu I  have on it I get close to to 60 million credits per day.

      PS i also like F1!!!

      Well in most races you can help your teammates, but there's no way you can do that here.  So apart from stopping processing for a bit (which isn't good for the science), I don't see how you can not pass them.  Unless you just work for another project?

      Anyway isn't the point of a team to achieve as much as you can collectively?  So what if you do more than them?  If you really want to help them out, post one of your GPUs to them :-)

      If this page takes an hour to load, reduce posts per page to 20 in your settings, then the tinpot 486 Einstein uses can handle it.

      Mr P Hucker
      Mr P Hucker
      Joined: 12 Aug 06
      Posts: 838
      Credit: 519325953
      RAC: 14237

      Zalster wrote:Yup, those are

      23 Apr 2020 22:07:36 UTC
      Message 177053 in response to message 177050
      (moderation:
        )

        Zalster wrote:
        Yup, those are them...I'm over 1150 pending so far, 2048 validated and climbing...

        I can give you two an hour but that's it, but Einstein keeps handing out gamma for some reason.  The server status shows crazy amounts of gravity waiting.  Maybe they consider gamma more important?

        If this page takes an hour to load, reduce posts per page to 20 in your settings, then the tinpot 486 Einstein uses can handle it.

        Tom M
        Tom M
        Joined: 2 Feb 06
        Posts: 6462
        Credit: 9587127867
        RAC: 6850344

        cecht wrote:Calling all

        23 Apr 2020 22:16:48 UTC
        Message 177054 in response to message 177044
        (moderation:
          )

          cecht wrote:
          Calling all volunteers! Now is the time to jump into GW GPU crunching because the current batch of GW GPU VelaJr1 tasks can be run rather quickly.

          I have fired up my Windows 10 box with its gpu pointed only at GW.  

          I expect to have an AMD video card show in the next couple of days.  I will look back and see if I can get the gpu/task segmentation to work so it is pumping out GW tasks.

          I have also switched back to pure gpu processing on both tasks (GW/Pulsar) on my Amd 3950x but who knows if/when it will download anything...

          Anyway, Windows with a pretty dinky Nvidia is running under 20min per task.  On the 3950x it is looking like near 12 hours for the CPU version of GW (it was an experiment, is running on 1 thread).

          Tom M

           

          A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)  I want some more patience. RIGHT NOW!

          Jimbeem
          Jimbeem
          Joined: 14 Aug 10
          Posts: 3
          Credit: 10742661
          RAC: 0

          What have you done to get

          23 Apr 2020 22:18:49 UTC
          Message 177055 in response to message 177044
          (moderation:
            )

            cecht wrote:
            Calling all volunteers! Now is the time to jump into GW GPU crunching because the current batch of GW GPU VelaJr1 tasks can be run rather quickly.  An RX570 at 3x (gpu utilization factor of 0.33) has average task times of 11 to 9 minutes, while a GTX-1080 or RTX-2080 can get one done in 7-8 min. As discussed in that linked forum, a paucity of hosts running those tasks is resulting in those that are not (yet) being assigned a wingman to provide a quorum of two to validate the tasks. At least that's the case for me with my growing "Pending" list.

            What have you done to get them crunch that quickly? I have two concurrent GW GPU VelaJr1 tasks, each always takes around 25 minutes to be completed. Have you raised the amount of assisting CPUs? Currently, I have 0.9 CPUs + 0,5 AMD GPUs. According to the AMD driver my card has a workload of 60-70%. Overall CPU load is at 25% in average.

            cecht
            cecht
            Joined: 7 Mar 18
            Posts: 1535
            Credit: 2910672045
            RAC: 2097546

            Jimbeem wrote:What have you

            23 Apr 2020 22:43:32 UTC
            Message 177056 in response to message 177055
            (moderation:
              )

              Jimbeem wrote:
              What have you done to get them crunch that quickly? I have two concurrent GW GPU VelaJr1 tasks, each always takes around 25 minutes to be completed. Have you raised the amount of assisting CPUs? Currently, I have 0.9 CPUs + 0,5 AMD GPUs. According to the AMD driver my card has a workload of 60-70%. Overall CPU load is at 25% in average.

              HA! Well, that speed spurt didn't last long. I just checked and those VelaJr1 tasks are now taking ~24 minutes. I guess the short tasks at the magic frequency were a "limited time offer" that I thought would last longer than it did. My CPU utilization is set to 0.5, but that doesn't affect CPU usage or thread allocation.  I can't recall why I set it to 0.5 from the default 0.9, but I'm sure I had a good reason at the time. LOL "If it ain't broke don't fix it", so I'll leave it at that setting for now.

              Ideas are not fixed, nor should they be; we live in model-dependent reality.

              Mr P Hucker
              Mr P Hucker
              Joined: 12 Aug 06
              Posts: 838
              Credit: 519325953
              RAC: 14237

              cecht wrote:HA! Well, that

              23 Apr 2020 23:46:08 UTC
              Message 177057 in response to message 177056
              (moderation:
                )

                cecht wrote:
                HA! Well, that speed spurt didn't last long. I just checked and those VelaJr1 tasks are now taking ~24 minutes. I guess the short tasks at the magic frequency were a "limited time offer" that I thought would last longer than it did. My CPU utilization is set to 0.5, but that doesn't affect CPU usage or thread allocation.  I can't recall why I set it to 0.5 from the default 0.9, but I'm sure I had a good reason at the time. LOL "If it ain't broke don't fix it", so I'll leave it at that setting for now.

                I gave up with that usage thing, and set all GPU tasks to request 0.01 CPU cores (0 seems to have hidden meaning).  Then I use the "use at most x% of the processors" in the basic Boinc settings.  Something to do with LHC's multiple CPU core tasks (Atlas).  I think what happened was since I had the computer set to use all 6 cores, minus whatever the Einstein GPU tasks subtracted, LHC still gave me 6 core Atlas tasks (a failing of Boinc), so Einstein kept getting throttled.  I just worked out the most CPU cores that would be needed for GPU work, and removed them permanently.

                If this page takes an hour to load, reduce posts per page to 20 in your settings, then the tinpot 486 Einstein uses can handle it.

                TBar
                TBar
                Joined: 3 Apr 20
                Posts: 24
                Credit: 891961726
                RAC: 0

                So... watch my back, I'm

                24 Apr 2020 0:16:37 UTC
                Message 177059
                (moderation:
                  )

                  So... watch my back, I'm going in.

                  I just received (hopefully) the last lockdown buster from eBay today. A low power i7-8700 to run a new MSI mining board. The 12 core 8700s are going for just a little more than the 8 core 6700s, and the board is going for $60. I decided it would be worth it to be able to run 12 GPUs again. I was running 14 GPUs on a 6700 at SETI, but the SETi App is happy with just Half a CPU core, unlike this place. The afternoon test shows a big improvement on the Gravity App, not so much on the Gamma App. I was just testing with a single GTX970, I'm now going to swap out the new board with the one running the 8 higher end nVidias. Hopefully it will only take an hour or two...

                   

                  mikey
                  mikey
                  Joined: 22 Jan 05
                  Posts: 12699
                  Credit: 1839101286
                  RAC: 3698

                  Keith Myers wrote:mikey

                  24 Apr 2020 0:22:14 UTC
                  Message 177060 in response to message 177051
                  (moderation:
                    )

                    Keith Myers wrote:
                    mikey wrote:
                    Keith Myers wrote:
                    Yes my pending list is predominantly those Velajr1 task awaiting a wingman.

                    So is that these tasks?

                    Gravitational Wave search O2 Multi-Directional GPU v2.07 () x86_64-pc-linux-gnu

                    Yes, one host is getting those mostly along with the FGRPB1G tasks.  The other host is getting mostly the FGRPB1G tasks and a smattering of GW 2.07 gpu task.

                    This host has a lot of the Velajr1 tasks in Pending and awaiting a wingman with a 1 Jan 1970 sent date.

                    https://einsteinathome.org/host/12444941/tasks/2/54

                    Or it did earlier.  Looks like I have a wingmen for those tasks finally.  The scheduler must have crawled through the pendings and sent out for wingmen since my last post.

                    So you don't need another wingman anymore?

                    mikey
                    mikey
                    Joined: 22 Jan 05
                    Posts: 12699
                    Credit: 1839101286
                    RAC: 3698

                    Peter Hucker wrote:mikey

                    24 Apr 2020 0:25:16 UTC
                    Message 177061 in response to message 177052
                    (moderation:
                      )

                      Peter Hucker wrote:
                      mikey wrote:
                      Peter Hucker wrote:
                      mikey wrote:
                      !! I'm at the point now where I'm trying to figure out which Project to run but have too many choices to make!! I have 15pc's running now and some can run this  Project but not that Project while others can't  run either of those. I'm also on a Team and trying not to pass my Teammates, just because I can, so that adds into things too. I will pass them if they quit a Project but not if  they are actively crunching it, I believe in the axiom 'just because you can doesn't mean you should'.

                      Agreed, apart from the teammate thing, I don't get why you wouldn't want to pass them.  A team in formula 1 could have one driver passing another.  If one can win the race and beat other teams, then they have to pass the rest of their own team first.

                      Oh I could easily pass any of them but that's not being a good TEAMmate, they earned their spot at that Project and why should I knock them off it just because I can. One teammate had over 20 million at one project while I had 3.5 million, 3 weeks later I had over 10 million and stopped. I did pass another teammate at another project during a challenge but then stopped and let him pas me back again after it was over, I passed him in under 4 days. At Collatz if I put every gpu I  have on it I get close to to 60 million credits per day.

                      PS i also like F1!!!

                      Well in most races you can help your teammates, but there's no way you can do that here.  So apart from stopping processing for a bit (which isn't good for the science), I don't see how you can not pass them.  Unless you just work for another project?

                      Anyway isn't the point of a team to achieve as much as you can collectively?  So what if you do more than them?  If you really want to help them out, post one of your GPUs to them :-)

                      Yes I switch Projects so I don't pass them, I'm #1 on my Team in enough places that I just pad my lead at other Projects.

                      I have sent my older gpu's to my teammates, I've also sold and given my pc's to teammates as well when they needed them.

                      Comment viewing options

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

                      Forums › Cruncher's Corner