Alrighty.

I have a Mac OS 10.6.2, running BOINC 6.10.33. I have an NVIDIA 8600M GT.

I also have a thread on this problem at the Collatz forum here: http://boinc.thesonntags.com/collatz/forum_thread.php?id=331

I'm running the Collatz Project, as my only GPU project. I've set it to only run GPU projects (turned off CPU tsks for that project).

It's running fine. That is, my computer downloads GPU tasks, runs them, and returns them without error.

There are two separate, but perhaps related issues:

One is that BOINC downloads more than 100 Collatz GPU tasks. My GPU seems to complete perhaps 6-10 a day, and my computer generally runs BOINC 22 hours/day. Previously, I had set my comp to have 4 days of additional work, but I have since (about a week ago) reduced it to 0.5.

The second is that BOINC, after downloading all these projects, starts a task, and then does not finish it. That is, it starts the task, switches to a new one, and so on and so on, rarely finishing a task. To give a snapshot of this, right now my computer is:

1 GPU task running
8 GPU tasks have begun and are waiting to run
102 GPU tasks are waiting to start

I have tried reseting the project. This does not seem to change anything. I have altered the amount of days extra of work it should have from 4 (which is what I've had for several years) to 0.5 (changed this about a week ago). This does not seem to do anything different.

None of the other projects I run seem to have this problem or are affected at all. I run World Community Grid, Einstein, Rosetta, SETI, and Climemateproject, in addition to Collatz.

Thanks for your time.

I have the same problem here, and it's not only limited to GPU, but CPU as well in Collatz.

At time of writing, I have over 60 CPU and 6 GPU workunits in a state of "waiting to run" hovering between 20% to 99% completion. This has caused me difficulty to download further workunits and maintain a cache of 1 week since my computer rarely go online.

My computer is a Mac Pro with nVidia GT 120 running 10.6.2 with BOINC 6.10.33 installed.

I have another Mac Pro at work, without any means of CUDA, so it run solely on CPU workunits. It's running 10.6.2 but with BOINC 6.6.36, and it has no issues of "waiting to run" whatsoever, and these workunits are completed sequentially as they are downloaded.

I have the same problem here, and it's not only limited to GPU, but CPU as well in Collatz.

At time of writing, I have over 60 CPU and 6 GPU workunits in a state of "waiting to run" hovering between 20% to 99% completion. This has caused me difficulty to download further workunits and maintain a cache of 1 week since my computer rarely go online.

My computer is a Mac Pro with nVidia GT 120 running 10.6.2 with BOINC 6.10.33 installed.

I have another Mac Pro at work, without any means of CUDA, so it run solely on CPU workunits. It's running 10.6.2 but with BOINC 6.6.36, and it has no issues of "waiting to run" whatsoever, and these workunits are completed sequentially as they are downloaded.

Hmm...that leads me to believe that it's an issue with this Alpha/Beta release of BOINC. I'd roll back normally, but the latest "stable" release of BOINC doesn't let me run GPU tasks. A conumdrum indeed.

Hmm...that leads me to believe that it's an issue with this Alpha/Beta release of BOINC. I'd roll back normally, but the latest "stable" release of BOINC doesn't let me run GPU tasks. A conumdrum indeed.

Noticed your first post, I've also got an earlier session of Collatz running on GPU only alongside Einstein@home CPU units, but it downloaded 150 Collatz GPU workunits, and my GT 120 completes one unit every 3 hours. With 14 days time frame, it was impossible to complete all 150 workunits. That was with BOINC 6.10.29 installed.

BOINC scheduler in a mess...?

Hmm...that leads me to believe that it's an issue with this Alpha/Beta release of BOINC. I'd roll back normally, but the latest "stable" release of BOINC doesn't let me run GPU tasks. A conumdrum indeed.

Noticed your first post, I've also got an earlier session of Collatz running on GPU only alongside Einstein@home CPU units, but it downloaded 150 Collatz GPU workunits, and my GT 120 completes one unit every 3 hours. With 14 days time frame, it was impossible to complete all 150 workunits. That was with BOINC 6.10.29 installed.

BOINC scheduler in a mess...?

I had the problem with 6.10.29 too.

Alrighty.

I have a Mac OS 10.6.2, running BOINC 6.10.33. I have an NVIDIA 8600M GT.

I also have a thread on this problem at the Collatz forum here: http://boinc.thesonntags.com/collatz/forum_thread.php?id=331

I'm running the Collatz Project, as my only GPU project. I've set it to only run GPU projects (turned off CPU tsks for that project).

It's running fine. That is, my computer downloads GPU tasks, runs them, and returns them without error.

There are two separate, but perhaps related issues:

One is that BOINC downloads more than 100 Collatz GPU tasks. My GPU seems to complete perhaps 6-10 a day, and my computer generally runs BOINC 22 hours/day. Previously, I had set my comp to have 4 days of additional work, but I have since (about a week ago) reduced it to 0.5.

The second is that BOINC, after downloading all these projects, starts a task, and then does not finish it. That is, it starts the task, switches to a new one, and so on and so on, rarely finishing a task. To give a snapshot of this, right now my computer is:

1 GPU task running
8 GPU tasks have begun and are waiting to run
102 GPU tasks are waiting to start

I have tried reseting the project. This does not seem to change anything. I have altered the amount of days extra of work it should have from 4 (which is what I've had for several years) to 0.5 (changed this about a week ago). This does not seem to do anything different.

None of the other projects I run seem to have this problem or are affected at all. I run World Community Grid, Einstein, Rosetta, SETI, and Climemateproject, in addition to Collatz.

Thanks for your time.

Slicker has been changing the flops values of tasks so first time users get a WU with a proper estimated duration, see: A rather long workunit?

My laptop with it's 8400M GS got 79 Cuda wu's each with 1 hour duration estimated, but with each wu taking over 10 hours, i had over 800hrs of work cached,
but less than 330 hours to do it, i had already suspended most of them, so i had to set NNT, and went through my tasks aborting a lot of them so i only had 2 or 3 tasks due on the same day,
Now because there has been a mixture of tasks with different flops values, i only have 4 tasks with 45hour duration estimated, DCF will fall in time and my cache will fill up again.
I suggest you do likewise, abort any surplus that haven't been started, and suspend some if necessary,
Note: Resetting the project (at Collatz) only re-downloads the apps and the very same wu's you had before!!

Claggy

I've just reported the problem on the BOINC_alpha mailing list. I hope the rest who encounters the same problem to do the same as well.

I've just downloaded a few CPU workunits today, and they've got extra long completion times, over 100 hours, and now BOINC is treating them as high priorities and running them once they've completely downloaded from Collatz servers. It has made things worse as older workunits are made "waiting to run". I have to babysit my Mac Pro from now on just to manually suspend new workunits so that older workunits are completed first before resuming the former.

Yesterday I had reported this thread to the attention of the developer for the Macintosh. He has told me he is aware of the problem, best even, he has one computer affected with the exact same problem on Einstein.

---

I aborted like 100 of my tasks, and am just having my machine run the rest.

Aborting all the tasks appears to have resolved the problem, at least for now. My BOINC finished all the tasks, and is now only downloading one at a time.

Aborting all the tasks appears to have resolved the problem, at least for now. My BOINC finished all the tasks, and is now only downloading one at a time.

Let me rephrase that. It's downloading one at a time at a 5-minute scheduler request interval.

After a couple days of this, it seems to have solved my problems.