Update - at about 17:00 UTC on Dec 26 the flood gates opened and projects started flowing. I now have 16 tasks for this CPU.
Didn't do anything from just after my last post to now except bring the laptop home and powered it up in my den but forgot to plug in the network cable - it ran for a day like this. When I eventually logged in and saw the network error messages the logs showed "Requesting new tasks for CPU". Sometime shortly after that I had projects for everything.
Thanks Gary. I will create the file that you recommended and if things get stuck again I'll put it into place. For now I think I'll just let it run (don't fidget with things that are working). I have all my projects at the default scheduling (100 for each). Found it worked best all around. Since this device tends to be faster then my other machines I allow it to pull a few more days already (helps when you tend to forget to plug in the network cable).
I appreciate everyone's help with this! I recognize that the opportunity may have passed but if you want me to try something else "for posterity's sake" let me know.
Was away for a few days and things stopped working again (i.e. no new projects - same error message about needing to update the client). Moved the laptop back to the office but didn't have the same luck as the last move (where it came online and immediately downloaded work).
Interesting that it still only effects this machine - the only one I have with a GPU.
Was away for a few days and things stopped working again (i.e. no new projects - same error message about needing to update the client).
Presumably you mean that E@H stopped downloading new CPU tasks? Please forget about the GPU distraction. It's wrong to refer to it as an error. It's simply a canned response from the server telling you one of the reasons that prevent it from responding positively to your client's request for GPU work.
According to your account page, you support 4 projects, Seti, E@H, CPDN and Rosetta. It would be quite normal to have periods where you don't actually have tasks on board for all four projects simultaneously. Presumably you do have work for some of these projects and you don't actually have idle CPUs? It's wrong to say, "things stopped working again", if what you really mean is that you don't currently have tasks for all 4 projects.
Can you please open BOINC Manager at the 'projects' tab and tell us the projects you see listed there, their resource shares and any information listed in the status column?
You said yourself in a previous message that you got a bunch of work in one hit when you turned on network access that had been left off. Now that that work has been crunched and returned, it's very likely that BOINC is simply repaying debts to other projects and will get further E@H tasks when it is good and ready. Another factor that will have an influence on this is the preference settings you have set for network usage, ie, 'connect to network every X days' and 'additional Y days work buffer'. What have you set for these two preferences?
Can you please open BOINC Manager at the 'projects' tab and tell us the projects you see listed there, their resource shares and any information listed in the status column?
Ok. Ignoring the GPU. I have 4 projects:
Climateprediction.net, Resource 100 (25%), cpu sched -47314, Fetch 0.00
E@H, Resource 100 (25%), cpu sched 0.00, Fetch 0.00 (deferred under just under 24hrs)
rosetta, Resource 100 (25%), cpu sched -39117, fetch -77833
Seti, Resource 100 (25%), cpu sched 0.00, fetch 0.00 (deferred 2 hrs)
Current Tasks:
Seti - 7 to report (100% done), 3 active/waiting for CUDA, 0 for CPU
Rosetta - 2 for CPU
Climate - 2 for CPU
E@H - None
Based on the time estimates I'll run out of Rosetta today.
Transfers:
Seti - 6 upload pending (the project isn't responding)
Messages:
E@H has the GPU thing (which we are ignoring)
Climate is now reporting no work for GPU but I'm ignoring it as I still have lots of time left on the 2 CPU tasks (300+ hrs)
Quote:
Another factor that will have an influence on this is the preference settings you have set for network usage, ie, 'connect to network every X days' and 'additional Y days work buffer'. What have you set for these two preferences?
Connect every: 0.10 (default) days
Additional work buffer: 5.00 days (I belive the default is 1.00)
Switch every 120 minutes
1/6/2010 7:38:26 AM Einstein@Home Requesting new tasks for GPU
1/6/2010 7:38:32 AM Einstein@Home Scheduler request completed: got 0 new tasks
1/6/2010 7:38:32 AM Einstein@Home Message from server: No work sent 1/6/2010 7:38:32 AM Einstein@Home Message from server: (your BOINC client is old - please install current version)
Today I received this:
1/7/2010 7:38:32 AM Einstein@Home Requesting new tasks for GPU
1/7/2010 7:38:37 AM Einstein@Home Scheduler request completed: got 0 new tasks
1/7/2010 7:38:37 AM Einstein@Home Message from server: No work sent 1/7/2010 7:38:37 AM Einstein@Home Message from server: NVIDIA display driver 19038 or later needed
And it downloaded a couple of CPU projects after.
I know that my NVIDIA driver can't be upgraded to this version (I'm stuck where I am) but at least it is now bypassing automatically!
Good work and Thanks to whomever made the changes! This is much better for those of us with ancient hardware. ;-)
1/21/2010 8:48:51 AM Einstein@Home Sending scheduler request: To fetch work.
1/21/2010 8:48:51 AM Einstein@Home Requesting new tasks for GPU
1/21/2010 8:48:56 AM Einstein@Home Scheduler request completed: got 0 new tasks
1/21/2010 8:48:56 AM Einstein@Home Message from server: No work sent
1/21/2010 8:48:56 AM Einstein@Home Message from server: (your BOINC client is old - please install current version)
I checked the settings for this Machine and it is still marked as "don't use GPU" but that (as it was pointed out) only works for 6.10.X clients - which I don't have because my Dell laptop video driver (and other supplied at Driver sites) can't be upgraded to a suitable version. :-(
Looks like my choices are slim now
- Stay at 6.6.38 to run the CUDA for SETI and cross my fingers that Einstein will give me a CPU project
- Upgrade the software and don't run CUDA
- Upgrade the hardware (ain't going to happen soon)
Thanks to all that have offered suggestions. I really have learned a lot through this process.
Update - at about 17:00 UTC
)
Update - at about 17:00 UTC on Dec 26 the flood gates opened and projects started flowing. I now have 16 tasks for this CPU.
Didn't do anything from just after my last post to now except bring the laptop home and powered it up in my den but forgot to plug in the network cable - it ran for a day like this. When I eventually logged in and saw the network error messages the logs showed "Requesting new tasks for CPU". Sometime shortly after that I had projects for everything.
Thanks Gary. I will create the file that you recommended and if things get stuck again I'll put it into place. For now I think I'll just let it run (don't fidget with things that are working). I have all my projects at the default scheduling (100 for each). Found it worked best all around. Since this device tends to be faster then my other machines I allow it to pull a few more days already (helps when you tend to forget to plug in the network cable).
I appreciate everyone's help with this! I recognize that the opportunity may have passed but if you want me to try something else "for posterity's sake" let me know.
Thanks!
Mike
Downdate. Was away for a
)
Downdate.
Was away for a few days and things stopped working again (i.e. no new projects - same error message about needing to update the client). Moved the laptop back to the office but didn't have the same luck as the last move (where it came online and immediately downloaded work).
Interesting that it still only effects this machine - the only one I have with a GPU.
RE: Was away for a few days
)
Presumably you mean that E@H stopped downloading new CPU tasks? Please forget about the GPU distraction. It's wrong to refer to it as an error. It's simply a canned response from the server telling you one of the reasons that prevent it from responding positively to your client's request for GPU work.
According to your account page, you support 4 projects, Seti, E@H, CPDN and Rosetta. It would be quite normal to have periods where you don't actually have tasks on board for all four projects simultaneously. Presumably you do have work for some of these projects and you don't actually have idle CPUs? It's wrong to say, "things stopped working again", if what you really mean is that you don't currently have tasks for all 4 projects.
Can you please open BOINC Manager at the 'projects' tab and tell us the projects you see listed there, their resource shares and any information listed in the status column?
You said yourself in a previous message that you got a bunch of work in one hit when you turned on network access that had been left off. Now that that work has been crunched and returned, it's very likely that BOINC is simply repaying debts to other projects and will get further E@H tasks when it is good and ready. Another factor that will have an influence on this is the preference settings you have set for network usage, ie, 'connect to network every X days' and 'additional Y days work buffer'. What have you set for these two preferences?
Cheers,
Gary.
Thanks again for the
)
Thanks again for the reply.
Ok. Ignoring the GPU. I have 4 projects:
Climateprediction.net, Resource 100 (25%), cpu sched -47314, Fetch 0.00
E@H, Resource 100 (25%), cpu sched 0.00, Fetch 0.00 (deferred under just under 24hrs)
rosetta, Resource 100 (25%), cpu sched -39117, fetch -77833
Seti, Resource 100 (25%), cpu sched 0.00, fetch 0.00 (deferred 2 hrs)
Current Tasks:
Seti - 7 to report (100% done), 3 active/waiting for CUDA, 0 for CPU
Rosetta - 2 for CPU
Climate - 2 for CPU
E@H - None
Based on the time estimates I'll run out of Rosetta today.
Transfers:
Seti - 6 upload pending (the project isn't responding)
Messages:
E@H has the GPU thing (which we are ignoring)
Climate is now reporting no work for GPU but I'm ignoring it as I still have lots of time left on the 2 CPU tasks (300+ hrs)
Connect every: 0.10 (default) days
Additional work buffer: 5.00 days (I belive the default is 1.00)
Switch every 120 minutes
The rest I haven't changed.
Mike
Cool! Something changed in
)
Cool! Something changed in the last 24 hours.
Yesterday I received this:
1/6/2010 7:38:26 AM Einstein@Home Requesting new tasks for GPU
1/6/2010 7:38:32 AM Einstein@Home Scheduler request completed: got 0 new tasks
1/6/2010 7:38:32 AM Einstein@Home Message from server: No work sent
1/6/2010 7:38:32 AM Einstein@Home Message from server: (your BOINC client is old - please install current version)
Today I received this:
1/7/2010 7:38:32 AM Einstein@Home Requesting new tasks for GPU
1/7/2010 7:38:37 AM Einstein@Home Scheduler request completed: got 0 new tasks
1/7/2010 7:38:37 AM Einstein@Home Message from server: No work sent
1/7/2010 7:38:37 AM Einstein@Home Message from server: NVIDIA display driver 19038 or later needed
And it downloaded a couple of CPU projects after.
I know that my NVIDIA driver can't be upgraded to this version (I'm stuck where I am) but at least it is now bypassing automatically!
Good work and Thanks to whomever made the changes! This is much better for those of us with ancient hardware. ;-)
Mike
Looks like my issue is back
)
Looks like my issue is back again.
1/21/2010 8:48:51 AM Einstein@Home Sending scheduler request: To fetch work.
1/21/2010 8:48:51 AM Einstein@Home Requesting new tasks for GPU
1/21/2010 8:48:56 AM Einstein@Home Scheduler request completed: got 0 new tasks
1/21/2010 8:48:56 AM Einstein@Home Message from server: No work sent
1/21/2010 8:48:56 AM Einstein@Home Message from server: (your BOINC client is old - please install current version)
I checked the settings for this Machine and it is still marked as "don't use GPU" but that (as it was pointed out) only works for 6.10.X clients - which I don't have because my Dell laptop video driver (and other supplied at Driver sites) can't be upgraded to a suitable version. :-(
Looks like my choices are slim now
- Stay at 6.6.38 to run the CUDA for SETI and cross my fingers that Einstein will give me a CPU project
- Upgrade the software and don't run CUDA
- Upgrade the hardware (ain't going to happen soon)
Thanks to all that have offered suggestions. I really have learned a lot through this process.
Just and update on
)
Just and update on this.
Dell finally released a 1.8 driver and I found a 1.9 driver on NVIDIA's site for this laptop (both released since the last update on this thread).
The new 6.10.18 client now works with my CUDA!
RE: Just and update on
)
I'm so glad you persisted with this battle and have ultimately been rewarded :-).
Thanks very much for letting us all know. It's always gratifying when a person goes to the trouble of 'closing the account' on a saga like this.
Cheers,
Gary.