AK V8 + CUDA MB team work mod

[Raistmer]

whaterrormessage, what stderr og stuck tasks, links on results... too littleinfo to go.

[Jason G]

If this is similar to the stuck tasks I 'used to' get, then there'll be no error messages, etc.. at all.  It just sticks at zero progress.  Restarting Boinc (service or boinc itself depending on installation type) should get them going normally. 

If so, then restarting boinc is the only way I found to make them 'un-stick' and happens to me every couple of days or so.  To stop from having to manually handle this, I installed the VBScript from the other thread here, then, as I run as a service, made the modification described later in that thread to restart the service instead of terminating the tasks.  I haven't had to do any manual intervention since, but noticed it automatically restarts boinc every couple of days or so.  I was not  able to pin the 'sticky tasks' to any particular angle range or other factor that I could determine.

Jason

[Raistmer]

His picture shows that CUDA app still running ~4% CPU is normal ...

[Archangel999]

If this is similar to the stuck tasks I 'used to' get, then there'll be no error messages, etc.. at all.  It just sticks at zero progress.  Restarting Boinc (service or boinc itself depending on installation type) should get them going normally. 

If so, then restarting boinc is the only way I found to make them 'un-stick' and happens to me every couple of days or so.  To stop from having to manually handle this, I installed the VBScript from the other thread here, then, as I run as a service, made the modification described later in that thread to restart the service instead of terminating the tasks.  I haven't had to do any manual intervention since, but noticed it automatically restarts boinc every couple of days or so.  I was not  able to pin the 'sticky tasks' to any particular angle range or other factor that I could determine.

Jason

me too i restart the boinc to go again there is no problem with the wu it is from the boinc or the app because it happen all the time

[Archangel999]

His picture shows that CUDA app still running ~4% CPU is normal ...

yes but it is the AK v8 app that is stuck i haven't got any problem with the gpu it is from the boinc or the app i don't know you are the pro wow 

[Jason G]

AKv8 stuck?  Uh Oh...(never heard of that before..)  I didn't do it    Exit stage left ...

[Archangel999]

i don't know it some boinc problem because before 5 hours it work 3 work unit now when i come back it is working all the 5 again from time to time stop working 1 or 2 unit but they as you see on the pic are in the memory and the boinc says that they are running i don't know maybe because my xp is x64 ??

[Jason G]

Ahh,  may not apply completely in your situation, but this is basically how it works for me.  I reckon it is the imperfect Boinc scheduling mechanism still.  Thinking logically through my situation (And extending it to quad core scenario, I have only dual core so I'm using my imagination):

-  5 tasks going ( 4 x CPU & 1 x Cuda, Total ~4.04 CPU + 1 Cuda), was possible because cuda started first!
-  Cuda task finishes.
-  Boinc says I have 4 CPUs & 4 apps allocated (no more room for 0.04 CPU required for CUDa)
-  Boinc waits until 1 CPU task finishes (Gets down to 3) then allocates a new cuda task, has available CPU so allocates another AK task afterwards.  (five running again)

For myself, it worked best without using any ncpus mod, and sticking with Boinc 6.5.0.

Yeah I found it frustrating enough that I went back to running the v7VLARkill mod & AstroPulse on the CPUs, but with AP off on main, no longer a suitable option.

I hope there might be some useful clues amongst that mess, that might apply to your situation.

Jason.

[Archangel999]

aham but doesn't explain that when there are only 2 working and 3 working not 5 will wait to see the answer for the Raimster  Thank you

Best Regards
D.Draganov

[Raistmer]

Still wanna see stderrs of hanging tasks.
There is update for "team" AK_v8 that will exit after checkpointing when it detects that CUDA not running. So, if CUDA task finishes, one of AK_v8 will exit with zero status.
BOINC will restart it then allowing CUDA app to run.
So, there should be no situations described by Jason now.

ADDON: here is update:
http://lunatics.kwsn.net/gpu-crunching/ak-v8-cuda-mb-team-work-mod.msg13780.html#msg13780

[Archangel999]

will test this night
Thanks

Best Regards
D.Draganov

[ofp1]

Hi,
I am using Raistmer's V8a SSSE3 on a Quadcore 9550 and a GTX260  with cc_config.xml <ncpus>5</ncpus> option  and setiathome 6.4.5 on vista ultimate 32 bits, it was running ok  5 workunits at a time and the computer RAC climbed from 4600 to 7400 and still up but now I noticed that still are procesing 5 WU at a time but the CUDA seems to be not working  the GPU temperature is down (when it was crunching was about 74 and now is 50) and the procesing time is the same in all 5 WU, also in the BOINC messages  says NO CUDA device detected and it was detected before, and I am running out of WU's because it says "Project has no jobs available". What can I do? how do  I know if the GPU is working?
Thanks

[Raistmer]

1) How you could know GPU is working - you already used right method - looked at GPU temp.
If temp = idle temp - GPU idle, if temp increased - GPU working hard.

2) If you have no work from SETI then you probably doing another non-CUDA enabled projects. So your CPU will do 5 CPU tasks and no CUDA ones.
Try to get new tasks for SETI MultiBeam.

[ofp1]

1) How you could know GPU is working - you already used right method - looked at GPU temp.
If temp = idle temp - GPU idle, if temp increased - GPU working hard.

2) If you have no work from SETI then you probably doing another non-CUDA enabled projects. So your CPU will do 5 CPU tasks and no CUDA ones.
Try to get new tasks for SETI MultiBeam.

Hi,
I am only running SETI, I still have some WU's but no one have been crunched by the GPU since yesterday and BOINC is not recognising my CUDA device.
How do I get new tasks from SETI MultiBeam?

[Raistmer]

Link on your host, please ?