It simply introduces too often crunching errors in the BRP4 application. I´m sure, it´s not the failure from BOINC but from BRP4 vers 1.25 and most in conjunction with WinVista and Win7. It´s allways error codes 2, 1002 and 1020 at me.
With error code 2 the files end up in the middle to end of the task > low number of absent output files, where as at error code 1002, 1020 the task are short running, sometimes a few seconds only - 150 taks within 10min - with high numbers of absent outputfiles. This happened also after taking out the virus scanner from observing and checking BOINC directories.
With BOINC client 6.12.34 I had also 1 crunching error today - just after writing the foregouÃng notice - but much, much less often.
But as said, soon more about this nerving failure.
I do not like the tab order and the fact the event log "messages" isn't in a tab in V7. Why was this changed ???. I will probably adapt to it, but for now I ask why was the tab order changed and event log removed from the tabs. Is there a way to change the order to the users preference?.
I reinstalled 6.10.60 and my GPU crunching now works again. I Uninstalled and manually deleted all remaining boinc files from both directories and started fresh.
I noticed about the time the GPU stopped crunching the event log said it deleted a file that was not needed anymore, it was "JPLEPH.405" I noticed on my new 6.10.60 this file is again in my program data directory. Could this file deletion have anything to do with my previous problems? CCleaner flagged it previously as an "unused file extension" and removed the registry entry for it, I assume boinc then deleted it as an unknown file..
Once I narrow down the problems I will probably install version 7 again. I know it has better features than 6 does.
For now though it's working again, I tried reinstalling version 7.0.28 two times before going back to V6 and it was still broken.
Reinstall of Nvidia drivers. detach and reattach, and reset project also had no effect.
It was only Einstein, all other GPU projects worked fine.
I do not like the tab order and the fact the event log "messages" isn't in a tab in V7. Why was this changed ???. I will probably adapt to it, but for now I ask why was the tab order changed and event log removed from the tabs. Is there a way to change the order to the users preference?.
The Messages tab was removed from BOINC 6.12 and later versions, as it was deemed 'too scary' for a lot of new users. This was the outcome of a poll held for over a year. Too many people thinking that all the messages visible there were errors, and them uninstalling BOINC immediately after seeing them. Remember that these were also visible in the Simple View.
The new Event Log window is rather nice for those of us who use it a lot. Now we can put the messages side by side with the actual tasks tab, see what our debug flags do.
So the Messages Log was removed, the Notices tab added, with the thought behind it that any messages in there would be things the user couldn't do a thing about anyway.
If you however do not like it, you can always keep running the old BOINC Manager with the newer BOINC versions. For this you need the e.g. 6.10 boincmgr.exe and boinc.dll files. Either rename that boincmgr.exe to something you want, or use it to overwrite the one you install next. But then you need to do that each time you (upgrade) install BOINC. BOINC 6.12 and later don't use boinc.dll anymore, so you can easily leave that in place, without renaming.
Or use an alternative to BOINC Manager, such as BOINCTasks.
It simply introduces too often crunching errors in the BRP4 application. I´m sure, it´s not the failure from BOINC but from BRP4 vers 1.25 and most in conjunction with WinVista and Win7. It´s allways error codes 2, 1002 and 1020 at me.
These are application error codes, not BOINC error codes.
As I read from the logs;
Exit code 2:
Continuing work on ../../projects/einstein.phys.uwm.edu/p2030.20101112.G45.61+00.64.N.b2s0g0.00000_1071.bin4 at template no. 3679
Input file on command line ../../projects/einstein.phys.uwm.edu/p2030.20101119.G46.75-00.08.C.b4s0g0.00000_3296.bin4 doesn't agree with input file ../../projects/einstein.phys.uwm.edu/p2030.20101112.G45.61+00.64.N.b2s0g0.00000_1071.bin4 from checkpoint header.
If you can't find you car keys, you can't drive your car anywhere either, so how is the science application supposed to run the work, when it can't find a specific input file? ;-)
Is a CUDA application error.
Error 1002 is:
Failed to enable CUDA thread yielding for device #0 (error: 999)! Sorry, will try to occupy one CPU core...
Couldn't acquire CUDA context of device #0 (error: 999)!
Demodulation failed (error: 1002)!
Is a CUDA application error, according to Google even one for Einstein/Albert specifically. You'll have to ask Oliver about that.
And exit code 1008:
Error during CUDA device->host power spectrum data transfer (error: 999)
[11:19:23][3784][ERROR] Demodulation failed (error: 1008)!
This is definitely another CUDA application error, it even says so!
Now, what you didn't mention was all those tasks you ran with 7.0.25, 7.0.26 and 7.0.28 that did run to completion, without any error whatsoever. Want to explain that to me then? :-P
Now, what you didn't mention was all those tasks you ran with 7.0.25, 7.0.26 and 7.0.28 that did run to completion, without any error whatsoever. Want to explain that to me then? :-P
At me in the mean 5.5% of all started BRP4 tasks ended up with crunching error, wasting 3.1% of crunching time within 30 days mostly in April. I think, that´s too much. There have been days with up to 23% failure rate and just 1 day without failure. Between March 29th and May 28th - the day I changed to BOINC 6.12.34 - I had 162 errors while computing. At normal days I´m crunching about 49 tasks/day. The 2 days before changing to 6.12.34 the situation escalated, as reported in my foregoing thread. Since changing to 6.12.34 I had just 1 error while computing with error code 2, which is about 0.4% failurerate, hopefully decreasing.
As said, I will report about my investigations hopefully in a week, as I have to finish some other work before. But I´m happy for every hint in this regard.
I found an old GT430 somewhere and installed that in my i3 system.
Got 6 wu's, all failing witch this exit-code.
Yes, I do have rights. nVidia driver 301.42
I tried to downgrade to 296.10
Now I do not get any wu's, host sceduler log says no app version available: APP#20 (Einstein_S6LV1) Platform#2
I found an old GT430 somewhere and installed that in my i3 system.
Got 6 wu's, all failing witch this exit-code.
Yes, I do have rights. nVidia driver 301.42
I tried to downgrade to 296.10
Now I do not get any wu's, host sceduler log says no app version available: APP#20 (Einstein_S6LV1) Platform#2
Win7 64, 7.0.28
Einstein dosnt send GPU work to hosts with drivers of the versions between 290 to 300 because that versions have a bug, if 301.24 dosnt work you need to go to older drivers (IIRC the latest that works is 285.xx).
My GT430 is working pretty well on Einstein with version 266.58.
I found an old GT430 somewhere and installed that in my i3 system.
Got 6 wu's, all failing witch this exit-code.
Yes, I do have rights. nVidia driver 301.42
I tried to downgrade to 296.10
Now I do not get any wu's, host sceduler log says no app version available: APP#20 (Einstein_S6LV1) Platform#2
Win7 64, 7.0.28
Einstein dosnt send GPU work to hosts with drivers of the versions between 290 to 300 because that versions have a bug, if 301.24 dosnt work you need to go to older drivers (IIRC the latest that works is 285.xx).
My GT430 is working pretty well on Einstein with version 266.58.
Hmmm.
I tried 285.79, then 280.26 and then, with a really clean installation, 266.58
Hmmm.
I tried 285.79, then 280.26 and then, with a really clean installation, 266.58
Sorry, still the same problem.
Yes, it seems there is another issue...
Ive seen that in your last request the log says:
Quote:
2012-06-04 18:30:37.9221 [PID=31148] [send] CPU: req 0.00 sec, 0.00 instances; est delay 0.00 2012-06-04 18:30:37.9221 [PID=31148] [send] CUDA: req 21780.00 sec, 1.00 instances; est delay 0.00
2012-06-04 18:30:37.9221 [PID=31148] [send] work_req_seconds: 0.00 secs
2012-06-04 18:30:37.9221 [PID=31148] [send] available disk 29.71 GB, work_buf_min 0
2012-06-04 18:30:37.9221 [PID=31148] [send] active_frac 0.999880 on_frac 0.954004 DCF 1.204564
2012-06-04 18:30:37.9234 [PID=31148] [send] [HOST#5394959] not reliable; max_result_day 28
2012-06-04 18:30:37.9234 [PID=31148] [send] set_trust: random choice for error rate 0.002028: yes
2012-06-04 18:30:38.0612 [PID=31148] [version] Checking plan class 'SSE2'
2012-06-04 18:30:38.0616 [PID=31148] [version] reading plan classes from file '../plan_class_spec.xml'
2012-06-04 18:30:38.0616 [PID=31148] [version] Don't need CPU jobs, skipping version 113 for einstein_S6LV1 (SSE2)
2012-06-04 18:30:38.0617 [PID=31148] [version] no app version available: APP#20 (einstein_S6LV1) PLATFORM#2 (windows_intelx86) min_version 0
2012-06-04 18:30:38.0712 [PID=31148] [version] Don't need CPU jobs, skipping version 23 for hsgamma_FGRP1 ()
2012-06-04 18:30:38.0712 [PID=31148] [version] no app version available: APP#17 (hsgamma_FGRP1) PLATFORM#2 (windows_intelx86) min_version 0
2012-06-04 18:30:38.0713 [PID=31148] [version] Checking plan class 'BRP4SSE'
2012-06-04 18:30:38.0713 [PID=31148] [version] project prefs setting 'also_run_cpu' (1.000000) prevents using plan class.
2012-06-04 18:30:38.0713 [PID=31148] [version] Checking plan class 'atiOpenCL'
2012-06-04 18:30:38.0713 [PID=31148] [version] parsed project prefs setting 'gpu_util_brp' : true : 0.500000
2012-06-04 18:30:38.0713 [PID=31148] [version] No ATI devices found
2012-06-04 18:30:38.0713 [PID=31148] [version] Checking plan class 'atiOpenCL6900'
2012-06-04 18:30:38.0713 [PID=31148] [version] parsed project prefs setting 'gpu_util_brp' : true : 0.500000
2012-06-04 18:30:38.0714 [PID=31148] [version] No ATI devices found
2012-06-04 18:30:38.0714 [PID=31148] [version] Checking plan class 'BRP4cuda32'
2012-06-04 18:30:38.0714 [PID=31148] [version] parsed project prefs setting 'gpu_util_brp' : true : 0.500000 2012-06-04 18:30:38.0714 [PID=31148] [version] Skipping CUDA version - user prefs say no CUDA
2012-06-04 18:30:38.0714 [PID=31148] [version] Checking plan class 'BRP4cuda32nv301'
2012-06-04 18:30:38.0714 [PID=31148] [version] parsed project prefs setting 'gpu_util_brp' : true : 0.500000
2012-06-04 18:30:38.0714 [PID=31148] [version] driver version required min: -30100, supplied: 26658
2012-06-04 18:30:38.0714 [PID=31148] [version] no app version available: APP#19 (einsteinbinary_BRP4) PLATFORM#2 (windows_intelx86) min_version 0
2012-06-04 18:30:38.0755 [PID=31148] [send] [HOST#5394959] is looking for work from a non-preferred application
2012-06-04 18:30:38.0816 [PID=31148] [debug] [HOST#5394959] MSG(high) No work sent
2012-06-04 18:30:38.0816 [PID=31148] [debug] [HOST#5394959] MSG(high) see scheduler log messages on http://einstein.phys.uwm.edu//host_sched_logs/5394/5394959 2012-06-04 18:30:38.0816 [PID=31148] [debug] [HOST#5394959] MSG( low) Jobs for NVIDIA GPU are available, but your preferences are set to not accept them
2012-06-04 18:30:38.0817 [PID=31148] Sending reply to [HOST#5394959]: 0 results, delay req 60.00
2012-06-04 18:30:38.0820 [PID=31148] Scheduler ran 0.274 seconds
So, you need to go to your preferences for the project and there you need to enable (set to yes) the option for USE NVIDIA GPU, also be sure that the "Run Only the selected applications" is enabled for "Binary Radio Pulsar Search (Arecibo)" (and optionally set the "Run CPU versions of applications for which GPU versions are available" to disabled, as they take longer on CPU)
Once the settings are changed, you will need to hit the update button in Boinc Manager to read the new settings and if everything works in the next update you will get GPU tasks.
Hallo
)
Hallo Ageless!
It simply introduces too often crunching errors in the BRP4 application. I´m sure, it´s not the failure from BOINC but from BRP4 vers 1.25 and most in conjunction with WinVista and Win7. It´s allways error codes 2, 1002 and 1020 at me.
With error code 2 the files end up in the middle to end of the task > low number of absent output files, where as at error code 1002, 1020 the task are short running, sometimes a few seconds only - 150 taks within 10min - with high numbers of absent outputfiles. This happened also after taking out the virus scanner from observing and checking BOINC directories.
With BOINC client 6.12.34 I had also 1 crunching error today - just after writing the foregouÃng notice - but much, much less often.
But as said, soon more about this nerving failure.
Kind regards and happy crunching
Martin
I do not like the tab order
)
I do not like the tab order and the fact the event log "messages" isn't in a tab in V7. Why was this changed ???. I will probably adapt to it, but for now I ask why was the tab order changed and event log removed from the tabs. Is there a way to change the order to the users preference?.
I reinstalled 6.10.60 and my GPU crunching now works again. I Uninstalled and manually deleted all remaining boinc files from both directories and started fresh.
I noticed about the time the GPU stopped crunching the event log said it deleted a file that was not needed anymore, it was "JPLEPH.405" I noticed on my new 6.10.60 this file is again in my program data directory. Could this file deletion have anything to do with my previous problems? CCleaner flagged it previously as an "unused file extension" and removed the registry entry for it, I assume boinc then deleted it as an unknown file..
Once I narrow down the problems I will probably install version 7 again. I know it has better features than 6 does.
For now though it's working again, I tried reinstalling version 7.0.28 two times before going back to V6 and it was still broken.
Reinstall of Nvidia drivers. detach and reattach, and reset project also had no effect.
It was only Einstein, all other GPU projects worked fine.
Thanks
RE: I do not like the tab
)
The Messages tab was removed from BOINC 6.12 and later versions, as it was deemed 'too scary' for a lot of new users. This was the outcome of a poll held for over a year. Too many people thinking that all the messages visible there were errors, and them uninstalling BOINC immediately after seeing them. Remember that these were also visible in the Simple View.
The new Event Log window is rather nice for those of us who use it a lot. Now we can put the messages side by side with the actual tasks tab, see what our debug flags do.
So the Messages Log was removed, the Notices tab added, with the thought behind it that any messages in there would be things the user couldn't do a thing about anyway.
If you however do not like it, you can always keep running the old BOINC Manager with the newer BOINC versions. For this you need the e.g. 6.10 boincmgr.exe and boinc.dll files. Either rename that boincmgr.exe to something you want, or use it to overwrite the one you install next. But then you need to do that each time you (upgrade) install BOINC. BOINC 6.12 and later don't use boinc.dll anymore, so you can easily leave that in place, without renaming.
Or use an alternative to BOINC Manager, such as BOINCTasks.
RE: It simply introduces
)
These are application error codes, not BOINC error codes.
As I read from the logs;
Exit code 2:
If you can't find you car keys, you can't drive your car anywhere either, so how is the science application supposed to run the work, when it can't find a specific input file? ;-)
Is a CUDA application error.
Error 1002 is:
Is a CUDA application error, according to Google even one for Einstein/Albert specifically. You'll have to ask Oliver about that.
And exit code 1008:
This is definitely another CUDA application error, it even says so!
Now, what you didn't mention was all those tasks you ran with 7.0.25, 7.0.26 and 7.0.28 that did run to completion, without any error whatsoever. Want to explain that to me then? :-P
Hallo Jord!! Thanks for your
)
Hallo Jord!!
Thanks for your reply.
At me in the mean 5.5% of all started BRP4 tasks ended up with crunching error, wasting 3.1% of crunching time within 30 days mostly in April. I think, that´s too much. There have been days with up to 23% failure rate and just 1 day without failure. Between March 29th and May 28th - the day I changed to BOINC 6.12.34 - I had 162 errors while computing. At normal days I´m crunching about 49 tasks/day. The 2 days before changing to 6.12.34 the situation escalated, as reported in my foregoing thread. Since changing to 6.12.34 I had just 1 error while computing with error code 2, which is about 0.4% failurerate, hopefully decreasing.
As said, I will report about my investigations hopefully in a week, as I have to finish some other work before. But I´m happy for every hint in this regard.
Kind regards and happy crunching
Martin
Hi, I found an old GT430
)
Hi,
I found an old GT430 somewhere and installed that in my i3 system.
Got 6 wu's, all failing witch this exit-code.
Yes, I do have rights. nVidia driver 301.42
I tried to downgrade to 296.10
Now I do not get any wu's, host sceduler log says no app version available: APP#20 (Einstein_S6LV1) Platform#2
Win7 64, 7.0.28
RE: Hi, I found an old
)
Einstein dosnt send GPU work to hosts with drivers of the versions between 290 to 300 because that versions have a bug, if 301.24 dosnt work you need to go to older drivers (IIRC the latest that works is 285.xx).
My GT430 is working pretty well on Einstein with version 266.58.
RE: RE: Hi, I found an
)
Hmmm.
I tried 285.79, then 280.26 and then, with a really clean installation, 266.58
Sorry, still the same problem.
RE: Hmmm. I tried 285.79,
)
285 should be ok for you, but the log has this to say about your host's most recent attempt to fetch work:
Can you check your preferences? You probably haven't enabled CUDA jobs there.
CU
HBE
RE: Hmmm. I tried 285.79,
)
Yes, it seems there is another issue...
Ive seen that in your last request the log says:
So, you need to go to your preferences for the project and there you need to enable (set to yes) the option for USE NVIDIA GPU, also be sure that the "Run Only the selected applications" is enabled for "Binary Radio Pulsar Search (Arecibo)" (and optionally set the "Run CPU versions of applications for which GPU versions are available" to disabled, as they take longer on CPU)
Once the settings are changed, you will need to hit the update button in Boinc Manager to read the new settings and if everything works in the next update you will get GPU tasks.