| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: OpenPandemics - COVID-19 Project
Thread: Crunching time doubled/trebled? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 18
|
|
| Author |
|
|
Sirius B
Advanced Cruncher Joined: Dec 11, 2007 Post Count: 67 Status: Offline Project Badges: 
|
For some time now, my dual core has been getting tasks that have been near enough treble what they used to be. They had been anywhere from 2.5 to 4 hours. Now they are all 7.75 hours.
I don't consider this an issue, however, on seeing that most, if not all take over 8 hours yet only get credited with 4. I do consider this to be a waste of money as it is costing power which I & not the project is paying for. It is currently running the last batch whose deadline is 31/12. I will download another batch after they complete. Should this continue to be the case, that batch will be completed, then rig detached. |
||
|
|
l_mckeon
Senior Cruncher Joined: Oct 20, 2007 Post Count: 439 Status: Offline Project Badges: 
|
All my Covid tasks are largely unchanged, which implies something wrong with your system.
If you're on Windows start up Task Manager and see what's using your CPU. If your system runs 24/7 it might be a good time for a cold reboot. |
||
|
|
William Albert
Cruncher Joined: Apr 5, 2020 Post Count: 41 Status: Offline Project Badges: 
|
I haven't noticed a change in the amount of time spent crunching OPN1 units across my fleet.
You may want to pause your crunching for a bit to see if anything else is consuming a significant amount of CPU time. Not getting credited for a significant amount of the time your WUs are running is a sign that the WUs are having to split CPU time with something else. |
||
|
|
alanb1951
Veteran Cruncher Joined: Jan 20, 2006 Post Count: 1317 Status: Offline Project Badges: 
|
@Sirius B
There is always some variability in run times, as you may have noticed even before you started to see this issue; however, what you are seeing definitely doesn't sound right, so I hope it can be resolved! As you didn't highlight a big difference between CPU time and elapsed time, I'm presuming that you are referring to CPU time increases, not just clock-on-the-wall increases! You say it's on a dual core system (presumably no hyperthreading); if so, are you running two BOINC tasks at a time? If you are, try only letting BOINC have one core to play with and see if it makes a significant difference - you might see quite a change in run-time (though that will depend on all sorts of things about your specific hardware!), and if you do it will provide a useful diagnostic. If you have already tried that, please accept my apologies for suggesting it... Good luck - Al. P.S. Not the same as it was at SETI@home, is it! |
||
|
|
Bryn Mawr
Senior Cruncher Joined: Dec 26, 2018 Post Count: 384 Status: Offline Project Badges:
|
I do consider this to be a waste of money as it is costing power which I & not the project is paying for. It is currently running the last batch whose deadline is 31/12. I will download another batch after they complete. Should this continue to be the case, that batch will be completed, then rig detached. Sorry if I’m being dense but I don’t see the logic of this. Your rig runs for as long as it runs - your costs have not changed. It is possibly the project that’s loosing out if they are really getting less scientific results back but how are you loosing? [Edit 1 times, last edit by Bryn Mawr at Dec 28, 2020 6:00:28 AM] |
||
|
|
sam6861
Advanced Cruncher Joined: Mar 31, 2020 Post Count: 107 Status: Offline Project Badges:
|
CPU overheating and throttling to very slow speed?
One year ago, my Ryzen 7 2700X Windows 10 went slow with tasks that took twice as long CPU time. Windows 10 task manager shows only about 50% CPU usage, with about 1.8 GHz speed. Eventually I check HWinfo64, shows CPU temperature at 105 C. Broken plastic hook for heatsink mount, changed plastic mount and re-attached heatsink. Now runs full speed with fast tasks, 3.7 GHz, CPU temperature 76 C. |
||
|
|
Sirius B
Advanced Cruncher Joined: Dec 11, 2007 Post Count: 67 Status: Offline Project Badges:
|
Sorry if I’m being dense but I don’t see the logic of this. Your rig runs for as long as it runs - your costs have not changed. It is possibly the project that’s loosing out if they are really getting less scientific results back but how are you loosing? Run time for starters. I'm well aware that there is a difference between cpu & elapsed times. On my rigs, that difference has been anywhere from 2 to 30 minutes. Over 50% difference is wasteful.The dual core was just an everyday rig for part time crunching & net access. With seti's long task deadlines, its 10/12 hour crunching days was okay. Entering our first lockdown, switched it to 24/7 because of Covid. 1st 6 tasks of 8 core: 2.88/2.91 - 2.80/2.82 - 2.75/2.78 2.73/2.75 - 2.90/2.91 - 2.89/2.91 1st 6 tasks of 2 core: 4.15/8.45 - 4.13/8.46 - 4.00/8.10 4.07/8.27 - 4.17/8.47 - 4.14/8.47 It costs to run inefficient equipment & that goes for anyone. |
||
|
|
PMH_UK
Veteran Cruncher UK Joined: Apr 26, 2007 Post Count: 786 Status: Offline Project Badges: 
|
That is consistent with something making 1 core near 100% busy, leaving 1 to do 2 BOINC tasks plus other windows tasks.
----------------------------------------As others have said check Task Manager to see what is taking CPU. Paul.
Paul.
|
||
|
|
Sirius B
Advanced Cruncher Joined: Dec 11, 2007 Post Count: 67 Status: Offline Project Badges:
|
All my Covid tasks are largely unchanged, which implies something wrong with your system. Thanks, that's what I was thinking, so rebooted. Issue came to light. Not a hardware problem but a Win 10 issue. On restarting, got a black screen with cursor, so rebooted. Managed to login. Clicked on task manager to see what the problem was & got black screen again. Tried several more times, but no joy so shutdown.If you're on Windows start up Task Manager and see what's using your CPU. If your system runs 24/7 it might be a good time for a cold reboot. Tried again approx. an hour later, no problem. 1st 6 tasks for today completed: 3.29/3.43 - 3.33/3.51 - 3.48/3.54 3.45/3.52 - 3.31/3.37 - 3.65/3.72 So far so good. Will monitor it for a few days. Personally, I do not like win 10. :-( |
||
|
|
Bryn Mawr
Senior Cruncher Joined: Dec 26, 2018 Post Count: 384 Status: Offline Project Badges:
|
Sorry if I’m being dense but I don’t see the logic of this. Your rig runs for as long as it runs - your costs have not changed. It is possibly the project that’s loosing out if they are really getting less scientific results back but how are you loosing? Run time for starters. I'm well aware that there is a difference between cpu & elapsed times. On my rigs, that difference has been anywhere from 2 to 30 minutes. Over 50% difference is wasteful.The dual core was just an everyday rig for part time crunching & net access. With seti's long task deadlines, its 10/12 hour crunching days was okay. Entering our first lockdown, switched it to 24/7 because of Covid. 1st 6 tasks of 8 core: 2.88/2.91 - 2.80/2.82 - 2.75/2.78 2.73/2.75 - 2.90/2.91 - 2.89/2.91 1st 6 tasks of 2 core: 4.15/8.45 - 4.13/8.46 - 4.00/8.10 4.07/8.27 - 4.17/8.47 - 4.14/8.47 It costs to run inefficient equipment & that goes for anyone. 30 minutes is excessive but not enough to explain a doubling of run time but from the last part of your post is that between two different machines? It is quite possible that an older dual core processor will take twice the time of a more modern 8 core. It would be interesting to know the CPU types involved. If the CPU differs markedly from elapsed then either the CPU is doing something else, whether Boinc or other or the Boinc tasks are being suspended and the CPU is semi idling. So either you need to find what is using the CPU and causing the inefficiency or which setting is suspending the processing. |
||
|
|
|