Unusual Milkyway behavior (at least to me)

Message boards : Number crunching : Unusual Milkyway behavior (at least to me)
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile George Project Donor
Volunteer tester
Avatar

Send message
Joined: 23 Oct 17
Posts: 154
Credit: 2,040,395
RAC: 21,868
United States
Message 2014749 - Posted: 9 Oct 2019, 15:07:02 UTC

I have an Intel Core 990X CPU with 6 cores and 12 threads with an EVGA RTX 2060 GPU and all has been running fine with BOINC. I have my preferences set so that I use 10 threads for BOINC leaving 2 threads to run anything else I want to. Lately I noticed Milkyway using all 10 threads for 1 N-Body simulation for a total of ~22 1/2 hours, and this is listed on a single line in BOINC Manager. All other projects in active tasks are "waiting to run", and surprisingly no SETI listed and no project is using GPU for any calculations. FWIW, I have SETI, Milkyway and Einstein as my only projects.

Comments?
George

ID: 2014749 · Report as offensive     Reply Quote
Profile tullio
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 7785
Credit: 2,910,026
RAC: 413
Italy
Message 2014762 - Posted: 9 Oct 2019, 16:20:05 UTC
Last modified: 9 Oct 2019, 16:25:42 UTC

Milkyway is using all 4 cores of my A10-6700 CPU. All other tasks are suspended. Only a GPUGRID GPU task is running on my GTX 1050 Ti board.
Tullio
ID: 2014762 · Report as offensive     Reply Quote
Profile Keith Myers Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 10244
Credit: 1,000,642,842
RAC: 1,419,066
United States
Message 2014768 - Posted: 9 Oct 2019, 18:04:54 UTC

Milkyway recently deprecated the single core N-body app leaving only the multi-core N-body cpu app for crunching. Unless you restrict the number of cores for the app, it will monopolize all available cpu cores. You can limit the number of cores used by setting a value in app_config. Set a mt plan class, and match the <avg_ncpus>x</avg_ncpus> to the nthreads value to how many cpu cores you want to restrict the app to.



<app_config>
<app>
<name>milkyway_nbody</name>
<max_concurrent>1</max_concurrent>
</app>
<app_version>
<app_name>milkyway_nbody</app_name>
<plan_class>mt</plan_class>
<avg_ncpus>8</avg_ncpus>
<cmdline>--nthreads 8</cmdline>
</app_version>
</app_config>


Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 2014768 · Report as offensive     Reply Quote
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 13267
Credit: 164,558,526
RAC: 216,553
United Kingdom
Message 2014769 - Posted: 9 Oct 2019, 18:08:01 UTC - in response to Message 2014749.  

An N-Body application with the (MT) multi-threaded plan class will normally be set to use all BOINC's available cores - 10 in your case.

A GPU task using less than one full CPU thread could be scheduled to run alongside, but it won't run if it needs a full CPU. What you've seen is correct, by the design specs.

You could fine-tune either project's settings with an app_config.xml file for either or both - set N-Body / MT to use say 8 CPUs, and give it an '--nthreads 8' command line, or reduce the CPU demand for SETI tasks so that they run, but use the two spare cores you've set aside for 'anything else'.

Have the User Manual open beside you while you work it out.
ID: 2014769 · Report as offensive     Reply Quote
Profile tullio
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 7785
Credit: 2,910,026
RAC: 413
Italy
Message 2014771 - Posted: 9 Oct 2019, 18:32:23 UTC

My GPUGRID GPU task is using 0.966 CPU. My A10-6700 is sold as 4 cores, but Task Manager says it has two cores and and 4 logical processors (threads?). Since I am running Milkyway@home as a Science United user I am unable to write an app_config.xml file, I am not allowed to do this since I have no password for Milkyway@home, only a single Science United password. But I have completed more than one thousand Milkyway tasks, both CPU and GPU, with a very small failure percentage.
Tullio
ID: 2014771 · Report as offensive     Reply Quote
Profile George Project Donor
Volunteer tester
Avatar

Send message
Joined: 23 Oct 17
Posts: 154
Credit: 2,040,395
RAC: 21,868
United States
Message 2014797 - Posted: 9 Oct 2019, 22:43:39 UTC - in response to Message 2014769.  

To Richard and Keith, thanks for your insight into the recent changes with Milkyway@home. I was unaware of the changes but I understand them now. I will look into and make necessary changes to my <app_config.xml> file when I get my 3950X CPU from AMD which will run only SETI on the Linux machine. The current machine is a Windows 10 machine which will then run only Milkyway and Einstein, and I'll only be running those two in Windows 10.
George

ID: 2014797 · Report as offensive     Reply Quote
Profile tullio
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 7785
Credit: 2,910,026
RAC: 413
Italy
Message 2014840 - Posted: 10 Oct 2019, 8:19:29 UTC
Last modified: 10 Oct 2019, 8:26:13 UTC

I still have one Milkyway@home task running along one Seti@home and two Asteroids@home plus a GPU task of GPUGRID. All this on a 4 Logical processors A10-6700 CPU.
Tullio
ID: 2014840 · Report as offensive     Reply Quote
Profile tullio
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 7785
Credit: 2,910,026
RAC: 413
Italy
Message 2014845 - Posted: 10 Oct 2019, 9:09:33 UTC

I have 4 one core (or logical processor) Milkyway@home tasks running on my A10-6700 CPU plus one GPUGRID running on the GTX !050 Ti. All this on a Windows 8.1 PC which hosts also a VirtualBox Linux SuSE Tumbleweed also enlisted in Science United.
Tullio
ID: 2014845 · Report as offensive     Reply Quote
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 3869
Credit: 229,342,399
RAC: 354,096
United States
Message 2015119 - Posted: 11 Oct 2019, 23:46:39 UTC - in response to Message 2014840.  

I still have one Milkyway@home task running along one Seti@home and two Asteroids@home plus a GPU task of GPUGRID. All this on a 4 Logical processors A10-6700 CPU.
Tullio


To make life more complicated that cpu model has one floating point node for each physical node. Which means heavy computation on 2 threads on the same core runs more than a little bit slower....

I had an A10-6700 (I think) and experienced all that under Window....

Tom
A proud member of the OFA (Old Farts Association)
"Over the hill? WHAT Hill? I don't REMEMBER any hill...." (from a bumper sticker I bought at a truck stop).
ID: 2015119 · Report as offensive     Reply Quote
Profile tullio
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 7785
Credit: 2,910,026
RAC: 413
Italy
Message 2015159 - Posted: 12 Oct 2019, 10:02:22 UTC

I am running ScienceUnited on that CPU. The only errors I had were GPU Asteroids@home. Then something happened and they run well with no errors. Speed is not important to me as long as tasks complete and validate. I am getting a weekly mail from Dave Anderson on my progress.
Tullio
ID: 2015159 · Report as offensive     Reply Quote
Profile tullio
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 7785
Credit: 2,910,026
RAC: 413
Italy
Message 2015185 - Posted: 12 Oct 2019, 17:30:12 UTC

The only report I get every week from Science United is the number of tasks I have done and the time it has taken. My last week reported less than one hour/task on my A10-6700 CPU with 4 logical processors. Of course this depends on the number of GPU tasks I am running on my GTX 1050 Ti.
Tullio
ID: 2015185 · Report as offensive     Reply Quote

Message boards : Number crunching : Unusual Milkyway behavior (at least to me)


 
©2019 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.