Index  | Recent Threads  | Unanswered Threads  | Who's Active  | Guidelines  | Search
 
World Community Grid Forums
  Category: Support
  Forum: GPU Support Forum
  Thread: CPU Benchmarks halt GPU Processing

Quick Go »
No member browsing this thread
Thread Status: Active
Total posts in this thread: 8
[ Jump to Last Post ]
Post new Thread
Author
Previous Thread This topic has been viewed 2266 times and has 7 replies Next Thread
Country Bumkin
Cruncher
Australia
Joined: May 14, 2008
Post Count: 14
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
CPU Benchmarks halt GPU Processing
I have been finding that the CPU benchmark disrupts GPU processing. When the benchmark runs (either automatically or manually activated) all jobs are suspended, and when the benchmark has completed all jobs appear to resume progress but when the GPU job gets to 49.707% or 99.707% the normal short pause continues indefinately. If the system is re-booted and the frozen job was at 99.707% it completes immediately BOINC starts and the result files are uploaded, if I manually suspend/resume the job it falls back to the last checkpoint and then processes to a normal completion.

If I use app_config to run more GPU tasks all are permanently paused by the CPU benchmark.

I have not found any log or error messages which suggest a problem.

CPU jobs are all FAAH.

Operating System is Linux Mint Maya 13, 64 Bit
Mon 31 Dec 2012 13:35:48 EST||Starting BOINC client version 7.0.42 for x86_64-pc-linux-gnu
Mon 31 Dec 2012 13:35:48 EST||log flags: file_xfer, sched_ops, task, checkpoint_debug
Mon 31 Dec 2012 13:35:48 EST||Libraries: libcurl/7.22.0 OpenSSL/1.0.1 zlib/1.2.3.4 libidn/1.23 librtmp/2.3
Mon 31 Dec 2012 13:35:48 EST||Processor: 8 GenuineIntel Intel(R) Core(TM) i7-2600 CPU @ 3.40GHz [Family 6 Model 42 Stepping 7]
Mon 31 Dec 2012 13:35:48 EST||Processor: 8.00 MB cache
Mon 31 Dec 2012 13:35:48 EST||Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush
 dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl
 xtopology nonstop_tsc aperfmperf pni pclmulqd
Mon 31 Dec 2012 13:35:48 EST||OS: Linux: 3.2.0-23-generic
Mon 31 Dec 2012 13:35:48 EST||Memory: 3.85 GB physical, 3.91 GB virtual
Mon 31 Dec 2012 13:35:48 EST||Disk: 33.27 GB total, 26.08 GB free
Mon 31 Dec 2012 13:35:48 EST||Local time is UTC +10 hours
Mon 31 Dec 2012 13:35:48 EST||CUDA: NVIDIA GPU 0: GeForce GTX 560 (driver version unknown, CUDA version 5.0, compute capability 2.1,
 1024MB, 972MB available, 1116 GFLOPS peak)
Mon 31 Dec 2012 13:35:48 EST||OpenCL: NVIDIA GPU 0: GeForce GTX 560 (driver version 304.43, device version OpenCL 1.1 CUDA,
 1024MB, 972MB available, 1116 GFLOPS peak)
Mon 31 Dec 2012 13:35:48 EST|World Community Grid|Found app_config.xml
Mon 31 Dec 2012 13:35:48 EST||Config: GUI RPC allowed from:
Mon 31 Dec 2012 13:35:48 EST||Config:   192.168.1.19
Mon 31 Dec 2012 13:35:48 EST||Config:   192.168.1.11
Mon 31 Dec 2012 13:35:48 EST|World Community Grid|URL http://www.worldcommunitygrid.org/; Computer ID xxxxxxxx; resource share 100
Mon 31 Dec 2012 13:35:48 EST|World Community Grid|General prefs: 
from World Community Grid (last modified 09-Dec-2012 09:09:40)
Mon 31 Dec 2012 13:35:48 EST|World Community Grid|Computer location: school
Mon 31 Dec 2012 13:35:48 EST||General prefs: using separate prefs for school
Mon 31 Dec 2012 13:35:48 EST||Reading preferences override file
Mon 31 Dec 2012 13:35:48 EST||Preferences:
Mon 31 Dec 2012 13:35:48 EST||   max memory usage when active: 3543.60MB
Mon 31 Dec 2012 13:35:48 EST||   max memory usage when idle: 3543.60MB
Mon 31 Dec 2012 13:35:48 EST||   max disk usage: 10.00GB
Mon 31 Dec 2012 13:35:48 EST||   max download rate: 102400 bytes/sec
Mon 31 Dec 2012 13:35:48 EST||   max upload rate: 10240 bytes/sec
Mon 31 Dec 2012 13:35:48 EST||   (to change preferences, visit the web site of an attached project, 
or select Preferences in the Manager)
Mon 31 Dec 2012 13:35:48 EST||Not using a proxy
Mon 31 Dec 2012 13:35:51 EST||Suspending computation - initial delay


app_config is....
<app_config>
 <app>
  <name>hcc1</name>
  <gpu_versions>
   <gpu_usage>1.0</gpu_usage>
   <cpu_usage>1.0</cpu_usage>
  </gpu_versions>
 </app>
</app_config>

----------------------------------------
Regards C Bumkin
----------------------------------------
[Edit 1 times, last edit by Country Bumkin at Dec 31, 2012 5:29:32 AM]
[Dec 31, 2012 5:21:22 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: CPU Benchmarks halt GPU Processing
Hello Country Bumkin,

Interesting. I have reported this to the staff.

Lawrence
[Dec 31, 2012 7:35:23 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: CPU Benchmarks halt GPU Processing
About two hours ago, I just saw unfold before my eyes a BOINC_v7.0.42-intiated CPU-benchmark. Two HCC7.08-GPU-WUs were running. At the time when the benchmark completed, the WU that was (or caught) at 99.707% had nothing indicated for it under the BOINC remaining-time column and appeared frozen, while the one that was less than 99.707% crunched along seemingly unaffected and completed successfully. The frozen WU's elapsed-time kept incrementing by the second. I gave that WU some time and when that WU reached 30-minutes elapsed-time, I opted to suspend and then resume. On resumption, the frozen WU grabbed the thread earlier released at suspension from another HCCv7.08-GPU-WU, then marked that WU as 'Waiting to run', and next used the thread: it then ran back to the start and eventually completed its WU run.

Another instance calling attention to the need for a mechanism to handle interrupted GPU-WUs. One method is for interrupting processes to wait for all running GPU-WUs to first complete, and temporarily hold-off all Ready-to-Start GPU-WUs from running. This method would also protect GPU-WUs from high-priority GPU-WUs.
;
; andzgridPost#782
;
[Jan 3, 2013 8:41:13 PM]   Link   Report threatening or abusive post: please login first  Go to top 
stoneageman
Advanced Cruncher
UK
Joined: Nov 21, 2005
Post Count: 104
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: CPU Benchmarks halt GPU Processing
I've been using Boinctasks rules feature to monitor for over running wu's, then perform a snooze for a few seconds, which restarts them all properly. Had not realised it was the benchmark causing this problem d oh so applause to CB
----------------------------------------
https://xs4s.org/index.php
https://discord.gg/rsCfAgSB
[Jan 5, 2013 8:25:14 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: CPU Benchmarks halt GPU Processing
For those that actually read manuals to see if there's a working solution: Insert in the <options> section of the cc_config.xml the following line:

<skip_cpu_benchmarks>1</skip_cpu_benchmarks>

This will forever postpone any periodic retesting of performance, in fact the manual benchmark test will not execute (MS would call that possible bug a feature).

Benchmarks are a waste of time, in fact, WCG and any other project that has adopted server 700 and up ignores [CPU] benchmarks based claims anyhow. Claims are computed from returned work throughput statistics, then adjusted a second time if it involves e.g. a quorum 2 science by whatever standard or legacy rules that WCG put in place.

Benchmarks used to run every 120 hours. As of some client version 6.xx only when upgrading or restarting, if the restart was longer than 120 hours ago, grosso modo. Not tracking the rules that force them since as said, it's to me a redundant legacy function only of use to projects pré server 700, and on top many of those just give dang anyhow... they make up their own credit awards.
[Jan 5, 2013 8:44:02 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: CPU Benchmarks halt GPU Processing
For those that actually read manuals to see if there's a working solution: ...
That's a nice way to introduce your ideas... Or is it? thinking
;
; andzgridPost#791
;
[Jan 5, 2013 10:11:30 PM]   Link   Report threatening or abusive post: please login first  Go to top 
armstrdj
Former World Community Grid Tech
Joined: Oct 21, 2004
Post Count: 695
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: CPU Benchmarks halt GPU Processing
I will look into this. For anyone experiencing this issue is it on both Windows and Linux?

Thanks,
armstrdj
[Jan 7, 2013 4:21:30 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: CPU Benchmarks halt GPU Processing
For those that actually read manuals to see if there's a working solution: ...
That's a nice way to introduce your ideas... Or is it? thinking
;
; andzgridPost#791
;

It's a working function of BOINC, for those understanding and interested in what's already on offer, *not* an introduction of my ideas to go over-head tongue
[Jan 7, 2013 4:41:07 PM]   Link   Report threatening or abusive post: please login first  Go to top 
[ Jump to Last Post ]
Post new Thread