Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
World Community Grid Forums
Category: Support Forum: GPU Support Forum Thread: Getting way too many CPU tasks |
No member browsing this thread |
Thread Status: Active Total posts in this thread: 9
|
Author |
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Since starting to run GPU tasks I have been getting way too many CPU tasks along with them.
----------------------------------------I have DDDT2 and DSFL selected and allow other tasks additional work buffer was set to 0.5 days, and yesterday I had to abort about 150 DSFL WU's, so I set buffer to 0.05 days, tonight I had to abort about another 150. Trying 0 days buffer now. Don't see it posted that anyone else is having this problem, perhaps I missed it, or have something set wrong elsewhere ? The number of CPU tasks is enough to stop GPU work and run all the CPU tasks at high priority,other than that, it runs great when not flooded with WU's ---- I thought maybe benchmarks were screwy, but seem ok 11/11/2012 6:54:27 PM 2806 floating point MIPS (Whetstone) per CPU 11/11/2012 6:54:27 PM 5855 integer MIPS (Dhrystone) per CPU and system startup messages i showing hardware and BOINC version 11/11/2012 6:51:53 PM Starting BOINC client version 6.10.58 for windows_x86_64 11/11/2012 6:51:53 PM log flags: file_xfer, sched_ops, task 11/11/2012 6:51:53 PM Libraries: libcurl/7.19.7 OpenSSL/0.9.8l zlib/1.2.3 11/11/2012 6:51:53 PM Data directory: C:\ProgramData\BOINC 11/11/2012 6:51:53 PM Running under account Steve 11/11/2012 6:51:53 PM Processor: 4 GenuineIntel Intel(R) Core(TM)2 Quad CPU Q8400 @ 2.66GHz [Family 6 Model 23 Stepping 10] 11/11/2012 6:51:53 PM Processor: 2.00 MB cache 11/11/2012 6:51:53 PM 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 htt tm pni ssse3 cx16 sse4_1 nx lm vmx tm2 pbe 11/11/2012 6:51:53 PM OS: Microsoft Windows 7: x64 Edition, Service Pack 1, (06.01.7601.00) 11/11/2012 6:51:53 PM Memory: 8.00 GB physical, 8.01 GB virtual 11/11/2012 6:51:53 PM Disk: 29.70 GB total, 2.65 GB free 11/11/2012 6:51:53 PM Local time is UTC -7 hours 11/11/2012 6:51:54 PM NVIDIA GPU 0: GeForce GTX 260 (driver version 29573, CUDA version 4020, compute capability 1.3, 896MB, 560 GFLOPS peak) 11/11/2012 6:51:54 PM NVIDIA GPU 1: GeForce GTX 260 (driver version 29573, CUDA version 4020, compute capability 1.3, 896MB, 537 GFLOPS peak) 11/11/2012 6:51:54 PM No general preferences found - using BOINC defaults 11/11/2012 6:51:54 PM Preferences: 11/11/2012 6:51:54 PM max memory usage when active: 4095.25MB 11/11/2012 6:51:54 PM max memory usage when idle: 7371.44MB 11/11/2012 6:51:54 PM max disk usage: 2.55GB 11/11/2012 6:51:54 PM don't use GPU while active 11/11/2012 6:51:54 PM suspend work if non-BOINC CPU load exceeds 25 % 11/11/2012 6:51:54 PM (to change preferences, visit the web site of an attached project, or select Preferences in the Manager) 11/11/2012 6:51:54 PM Found project_init.xml for http://www.worldcommunitygrid.org/ [Edit 4 times, last edit by Former Member at Nov 15, 2012 5:31:09 AM] |
||
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges: |
11/11/2012 6:51:53 PM Starting BOINC client version 6.10.58 for windows_x86_64 Upgrade to v7.0.28 or later and your problem with too much work should go away. "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Your ADDITIONAL work buffer was set to 0.5, but what is yourMAIN one set to?
|
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Steven,
If you open up the BOINC Data Dir, client_state.xml and search for the first Worldcommunitygrid line, then after the first <duration_correction_factor> entry, what is the value that has? A client that crunches exactly as per benchmark usually has one somewhere around 1.000000. (Mine has 1.46 ATM due the elongation of the HCC I guess or some other tasks that ran longer than originally projected]. At any rate, the maximum "In Progress" is 40 per CPU processor core [160 for a quad]. Can't remember the GPU exactly, but think that was latest set to ~250. Yes, upgrading to 7 lets the problem go away [you cant go back to 6 without loosing work that's not finished/reported]. A 0.0 connect and 0.0 additional for 6.10.58 solves it too. Whether 7.0.28 or 7.0.38 [latest beta], no matter how you look at upgrading to 7, a new "recommended" will arrive in the near future... one that also handles properly on W8 [WCG is testing and security auditing the code momentarily]. W* specifically requires 7.0.31 to install correctly, which should be done whilesigned in under a local account, *not* a Microsoft account. |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
P.S. Upgrading... as so often happens, Berkeley is off the air again [noticed this yesterday too]. It's the presumed reason why WCG started to host their own skinned kit. Try later.
|
||
|
tmedve
Senior Cruncher USA Joined: Nov 16, 2004 Post Count: 182 Status: Offline Project Badges: |
Now I am confused. I know how to set the "additional work buffer", but where does one set the "main work buffer". I did not realize that there were 2 separate things or that they were really not the same thing. Thanks.
---------------------------------------- |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The "Connect every..." in client 6 or earlier or "Minimum work buffer" in client 7 is the primary buffer driver. The "[Max] Additional work buffer" is secondary, where the functionality was changed between version 6 and 7 [The MinB/MaxAB extensively explained in a draft FAQ, I think to be lurking in the Website support forum]. In a nutshell, the MaxAB in v7 will not be regarded until the cache drops below the MinB. In practice, I've discontinued using that option... set it to 0.0. Just MinB takes care of a continuous backfill [when the selected projects have work available].
|
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Actually the draft FAQ describing how it was, and how it will be / is [at the time] sits in the BOINC Support forum: https://secure.worldcommunitygrid.org/forums/wcg/viewthread_thread,32530
|
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Thank you all for the help. The connect every was 0.10 days, and it has probably changed since I posted, but have been pretty hectic, now the dcf is <duration_correction_factor>1.435338</duration_correction_factor>
Seems to be ok with 0 days cache, I wil upgrade the client this weekend. Thanks ;) |
||
|
|