KWSN Orbiting Fortress Forum Index KWSN Orbiting Fortress
KWSN Distributed Computing Teams forum
 
 FAQFAQ   SearchSearch   MemberlistMemberlist   UsergroupsUsergroups   RegisterRegister 
 ProfileProfile   Log in to check your private messagesLog in to check your private messages   Log inLog in 

Anyone noticed core client 7.0.28 ignoring preferences?

 
Post new topic   Reply to topic    KWSN Orbiting Fortress Forum Index -> KWSN BOINC'ers
View previous topic :: View next topic  
Author Message
Nuadormrac
Prince
Prince


Joined: 13 Sep 2009
Posts: 506

PostPosted: Thu Jul 12, 2012 8:54 pm    Post subject: Anyone noticed core client 7.0.28 ignoring preferences? Reply with quote

OK, so the new laptop arrived, and I have it set to take work for nVidia, but not for CPU from Milkyway. However it downloaded tasks for CPU, and seems to be ignoring project preferences on this. Anyone noticed?

Outside of that difference, the GPU crunching does seem to go a lot faster then the CPU. Though of a somewhat unsuspected surprise, both Windows 7 and BOINC sees 8 CPU cores, not 4. It's an i7 quad core, I guess each core being dual threaded, software sees it as 2 physically seperate ones, so it pulls 8 CPU tasks to crunch at a time.... It does recognize the nVidia GPU, so it's been pulling 9 tasks to run at once... Have sorta wanted to leave the CPU for other projects though which wouldn't have the biggest difference for them based on the GPU....
_________________


.
Back to top
View user's profile Send private message
The King's Own
Baron
Baron


Joined: 09 Jun 2012
Posts: 232
Location: Harbor Springs, Michigan

PostPosted: Thu Jul 12, 2012 10:46 pm    Post subject: Reply with quote

In BOINC Manager Adjust the Tools--Computing Prefernces--Other Options to 75% or 87.5 %. That will leave one or two CPU threads free. Also Ubse Windows Environment Variables set Swan Sync=0 to dedicate the GPU to a core
_________________
What are you going to do, bleed on me?

Back to top
View user's profile Send private message Send e-mail
Concrete-mixing Moose
Prince
Prince


Joined: 30 Apr 2012
Posts: 567
Location: The Joyce Grenfell Home for the Distressed

PostPosted: Thu Jul 12, 2012 11:13 pm    Post subject: Reply with quote

CPUs with HT are seen by Boinc and non-Boinc projects like Folding@home as CPUs with double so many physical cores. Task Manager in Windows also sees double physical cores.
_________________

Save my home - click every day on the picture!
Back to top
View user's profile Send private message Send e-mail
Nuadormrac
Prince
Prince


Joined: 13 Sep 2009
Posts: 506

PostPosted: Thu Jul 12, 2012 11:17 pm    Post subject: Reply with quote

Well it's not that I was looking to not have BOINC run on one of the cores so much, as I was looking to keep Milkyway on the GPU (except for the 0.05% it says it's taking of the CPU), so that the other cores would be dedicated to other projects that don't have a GPU option... But it seems that having set that in the preferences, the manager is ignoring that, and is downloading CPU tasks in addition to GPU tasks...

Quote:
7/13/2012 12:00:45 AM | Milkyway@Home | Sending scheduler request: To fetch work.
7/13/2012 12:00:45 AM | Milkyway@Home | Requesting new tasks for NVIDIA
7/13/2012 12:00:48 AM | Milkyway@Home | Scheduler request completed: got 0 new tasks
7/13/2012 12:00:48 AM | Milkyway@Home | No tasks sent
7/13/2012 12:00:48 AM | Milkyway@Home | This computer has reached a limit on tasks in progress

...7/13/2012 12:06:12 AM | Milkyway@Home | Sending scheduler request: To fetch work.
7/13/2012 12:06:12 AM | Milkyway@Home | Requesting new tasks for CPU
7/13/2012 12:06:15 AM | Milkyway@Home | Scheduler request completed: got 0 new tasks
7/13/2012 12:06:15 AM | Milkyway@Home | No tasks sent
7/13/2012 12:06:15 AM | Milkyway@Home | This computer has reached a limit on tasks in progress
7/13/2012 12:15:40 AM | Milkyway@Home | Computation for task de_separation_09_2s_sample_2_1341007502_7315359_0 finished
7/13/2012 12:15:40 AM | Milkyway@Home | Starting task de_separation_09_2s_sample_2_1341007502_7292423_1 using milkyway version 102 (opencl_nvidia) in slot 9
7/13/2012 12:15:40 AM | Milkyway@Home | Sending scheduler request: To fetch work.
7/13/2012 12:15:40 AM | Milkyway@Home | Reporting 1 completed tasks, requesting new tasks for NVIDIA
7/13/2012 12:15:43 AM | Milkyway@Home | Scheduler request completed: got 1 new tasks


So it's definitely getting tasks for the GPU, but it's requesting tasks for the CPU even though on project preferences I unchecked CPU tasks, and set it to send up GPU tasks only... The manager doesn't seem to be acknowledging, or following this project setting to request nVidia tasks only...
_________________


.
Back to top
View user's profile Send private message
Gemjunkie
Prince
Prince


Joined: 03 Jul 2010
Posts: 3519
Location: Earth, lately

PostPosted: Fri Jul 13, 2012 12:12 am    Post subject: Reply with quote

I'd double check preferences at your MW account, especially if you're using more than one profile (default, home, work, school). I've had unexpected results attaching a new computer where I've forgotten I've been crunching under the home profile for months. Embarassed

Also make sure the option:
Quote:
If no work for selected applications is available, accept work from other applications?

is set to No.
_________________




(older, before split CPID)
Back to top
View user's profile Send private message
Nuadormrac
Prince
Prince


Joined: 13 Sep 2009
Posts: 506

PostPosted: Fri Jul 13, 2012 10:37 am    Post subject: Reply with quote

Yeah the accept other work was also disabled, and this computer is not running in a profile....

http://milkyway.cs.rpi.edu/milkyway/show_host_detail.php?hostid=457135
_________________


.
Back to top
View user's profile Send private message
Gemjunkie
Prince
Prince


Joined: 03 Jul 2010
Posts: 3519
Location: Earth, lately

PostPosted: Fri Jul 13, 2012 10:47 am    Post subject: Reply with quote

I don't think it's the client, though. It shouldn't matter what it asks for, if your preferences at the project say no cpu WUs the server shouldn't send any. Confused
_________________




(older, before split CPID)
Back to top
View user's profile Send private message
Nuadormrac
Prince
Prince


Joined: 13 Sep 2009
Posts: 506

PostPosted: Fri Jul 13, 2012 10:58 am    Post subject: Reply with quote

Yeah, I was shocked myself, but CPU is unticked in the preferences, so I have no idea why the client has been asking for CPU tasks... As in is the project not passing that preference up, or is 7.0.28 not acknowledging it. I do have an overide file set for things like the work cache (a lot easier then going into a project and setting all that, and having to force up update whenever one makes changes), so maybe.... But project specific preferences don't even enter into it there, so it should still acknowledge those.... There's nothing locally set in the tools/computing preferences that touches upon project specific settings...
_________________


.
Back to top
View user's profile Send private message
Gemjunkie
Prince
Prince


Joined: 03 Jul 2010
Posts: 3519
Location: Earth, lately

PostPosted: Fri Jul 13, 2012 11:04 am    Post subject: Reply with quote

From your account page try "Edit Computing preferences" and set

On multiprocessors, use at most 0 or 1processors
On multiprocessors, use at most 0 or 5% of the processors


All I can I think of.
_________________




(older, before split CPID)
Back to top
View user's profile Send private message
Nuadormrac
Prince
Prince


Joined: 13 Sep 2009
Posts: 506

PostPosted: Fri Jul 13, 2012 3:11 pm    Post subject: Reply with quote

I do have a preferences file which would over-ride, but the way it's classified it looks like those settings apply to all BOINC projects. It's not that I don't want BOINC to use the CPUs, I just want BOINC to run othr projects on the CPUs... AKA Milkyway seems to be more optimized for a better return on the GPU, many projects are CPU only or do work that'd be better optimized for the CPU.

At least from the look of it it, setting this wouldn't do what I would want, if it would prevent BOINC from getting CPU tasks from projects other then this one. The note at the top of that page also says it effects all projects, not what I'd want #ni-1 Perhaps if it would still grab GPU tasks with a 0 resource share (so the project couldn't accumulate project debt) that would prevent it from requesting CPU projects, if no other GPU project is active at the time and accepting new tasks, hmm... It depends how it treats resource share, and if it defines it for the CPU only, or not.... Though if the GPU counts towards project debt if it was low enough that the GPU crunched a higher % then necessary it might keep everything else in debt to it, not sure. I did have to bump that back as from when I was crunching it on a CPU and had the opti app for extra credits (years ago) it was set to a 500 resource share....

I guess I could try a resource share of nothing, and see if it still gets GPU tasks, hmm... Not sure that setting effects CPUs only, though or how debt would figure in when a GPU project is compared to a CPU project (they are seperate if they're running on different devices in practice, but does the code treat it that way for scheduleing, hmm)
_________________


.
Back to top
View user's profile Send private message
Gemjunkie
Prince
Prince


Joined: 03 Jul 2010
Posts: 3519
Location: Earth, lately

PostPosted: Sat Jul 14, 2012 10:55 am    Post subject: Reply with quote

Do you use BAM? I don't.

I don't know what else to suggest. I haven't seen that problem unless I had a computer on a wrong profile or had a checkmark in the wrong place. (Except for Einstein, which sends WUs against your settings.) MW works fine for me as a GPU only project. My preferences:



#ni-2
_________________




(older, before split CPID)
Back to top
View user's profile Send private message
Nuadormrac
Prince
Prince


Joined: 13 Sep 2009
Posts: 506

PostPosted: Sat Jul 14, 2012 1:05 pm    Post subject: Reply with quote

I'm beginning to suggest resource share has a lot to do with it.... Initially MW was 500 resource share, and being in a rush when I was setting it up to go grab a bite to eat, aside from MW I had only attached FreeHAL... I was surprised it started getting GPU projects with it set not to, but I was less concerned about the CPU crunching then. I got the computer at 5 pm, had a bunch of stuff to do from video drivers to Windows update, to loading my AV software, so it was after 7 and I didn't want to late longer for dinner. I took someone out for pizza...

Latter I attached another project but it was still getting CPU tasks until I stripped the resource share down to 200 (the other projects were 100).... It's stopped doing that. Only thing, I was in such a rush and slapped something on so quickly I ended up putting on the x86 client; gotta run my queue out now, to get the x64 client. Well a lot of the projects probably don't have an x64 app, but some perhaps do. Also, though it might have been corrected I seem to remember win7 having a slow memory leak wrt 32-bit programs, but specifically on the win64 version of it (the win32 version didn't have that leak)... It was over a year ago I remember reading about that one though.... I do NOT shut down every night, though as it's a laptop I do move it more then I would a desktop. Well anyhow.... MW had such a high resource share because when I last crunched it (when I didn't have a GPU), I was using the opti-app that would complete tasks in like 45 mins or so, and I think they were paying about 200 credits per task, vs the current. So it payed better to run it on the opti-app more, then other stuff. I'm guessing it ignores the selection to not use the CPU if either:

- there are no CPU tasks
- the resource share is such that it's amassed a lot of project debt, and GPU crunching only wouldn't pay it off

Still one would expect that debt on the GPU would be defined different then on the CPU, because they're different crunching devices, and for instance a CPU only project could never get GPU time to help collect on debt owed it by other projects. But perhaps they don't define it that way in their code, hmm....

Things would have to be further tested to see if that pans out. At first I was like "oh bother, 7.0.28 has some glitch); wouldn't be the first time, the bug fix list has addressed issues dealing with preferences in various builds before. 7.0.31 woulda been the next thing to try if something didn't adjust....
_________________


.
Back to top
View user's profile Send private message
Pooh Bear 27
Prince
Prince


Joined: 28 Jan 2005
Posts: 1358
Location: Fond du Lac, WI

PostPosted: Sat Jul 14, 2012 1:09 pm    Post subject: Reply with quote

I found that GPU projects can now steal CPU cores in version 7+. I am running POEM on GPU only, but my CPU project I had my CPUs set to 50%. Well, I was only running 3 CPU projects with the GPU project instead of 4 plus. I had to change preferences to set the percentage as 5 CPUs to get 4 CPUs plus the GPU to run.

Kind of bothersome, but at least I have it figured out.
Back to top
View user's profile Send private message
Gemjunkie
Prince
Prince


Joined: 03 Jul 2010
Posts: 3519
Location: Earth, lately

PostPosted: Sat Jul 14, 2012 1:26 pm    Post subject: Reply with quote

Pooh Bear 27 wrote:
I found that GPU projects can now steal CPU cores in version 7+. I am running POEM on GPU only, but my CPU project I had my CPUs set to 50%. Well, I was only running 3 CPU projects with the GPU project instead of 4 plus. I had to change preferences to set the percentage as 5 CPUs to get 4 CPUs plus the GPU to run.

Kind of bothersome, but at least I have it figured out.


I understood none of that. Confused
_________________




(older, before split CPID)
Back to top
View user's profile Send private message
Gemjunkie
Prince
Prince


Joined: 03 Jul 2010
Posts: 3519
Location: Earth, lately

PostPosted: Sat Jul 14, 2012 1:30 pm    Post subject: Reply with quote

Nuadormrac wrote:
I'm beginning to suggest resource share has a lot to do with it.... Initially MW was 500 resource share, and being in a rush when I was setting it up to go grab a bite to eat, aside from MW I had only attached FreeHAL... I was surprised it started getting GPU projects with it set not to, but I was less concerned about the CPU crunching then. I got the computer at 5 pm, had a bunch of stuff to do from video drivers to Windows update, to loading my AV software, so it was after 7 and I didn't want to late longer for dinner. I took someone out for pizza...

Latter I attached another project but it was still getting CPU tasks until I stripped the resource share down to 200 (the other projects were 100).... It's stopped doing that. Only thing, I was in such a rush and slapped something on so quickly I ended up putting on the x86 client; gotta run my queue out now, to get the x64 client. Well a lot of the projects probably don't have an x64 app, but some perhaps do. Also, though it might have been corrected I seem to remember win7 having a slow memory leak wrt 32-bit programs, but specifically on the win64 version of it (the win32 version didn't have that leak)... It was over a year ago I remember reading about that one though.... I do NOT shut down every night, though as it's a laptop I do move it more then I would a desktop. Well anyhow.... MW had such a high resource share because when I last crunched it (when I didn't have a GPU), I was using the opti-app that would complete tasks in like 45 mins or so, and I think they were paying about 200 credits per task, vs the current. So it payed better to run it on the opti-app more, then other stuff. I'm guessing it ignores the selection to not use the CPU if either:

- there are no CPU tasks
- the resource share is such that it's amassed a lot of project debt, and GPU crunching only wouldn't pay it off

Still one would expect that debt on the GPU would be defined different then on the CPU, because they're different crunching devices, and for instance a CPU only project could never get GPU time to help collect on debt owed it by other projects. But perhaps they don't define it that way in their code, hmm....

Things would have to be further tested to see if that pans out. At first I was like "oh bother, 7.0.28 has some glitch); wouldn't be the first time, the bug fix list has addressed issues dealing with preferences in various builds before. 7.0.31 woulda been the next thing to try if something didn't adjust....

I always include
<zero_debts>1</zero_debts>
in my cc_config.xml file, but I heard version 7.xx doesn't support it. Another huge reason to stick with 6.xx. Twisted Evil
_________________




(older, before split CPID)
Back to top
View user's profile Send private message
Pooh Bear 27
Prince
Prince


Joined: 28 Jan 2005
Posts: 1358
Location: Fond du Lac, WI

PostPosted: Sat Jul 14, 2012 4:19 pm    Post subject: Reply with quote

Gemjunkie wrote:
I understood none of that. Confused

CPU set to 50% on my I7 to use 4 of 8. Problem is POEM is setting their GPU at a threshold on the server side so that takes a whole CPU core as a default so now I was only running 3 CPUs and 1 GPU whereas any other project I set the same settings and I still run 4 CPUs and the 1 GPU. So I had to set to 62.5% to get 4 CPUs and the POEM GPU to run simultaneously.
_________________
Watch the movie about me The 4-1-4s: The Original Teenage Hackers (It is only about 12 minutes long.)
My lucky numbers are 121*2^4553899-1 and 3756801695685*2^666669±1


Back to top
View user's profile Send private message
Gemjunkie
Prince
Prince


Joined: 03 Jul 2010
Posts: 3519
Location: Earth, lately

PostPosted: Sat Jul 14, 2012 5:08 pm    Post subject: Reply with quote

I see, other projects mostly use the 4 cores and Poem effectively claims one hyperthread "core". In effect, if not technically accurate?


#ni-1
_________________




(older, before split CPID)
Back to top
View user's profile Send private message
Nuadormrac
Prince
Prince


Joined: 13 Sep 2009
Posts: 506

PostPosted: Sat Jul 14, 2012 7:56 pm    Post subject: Reply with quote

Gemjunkie wrote:
Nuadormrac wrote:
I'm beginning to suggest resource share has a lot to do with it.... Initially MW was 500 resource share, and being in a rush when I was setting it up to go grab a bite to eat, aside from MW I had only attached FreeHAL... I was surprised it started getting GPU projects with it set not to, but I was less concerned about the CPU crunching then. I got the computer at 5 pm, had a bunch of stuff to do from video drivers to Windows update, to loading my AV software, so it was after 7 and I didn't want to late longer for dinner. I took someone out for pizza...

Latter I attached another project but it was still getting CPU tasks until I stripped the resource share down to 200 (the other projects were 100).... It's stopped doing that. Only thing, I was in such a rush and slapped something on so quickly I ended up putting on the x86 client; gotta run my queue out now, to get the x64 client. Well a lot of the projects probably don't have an x64 app, but some perhaps do. Also, though it might have been corrected I seem to remember win7 having a slow memory leak wrt 32-bit programs, but specifically on the win64 version of it (the win32 version didn't have that leak)... It was over a year ago I remember reading about that one though.... I do NOT shut down every night, though as it's a laptop I do move it more then I would a desktop. Well anyhow.... MW had such a high resource share because when I last crunched it (when I didn't have a GPU), I was using the opti-app that would complete tasks in like 45 mins or so, and I think they were paying about 200 credits per task, vs the current. So it payed better to run it on the opti-app more, then other stuff. I'm guessing it ignores the selection to not use the CPU if either:

- there are no CPU tasks
- the resource share is such that it's amassed a lot of project debt, and GPU crunching only wouldn't pay it off

Still one would expect that debt on the GPU would be defined different then on the CPU, because they're different crunching devices, and for instance a CPU only project could never get GPU time to help collect on debt owed it by other projects. But perhaps they don't define it that way in their code, hmm....

Things would have to be further tested to see if that pans out. At first I was like "oh bother, 7.0.28 has some glitch); wouldn't be the first time, the bug fix list has addressed issues dealing with preferences in various builds before. 7.0.31 woulda been the next thing to try if something didn't adjust....

I always include
<zero_debts>1</zero_debts>
in my cc_config.xml file, but I heard version 7.xx doesn't support it. Another huge reason to stick with 6.xx. Twisted Evil


Yeah, but this laptop uses nVidia Optimus to switch between the built in Intel gfx on the CPU and the nVidia gfx chip... I've heard that the 6.x client doesn't play nicely as it can't detect Optimus properly and hence only sees the Intel gfx (which isn't CUDA/OpenCL capeable)... So it's 7.x to play on this laptop it appears....
_________________


.
Back to top
View user's profile Send private message
Gemjunkie
Prince
Prince


Joined: 03 Jul 2010
Posts: 3519
Location: Earth, lately

PostPosted: Sat Jul 14, 2012 11:29 pm    Post subject: Reply with quote

Between a rock and a hard place. #Mad
_________________




(older, before split CPID)
Back to top
View user's profile Send private message
Display posts from previous:   
Post new topic   Reply to topic    KWSN Orbiting Fortress Forum Index -> KWSN BOINC'ers All times are GMT - 5 Hours
Page 1 of 1

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum


Powered by phpBB © 2001, 2005 phpBB Group
Optimized Seti@Home App | BOINC Stats