| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Help Conquer Cancer
Thread: HCC seems to be ignoring suspsend preferences for CPU |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 15
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Just started working HCC work units with my GPU (ATI Mobility Radeon 5000 series). Most of my work units now have the notation 1.00 CPUs + 1.00 ATI GPUs.
However, those work units seem to be ignoring my preference to suspend when user is active. And it's not the GPU part I'm concerned about, but the CPU. In the past, everything has always repected that preference. Now no matter what I do, including suspending the task and suspending the project, the HCC WU continues chweing up CPU processing resources and the progress bar percentage continues increasing. I'm not sure what other details I need to post as this is the first problem I've ever had with WCG in the year I've been runnning it. So if there are details you need, please let me know. |
||
|
|
Bearcat
Master Cruncher USA Joined: Jan 6, 2007 Post Count: 2803 Status: Offline Project Badges: 
|
Disregard
Crunching for humanity since 2007!
---------------------------------------- [Edit 1 times, last edit by Bearcat at Oct 14, 2012 1:47:02 AM] |
||
|
|
Coleslaw
Veteran Cruncher USA Joined: Mar 29, 2007 Post Count: 1343 Status: Offline Project Badges: 
|
Bearcat...the answer to that question is in the details jasonvr presented. If BOINC was installed as a service, then it would not use or detect the GPU in the first place. Since it is detecting, you can rule that out.
----------------------------------------jasonvr, can you tell us all the preferences you have selected? That may help us narrow down the trouble.     |
||
|
|
Bearcat
Master Cruncher USA Joined: Jan 6, 2007 Post Count: 2803 Status: Offline Project Badges:
|
Oopps, my bad. Should have caught that.
----------------------------------------
Crunching for humanity since 2007!
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Bearcat...the answer to that question is in the details jasonvr presented. If BOINC was installed as a service, then it would not use or detect the GPU in the first place. Since it is detecting, you can rule that out. jasonvr, can you tell us all the preferences you have selected? That may help us narrow down the trouble. Here's the run down: Activity Menu Run based on preferences Use GPU based on preferences Preferences-->processor usage While computer is on batteries-->unchecked While computer is in use-->unchecked Use GPU while computer is in use-->checked Only after computer has been idle for 0.50 minutes While processor usage is less than 30 percent Every day between hours of 00:00 and 00:00 No day-of-week overrides set Other options Switch between applications every 120.00 minutes On multiprocessor systems, use at most 60.00% of the processors Use at most 50.00% CPU time I think that is all the relevant settings, but if not, let me know which others you need |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello jasonvr,
When HCC started running on the GPU, one post mentioned that on one system, the HCC task ran on the CPU for 20 seconds of setup, then ran on the GPU for 40 seconds, then ran on the CPU for 40 seconds to create the result to send back. In all, 100 seconds on the system. I assume that you are willing to allow some CPU activity if you allow the GPU to run while the user is active. Lawrence |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Use GPU while computer is in use-->checked ...uncheck it then it stops when you use it...  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello jasonvr, When HCC started running on the GPU, one post mentioned that on one system, the HCC task ran on the CPU for 20 seconds of setup, then ran on the GPU for 40 seconds, then ran on the CPU for 40 seconds to create the result to send back. In all, 100 seconds on the system. I assume that you are willing to allow some CPU activity if you allow the GPU to run while the user is active. Lawrence So I just changed my preferences to not allow GPU while user is active. After the tasks went active due to user inactivity I started a stopwatch and kept up constant activity. After 5 minutes the HCC task was still consuming 13% of my CPU. Since I am on a quad core with hyperthreading, it means the HCC task is essentially taking one entire physical core of the CPU and not releasing it at all, even with constant user activity. In fact, the HCC task is continuing to consume 13% of my CPU as I type this :( All other tasks are properly suspended (Leishmaniasis, HCC w/o GPU, and Malaria) So I don't think your theory is correct |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Use GPU while computer is in use-->checked ...uncheck it then it stops when you use it... I don't think you read the original post correctly. My CPU is being used when the work on the WU should be suspended. I'm perfectly willing to let the GPU be used while I'm active, but the CPU being used is seriously slowing down the computer |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Use GPU while computer is in use-->checked ...uncheck it then it stops when you use it... And don't forget to set the activity menu to "Run based on preferences", for CPU and or GPU. Can though see that the scheduler does not see the CPU portion of a GPU task as CPU processing. Someone, pirogue IIRC, noted that even if LAIM is on, and a GPU task gets paused, on resume it skips back to the start. Don't know why. Suggest TThrottle as a solution. By artificially setting a lower temperature, the crunching will be smooth but uses a lower number of cycles so more are available for the user... which then maybe allows a "Run Always" on the CPU. Moi, I continue to use ThreadMasterGUI [See Start Here FAQs], as that gives me science application level throttling control [some generate more heat than others, where TTHrottle is indiscriminate... measures high temp, so slows all the sciences down]. The GUI portion would maybe still run 100%, but you could set a lower value so you wont notice. Whilst, I've never noticed CPU crunching on an 8 core hyperthreaded. I'm doing it now and there's a CEP2 in amongst the running collective. The TMG service is stopped, and ambient and crunching temps are not high now. Just saying, there are well working tricks and tools. Else if crunching while using interferes, make your choices [Run based on preferences], or don't run it. HCC completed 726,168 results yesterday, which is about 3.25 times as much as 3 days ago. With enough statistics data, we can pretty much predict this science is going to finish much sooner. Even backfill of new work by the scientists wont hold that up. Crunch On. |
||
|
|
|