| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
| World Community Grid Forums
|
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 16
|
|
| Author |
|
|
3rkko
Advanced Cruncher Finland Joined: Aug 2, 2008 Post Count: 105 Status: Offline Project Badges:
|
I have always let Boinc use 100% of resources and I have only sometimes had problems that were caused by Boinc. Now that I have set MIP as the only project it seems that my computer is sometimes less responsive and some things that used to be instantaneous now take up to several seconds. Suspending Boinc solves these problems, so it’s clearly caused by MIP.
----------------------------------------![]() |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The short answer is probably "Yes". If you look through this forum you will find advice on limiting the number of MIP WUs that can run concurrently. Some work is being done on the app which may improve things, but no timetable has been disclosed.
|
||
|
|
3rkko
Advanced Cruncher Finland Joined: Aug 2, 2008 Post Count: 105 Status: Offline Project Badges:
|
Thanks! I'll try limiting the number of cores available to Boinc.
----------------------------------------![]() |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
You can use BOINC control files to limit the number of a particular sub-project's WUs that run concurrently, but still let others run freely. You don't have to run just one sub-project, you can mix them up so that you're still supporting WCG on all your machine's threads. This is more advanced stuff and you'd be better to do some reading rather than rely on me to tell you what or how to do it. If you don't want to dig that deep, maybe just don't select the MIP sub-project at all?
|
||
|
|
Seoulpowergrid
Veteran Cruncher Joined: Apr 12, 2013 Post Count: 823 Status: Offline Project Badges:
|
Your system also might not have enough ram available to it. If you look at the system requirements page (link) you can see that MIP requires more computer memory than other projects; currently at 650 megs per WU compared to 250 per other sub projects, with exception to MCM which needs 400 megs per. I ran into this problem myself.
----------------------------------------![]() |
||
|
|
3rkko
Advanced Cruncher Finland Joined: Aug 2, 2008 Post Count: 105 Status: Offline Project Badges:
|
>650mb ram per wu
----------------------------------------Thanks! That must be the reason for my problems. ![]() |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
>650mb ram per wu Thanks! That must be the reason for my problems. Honestly, I do not think that is the reason. From my experience, every MIP unit uses about 300-350 MB RAM, 650 MB seems more of a worst case number. How many MIP units do you run simultaniously on which type of CPU with how much RAM? I would guess the issue has more to do with MIP using high amounts of L3 Cache (see official statement here ). If you run too many MIP units at once, Cache is not sufficient, which slows down processing of MIP units dramatically, and might also have an impact on other applications running. A good solution is, to limit the number of MIP units to one per 4 MB L3 cache your CPU has. For most modern AMD and Intel CPUs that translates to 1/4th of the overall threads running MIP, the other threads can run other projects. (see an observation I posted here ). To achieve that, there are two solutions. The easy one: just leave all projects activated, that ensures you do not have too many MIP units most of the time. The better, but less-easy one: Limit the number of MIP units running at once by editing your app_config.xml in the BOINC data directory. Here you can find the code to add to the app_config.xml to limit the number of MIP units processed simultaniously. Unfortunately, when using this, you will have to micro-manage BOINC from time to time, as it sometimes happens, that your queue fills up with MIP units and your machine cannot then be utilized 100%, as the number of MIP units to process is limited. So you have to enable/disable MIP in the device profile from time to time to get enough MIP units, but not too many. |
||
|
|
Macromancer
Veteran Cruncher United States Joined: Sep 6, 2016 Post Count: 994 Status: Offline Project Badges:
|
How many MIP units do you run simultaniously on which type of CPU with how much RAM? Although this question isn't directed at me specifically, I am following this discussion because I primarily run MIP1 tasks. All of my PCs are running Win10 Pro. I have a Ryzen 5 1600, with 16GB of RAM (two 8GB DDR4 2667 sticks). I have a mild overclock with all 6 cores running at 3.65GHz. I run 6 cores/12 threads at 100% 24/7. Typically 75 - 80% of my 12 threads are running MIP1 jobs. Based on recent results returned, it takes on average 2.18 hrs for this PC to complete a MIP task. What is the average time to complete a MIP task on your R7 1700? (I'm guessing Linux is much faster than Win10, but you have only 4GB of Ram, so curious). FWIW, I haven't noticed any slowdown in web browsing or in MS Office applications while all of these BOINC related jobs are running on this PC. I also run an i5 4670K at 4.0GHz with 16GB of DDR3 RAM. I run 4 cores/4 threads at 100% 24/7. Typically 75% of my 4 threads are running MIP1 jobs. Based on recent results returned, it takes on average 1.17 hrs for this PC to complete a MIP task. I do notice a slow down in MS Office/browsing when running BOINC at 100% on this PC. [Edit 1 times, last edit by Macromancer at May 3, 2018 11:46:40 AM] |
||
|
|
3rkko
Advanced Cruncher Finland Joined: Aug 2, 2008 Post Count: 105 Status: Offline Project Badges:
|
I would guess the issue has more to do with MIP using high amounts of L3 Cache (see official statement here ). Thanks! I hadn’t looked at that thread because I don’t care about credits. L3 cache choke might be the right answer. I have i7-3770 just like you, but OS is Win10. It’s also good to see that the project scientists know about this problem and are working on to solve it! ![]() |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7846 Status: Offline Project Badges:
|
I have a Ryzen 5 1600, with 16GB of RAM (two 8GB DDR4 2667 sticks). I have a mild overclock with all 6 cores running at 3.65GHz. I run 6 cores/12 threads at 100% 24/7. Typically 75 - 80% of my 12 threads are running MIP1 jobs. Based on recent results returned, it takes on average 2.18 hrs for this PC to complete a MIP task I would have thought you would do somewhat better with that machine. I have an older 4 core Q6600 running stock with 2GB of memory running Windows Vista. I am averaging 3.13 hours per work unit. Granted, there is some variability in the run times of 1.89 to 4.27 hours. I can undestand if you got a rash of the longer running units. Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
|