| 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: 117
|
|
| Author |
|
|
Aurum
Master Cruncher The Great Basin Joined: Dec 24, 2017 Post Count: 2391 Status: Offline Project Badges:
|
4 MB of L3 cache per instance is probably a good rule of thumb for now... A disproportionate hit on higher core CPUs. E.g., an E5-2699v4 with 22c44t and a 55 MB cache could only support 11 of 44 threads or 25%.BTW, I dedicated all my cores to MIP1 this week, and although run time has remained constant from the previous Ebola run, the credits dropped in half. ![]() ...KRI please cancel all shadow-banning |
||
|
|
Aurum
Master Cruncher The Great Basin Joined: Dec 24, 2017 Post Count: 2391 Status: Offline Project Badges:
|
Running app_config.xml causes a serious BOINC problem. E.g., my DC rig had 4 GPUs running Einstein@Home and then I installed app_config.xml and only one GPU will run E@H. That's unacceptable so MIP & E@H will not be able to run on the same rig.
----------------------------------------I divide my cache by 5 MB then subtract one and that's my max_concurrent. Then I have to run other CPU projects to fill out the threads. Too bad BOINC doesn't have an Edit app_config.xml button and not just a Read Config Files. <app_config>BOINC: Instead of using "At most use 88% of the CPUs" it should be "At most use 7 CPU threads". ![]() ...KRI please cancel all shadow-banning[Edit 1 times, last edit by Aurum420 at Mar 11, 2018 5:19:41 PM] |
||
|
|
Aurum
Master Cruncher The Great Basin Joined: Dec 24, 2017 Post Count: 2391 Status: Offline Project Badges:
|
I just noticed that not all CPU threads are active. E.g., an E5-2678v3 has 24 threads & a 30 MB cache. I specified use 84% of the CPUs to reserve four threads for Einstein@Home running on the GPUs. I used the above app_config.xml file limiting MIP to 5 threads. But only 11 threads are running WCG WUs instead of 20.
----------------------------------------![]() ...KRI please cancel all shadow-banning |
||
|
|
Aurum
Master Cruncher The Great Basin Joined: Dec 24, 2017 Post Count: 2391 Status: Offline Project Badges:
|
I deleted the app_config.xml file and clicked Read Config Files and I'm back to 20 threads running WCG WUs but too many MIPs.
----------------------------------------I guess MIP will have to be banned since it causes too many problems. Shame, I was looking forward to a time when flatulence smells like a rose garden :-) ![]() ...KRI please cancel all shadow-banning |
||
|
|
Byteball_730a2960
Senior Cruncher Joined: Oct 29, 2010 Post Count: 318 Status: Offline Project Badges:
|
Hi,
----------------------------------------I am so glad that you guys made this thread. I was focussed on getting 100 years on some of the older projects before switching to this project but got a shock when I saw a massive drop in my points and also my runtime. Most of my points come from 2 dual xeon 40c/80t machines which each have 16gb of ram and work super well. Switching to 100% MIP resulted in massive drops in points and runtime too. I can see the points drop that everyone is talking about, but I didn't see where the drop in runtime came from. Maybe there was a fight for resources? I don't really know. Anyway, if I adjust the config files to run a limited number of WUs, how many should I run max without affecting performance? I've seen a lot of talk about 2, but with my rigs, would 4 be ok? Edited as I obviously can't write English properly anymore. [Edit 1 times, last edit by vcd683s at Jun 21, 2018 8:35:49 AM] |
||
|
|
ca05065
Senior Cruncher Joined: Dec 4, 2007 Post Count: 328 Status: Offline Project Badges:
|
One suggestion I have heard before is to restrict the number of MIP running to one per core. The remaining threads can be used for other WCG projects. You may even try increasing the MIP work units running if the suggested limit is OK - I think I used to run 6 MIP on a 4c/8t i7.
----------------------------------------[Edit 1 times, last edit by ca05065 at Jun 21, 2018 7:23:32 AM] |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1684 Status: Offline Project Badges:
|
The real pity is that I am convinced that the science could probably be optimized for avoiding such bad performances. Nevertheless within 9 months no significant solution has been provided and the scientists do not give the feeling that they become aware and that they are willing to take care of this issue. It is especially bad and inappropriate because the MIP project should probably run for about one decade.
----------------------------------------We contribute as volunteer but not for free (we have to pay the electricity bill) and it is not good to have the feeling that the resources are partially wasted. I am still considering that the fairness on scientists side is that they have to take care to optimize the code for the best efficiency as possible. Cheers, Yves |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The real pity is that I am convinced that the science could probably be optimized for avoiding such bad performances. Nevertheless within 9 months no significant solution has been provided and the scientists do not give the feeling that they become aware and that they are willing to take care of this issue. Doug Renfrew replied in this thread (about half way down on page 4) that they understand the problem and, although they can't make any promises, they are looking into it. It, most likely, is not a trivial fix. |
||
|
|
Byteball_730a2960
Senior Cruncher Joined: Oct 29, 2010 Post Count: 318 Status: Offline Project Badges:
|
Hmmmm, I might just start with 2 and increase it slowly until I find the right balance.
Until a fix is implemented. I agree it is probably not trivial at all. |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1684 Status: Offline Project Badges:
|
Hi Doneske,
----------------------------------------thank you for reminding me to Doug's message. Indeed, in the mean time, I forgot it. Even if the problem resolution is not trivial, it would be good to have some information, just to know that the problem is still being addressed. Cheers, Yves |
||
|
|
|