| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: The Clean Energy Project - Phase 2 Forum
Thread: CEP2 AMD 4P Optimization |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 58
|
|
| Author |
|
|
OldChap
Veteran Cruncher UK Joined: Jun 5, 2009 Post Count: 978 Status: Offline Project Badges: 
|
Yes on the first
----------------------------------------Don't know on the second  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Ok
----------------------------------------got Fatdog64 v701 waiting on bits n pieces when they are installed we will press the test looking for 5.5% improvement on the 6276s to get me under 18 hours my Linux is pitiful on a good day, so i may come crying on How 2, this or that Interesting suggestion Thanks  [Edit 1 times, last edit by Former Member at Dec 7, 2015 2:41:07 AM] |
||
|
|
OldChap
Veteran Cruncher UK Joined: Jun 5, 2009 Post Count: 978 Status: Offline Project Badges:
|
Good luck with that (asking me questions)
----------------------------------------I have had it running wcg but that is all |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I have had it running wcg "HAD" what are you using now and why did you switch?  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
It's 18 -CPU- time hours, not 19 when the cut-off happens [no debate on the why plz... off-topic]. Depending on efficiencies, my laptop goes daytime about 70-75%, nighttime about 85-95%, the elapsed time is proportionally longer. The scripted staggered start can be achieved with the app_config.xml. On my work-system it can take -in use- about 7 minutes -Elapsed- before the setup is complete and the CPU time starts ticking, so that would be my used interval to increment the <max_concurrent> tag value for <name>cep2</name> app and re-read [using the boinccmd.exe tool in a terminal window]. An app_config.xml just for CEP2 looks like this: <app_config>It's been done before, so search the forums, kind of a post-bootup script or one that's kicked when the boinc client is started [via a batch]. SekeRob* i have serched the forums for this and there are hundreds and Hundrds of pages i have read some ... and from what they said the app_config.xml you put above will just run one instance of CEP2 at a time since i believe you know what you are talking about i must be looking in the wrong place can i get some guidance to the thread that tells me how to delay the start of each CEP2 WU by a couple minutes ... then starts the next ... a few minutes then the next and so on ... the plan is ... an OS that is totally in ram and a totally dedicate 64 core 64 Gig RAM, box that's only crunching CEP2 WUs (if i can make this work ... ill clone this box) so a delayed start between WUs would help make this a "maybe could happen" but is a ... "Definitely NOT Happening" without it thanks ... sorry to be a bother [Edit 1 times, last edit by Former Member at Dec 7, 2015 11:24:11 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Take a good close look at Fatdog64 as an OS. Runs completely in ram as does boinc and wcg. In fact you have to learn quite a bit to get things to save to disk when you need that "How To Backup FatDog64 - So Simple It's Awesome" https://www.youtube.com/watch?v=wLiHtSWzW5Y this would make FD64 a "persistant" configuration once the 32 bit libraries and video drivers and such were installed in the event of a power failure the only loss would be the current WUs unless an "auto save every X minutes" script could be implemented then there would be only minimal "crunch time" lost but 64 gigs is quite a bit ...so ... not sure how practical "auto save" would be [Edit 2 times, last edit by Former Member at Dec 8, 2015 4:26:43 AM] |
||
|
|
Yarensc
Advanced Cruncher USA Joined: Sep 24, 2011 Post Count: 136 Status: Offline Project Badges: 
|
can i get some guidance to the thread that tells me how to delay the start of each CEP2 WU by a couple minutes ... then starts the next ... a few minutes then the next and so on ... so a delayed start between WUs would help make this a "maybe could happen" but is a ... "Definitely NOT Happening" without it thanks ... sorry to be a bother As far as I know, there is no functionality that automatically does that. The way you could do it with app_config is: 1) Start with it set to 1 2) Wait a few minutes, change it to 2 3) Reread config files on the client (I think under the advanced drop down? don't have it installed at work, so I can't check right now) 4) repeat steps 2 and 3 until up to the desired number of WU's This should get you a staggered start, which hopefully further randomizes since the WU's are of different lengths and won't all be finishing at the same time. |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
There's various ways to skin the cat, the simplest is to create a whole swat of app_configs called app_cep2-01, app_cep2-02, app_cep2-03 and so on, then with a batch copy -01 over app_config.xml, let boinccmd read in the config, insert a sleep of whatever interval, copy -02 over app_config.xml, read config in with boinccmd and on and on.
----------------------------------------This is for the initial launching of CEP2. On restarting the system, you have to repeat this as 8-6-32-48-64 simultaneous starts/resumes will just kill lots to the point that a host will get down to quota 1 per day. Resumes can have a shorter start-interval, the initial job #01 needing most time. The advanced [not me], can have a scriptor insert the incremented number and read, increment, read etc, but as said, on the forums this has been discussed and using advanced search filters, and limiting to CEP2 forum its finding can be narrowed down. Keywords e.g. 'delay OR script OR staggered OR sleep or batch'. Could do it in Basic [really old school] with a for-next, but not going to venture back into dementia territory. Beyond a certain number, no randomizing of various length tasks will prevent from simultaneous finished and start of multiple... trial and error what the sweetpoint is. [Edit 1 times, last edit by SekeRob* at Dec 8, 2015 3:32:14 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
ok ... let me chew on this for a while ...
i am very adverse to 1/2 step upgrades on my way to 6380s ... however ... sometimes there is a deal to sweet to pass up so ... that and getting TPC set up, kinda has my attention at the moment. "petty much" new 6282SEs for $63.77/ea ... to my door!!! ... Aaaaaa??? ... ok ill take 8 all those 6174s ... for sale ... CHEAP!!!! make an offer  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
a little update ... for those interested
i closed the deal on the new CPUs and that is significant because my "ballpark Guesstimate" of the Mhz required to consistently make the 18 hour cut off (with Opteron 6000 series processors in a P-state of, full load turbo) is 2750 Mhz ... so the new processors make that part work and the other thing is Team leader pointed me to this thread and said: "last post in this thread has the contents of the cc_config.xml that you need to have to delay BOINC" http://hardforum.com/showthread.php?p=1042020176#post1042020176 so ... I will be running trials of FatDog64, TPC, and this script when the new CPUs arrive  |
||
|
|
|