| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Is there a way to reduce the number of tasks that run? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 22
|
|
| Author |
|
|
genhos
Veteran Cruncher UK Joined: Apr 26, 2009 Post Count: 1103 Status: Offline Project Badges: 
|
While I expected BOINC had been written this way for a reason (Scribe!
---------------------------------------- ), it was what that reason was. Thanks for the info, lavaflow.---------------------------------------- [Edit 1 times, last edit by genhos at Apr 19, 2014 4:01:05 PM] |
||
|
|
Alex_Ross83
Cruncher Joined: Jul 28, 2013 Post Count: 11 Status: Offline |
Thanks again for all the replies. Generally it would take between 4-8 hours - I do the Clean Energy Project and Fights AIDS at Home stuff.
It was quicker but I reduced the CPU power being used as it was heating up my laptop and making a fair whack of noise. I followed the guidelines which suggested setting CPU usage to 50% or something. |
||
|
|
noderaser
Senior Cruncher United States Joined: Jun 6, 2006 Post Count: 297 Status: Offline Project Badges: 
|
Yep, that's pretty much it, 1 task 1 core. As for why you can't get all cores to crunch 1 task, I don't know but hopefully someone can explain. Basically 'cos BOINC is not designed that way, so you have no choice.  BOINC has been able to handle multi-threaded apps for some time, but the projects have been slow to adopt them. I've heard that running multi-threaded apps does not always yield better performance and sometimes even yields slower performance, though that largely depends on the app and the type of computation. |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3715 Status: Offline Project Badges: 
|
Reducing the number of cores from 4 to 2 will NOT cause the remaining 2 to run twice as fast and finish tasks in half the time The way you word it, Scribe, it's true. However when it is a hyperthreaded processor, reducing the number of active tasks to the number of physical cores allows tasks to need less elapsed time to complete. Say about 20-30 % less. Unfortunately in this precise case Alex has forgotten to tell us about his processor model. Also Alex, if you use your laptop less frequently you should 1. avoid applications like CEP2 which need much elapsed time and don't checkpoint frequently 2. avoid shutting down your laptop when you stop using it; prefer putting it to sleep so that it resumes exactly where it was when you use it again. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Not sure about the math, but if running on my ht machine and limiting the processing to just physical cores, the jobs run about 80 percent faster, and hardly cooler. Hyperthreading gives an overall completed work uptick of 15-30 percent, depending on what age the cpu is from. Disabling of ht is not necessary as the cycles will be used anyhow. Whilst, never tried disabling this feature to see if there's crunching improvement. Possibly quite marginal and at the same time loosing the smartness of the OS when there's incidental higher user application demand.
|
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3715 Status: Offline Project Badges:
|
Alex, you can trust lavaflow's figures: I have not run a HT machine for several years and I have been lazy to look back for my older posts.
---------------------------------------- I had 30 % in memory but it is indeed for the throughput, not the running time. |
||
|
|
Alex_Ross83
Cruncher Joined: Jul 28, 2013 Post Count: 11 Status: Offline |
Sorry - didn't mention my own specs - 'tis only an Acer Aspire 4820T - so Intel Core i3-350M processor (2.26ghz).
Good idea re: projects - I chose CEP because it was something I would want to support over the other projects - any ones people could recommend that tend to be shorter tasks? Funnily enough, it feels like the tasks on my Nexus 7 for the same project are shorter - but could it be that the cores can be more dedicated on a tablet with a simpler OS? |
||
|
|
cjslman
Master Cruncher Mexico Joined: Nov 23, 2004 Post Count: 2082 Status: Offline Project Badges: 
|
Funnily enough, it feels like the tasks on my Nexus 7 for the same project are shorter Sorry, but you are not running CEP2 WUs on a Nexus 7... that tablet runs on Android and there are only FAHV WUs (or FAAH, can't remember which it is) for Android. CJSL Crunching for a better world... |
||
|
|
Alex_Ross83
Cruncher Joined: Jul 28, 2013 Post Count: 11 Status: Offline |
That'll be why then! :)
|
||
|
|
Alex_Ross83
Cruncher Joined: Jul 28, 2013 Post Count: 11 Status: Offline |
On a related subject then - if my android device has four cores and 1GB RAM, will tasks complete faster running two cores on 100% RAM or four cores on 100% RAM? i.e. does the RAM cap out anyway so it won't make any difference (e.g. each core takes, say, 250mb regardless) or will it mean the 1GB is shared out between fewer cores and therefore get twice as much and therefore run, if not twice as fast, certainly a little bit faster? e.g. two cores getting 500MB RAM instead of 250MB RAM?
Again, I know it's super simple stuff but I'd just like to have the settings done in a certain way then I can forget about it and know I have it relatively efficiently set up. :) Thanks again for everyone's help. |
||
|
|
|