| 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: Questions about participation drive for CEP2 |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 152
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I'm resurrecting this somewhat ancient thread because i found it by Google when looking for information about running multiple CEP2 tasks. Better than starting a new one, i assume.
----------------------------------------I have a three-disk setup; an SSD for my OS and two SATA2 data drives, one of which is dedicated only to long-term storage and has a partition for BOINC use. I've been crunching solely CEP2 workunits with 8 threads and haven't noticed any slowdowns whatsoever, and out of 8GB RAM only about 35-40% is constantly in use, leaving plenty for the rest of my needs. I can see a heap of I/O activity with Process Explorer but benchmarking the other 2 drives yields similar performance with or without BOINC/8x CEP2, so it seems there's no interference. Am i just lucky or is there something special in my setup, given the apparent amount of trouble people are having while running more than a couple CEP2 workunits? Also, i read an older post somewhere that CEP tends to use the swap file a LOT regardless of having plenty of free memory, i'd like to know if this is still true? edit: typo [Edit 1 times, last edit by Former Member at Apr 18, 2013 4:58:42 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I think you would only notice the slowdown of WUs runtime if you compare to how long it would take if you ran 1,2, or 3 (maybe even 4). It should not affect your other drives benchmarks much as the CPU can handle managing the I/O it's the harddrive itself that is the bottleneck. Yes, still uses lots of swap.
Test running a single WU compared to running 8 and let us know if you still ahve them running in about the same time, we'd like to know if you found "the magic" <thumbsup> |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
That's an interesting thought, i'll test it and post results. Average runtime of the 65 CEP2 workunits i've returned so far with 8 simultaneous threads is 6h 33m per WU, which according to the global CEP2 statistics is slightly faster than the average runtime. Running a single unit now.
----------------------------------------[Edit 1 times, last edit by Former Member at Apr 18, 2013 5:13:08 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Ha, no magic or silver bullets here. That was a quick CEP2 WU; 3h 37m, almost half the time of my averages. I'll need to test this further but if it really affects progress this much, even if i have no system slowdowns with 8 threads it might indeed be more beneficial to run 3-4 for CEP2 and give the rest to another project as some have pointed out.
Funny though, the difference between CPU time and elapsed time has been around 4-7 minutes for all the workunits that ran simultaneously, and it was the same for the single workunit processed alone. Shouldn't that indicate the opposite, that there's no ill effect? The single CEP2 job took 12,5% of my total CPU power, and the same percentage is allocated per job with 8 threads as well. How come it ran so much faster? How does this stuff work anyway?  |
||
|
|
Yarensc
Advanced Cruncher USA Joined: Sep 24, 2011 Post Count: 136 Status: Offline Project Badges: 
|
just a thought, assuming your using one of AMD's FX 8 cores (I believe everything else above 4 cores is server) it isn't actually 8 cores, there are 4 'modules' each of which counts, according to AMD, as 2 cores. However they share some of the hardware, so this could account for some of the speed increase, but I wouldn't imagine this would account for your almost half time.
yaren |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I have an Intel i7 with 4 cores and 8 threads. I'm testing additional single workunits, but i'm already puzzled so it's likely to only add to the confusion.
 |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
It seems i can only get 3,5 hour WU runs when i run a solo CEP2 unit with the other 7 threads sitting idle. I tested by crunching all other projects with 6 threads and 2 for CEP2, and again the average is over 6 hours for each CEP2 WU, same as if i was crunching it with all available threads.
Oh well. |
||
|
|
ca05065
Senior Cruncher Joined: Dec 4, 2007 Post Count: 328 Status: Offline Project Badges: 
|
If you are willing to perform further tests, it would interesting to see if it is the CEP2 work units interfering with each other or the effect of other work units.
test 1 : two CEP2 units and 6 idle theads test 2 : one CEP2 and 7 other WCG work units |
||
|
|
Mamajuanauk
Master Cruncher United Kingdom Joined: Dec 15, 2012 Post Count: 1900 Status: Offline Project Badges: 
|
Over the last few days I've ramped up the number of CEP2 wu's I'm running. I'm now running 8 wu's at a time on my 8 core machines, they don't seem to have any problems, all appear to be returning 'Valid' with NO wu's being returned as 'Invalid'
----------------------------------------Wind them up, runch more!
Mamajuanauk is the Name! Crunching is the Game!
  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
If you are willing to perform further tests, it would interesting to see if it is the CEP2 work units interfering with each other or the effect of other work units. test 1 : two CEP2 units and 6 idle theads test 2 : one CEP2 and 7 other WCG work units I doubt just one or two runs in these conditions will prove anything useful but your "test 2" scenario didn't pan out any different - i just had one CEP2 unit take 9 hours with 6 HPF2 units and 1 idle thread which i reserved for other computer activities, not gaming though. I remember reading there are size differences between the CEP2 workunits themselves so i'd definitely need to test a lot of them to draw any long-term conclusions. I think i'll just queue up some work instead and let the client do its job.  |
||
|
|
|