| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Performance Differences between Boinc and UD client |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 10
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello everyone, My current setup is that I have a win2k and rhel 4 box and for the last few weeks I've had both running world community grid clients each night. It took me a few weeks to get the boinc client going as I didn't realise one existed until I took the time to have a look. Anyway I've noticed something that's a little surprising about the results that are being returned. The linux box takes around 60% less time than the windows box to complete a result. Now in fairness my linux box is a big newer but the biggest difference is a 3.0 Ghz cpu to a 2.4Ghz cpu in the windows box. Anyway here's a table of the brakedown of my results:
Days Hours Minutes Total in Minutes Results Returned Time taken per result Windows Box 10 22 45 15765 29 543.6206897 Linux Box 1 6 0 1800 5 360 I wonder if this difference is down to the difference in the power of my two machines or has anyone seen similar results with the same machine? Is it due to underlying differences in the two OSs or is the UD client just a bit slower to perform then the linux client? Would be a bit disappointing if it wasn't an efficient piece of code, its almost like charity organisations squadering donations. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Gah just read over it and noticed all my spelling mistakes, sorry must be too earily in the morning to be posting :). One other thing to mention that in the most part the results I've been processing on the win box have been FightAids@Home, I think I only remember seeing one Folding project work unit. I can't seem to find a way to tell which project is being worked on the Boinc client.
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
There has been some debate about this. The bottom line is both clients use almost identical science applications. The software that does the work is entirely seperate from BOINC and UD. HPF uses Rosetta, and FAAH uses AutoDock. In a controlled test, you should see a negligible difference between BOINC and UD. Unfortunately, none of us can reasonably set up a controlled test, since work units are assigned randomly.
At the moment, more FAAH units are being sent out than HPF units. This is because the HPF project is ending one major phase, and starting the next. We're expecting an announcement and a flood of new HPF work any time now. If you want to correct your spelling, there is an "edit" button in the top left of your posts. It is hard to see (white text on white.... *sigh*). |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello barryodriscoll,
All but a few seconds of CPU time is spent in the application program. Differences between OS should be unnoticeable. There could be a substantial difference in compiler efficiency, compiling the application program for each OS. But my bet is that you are running into the ordinary variation in work units. Anyway, the Windows compilers tend to be more efficient than the Linux compilers, and you are seeing what seems to be the opposite, so it is not the compiler. When I click 'My Grid' - 'Device Manager' - 'Results Status' I see that my most recent HPF unit took 1.96 hours but the one before it took 8.14 hours. For FAAH it is 7.53 and 5.76 hours. The work units vary greatly, so you are probably seeing CPU speed differences plus chance variations. If you bring up BOINC Manager and click on the Work tab you can see what is running. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Actually I was thinking about this since and you're right, I had automatically assumed that each work unit was a standard size within each project. Thanks for the replies, just out of curiousity is there any information available about the underlying scientific application which is doing the processing. Who made it, what did they use, are there any benchmarks/profilling data on the code? Anything to compare it too?
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
There is lot of information if you know what to look for:
Rosetta: http://boinc.bakerlab.org/rosetta/rah_science_faq.php AutoDock: http://www.scripps.edu/mb/olson/doc/autodock/ From these starting points, you should be able to find out as much or as little as you want. :-) |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
hey, a nice little challenge for me to play with, i have 2 indentical spec machines ( pent 4 3.0ghz etc ) i will try to grab the same w/u for each, the ud just does new unit when it finishes last one ( ok unless use udmon) so i will try to compare the exact w/u on 2 machines with only diff the boinc-win or ud setup, yah this will be interesting , will let youse know bfn
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi retep57,
The server does not hand out identical work units to different platform types. So I do not think that you can make this comparison. Lawrence |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
oh, doh , thanx for that, assumptions assumptions, silly me
 ok , so much for that, i am warming to the boinc version esp when i am messing around with computer setups, i can abort w/u if necessary eg when i am changing ubuntu to suse O/S etc, dont want to hold onto a w/u esp if others are waiting for unti return for points validation etc, also gives a chance to watch for return date deadlines to make sure i get work done well in time.. i plan to get the suse setup runningon BOTH cores then i will aim to complete the w/u on the ubuntu disk then stay with suse, otherwise just abort them (old ubu setup),. it is interesting to compare and contrast the boinc and ud, fors and against , pros amd cons for both, interesting and educational to check it all out.. no last thing is to get that 2nd core running grrrr.. i checked the default setting for multi core i am sure, doh must have skipped a step as i have done b4 on many occasions... |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
all sorted, the client somehow noticed the other core, mst have had some more communication part way thru w/u cool! all problems solved YAY!!
will finish these 2 w/u then temp pause this disk then cleanup or abort on old disk yay keep on crunching! |
||
|
|
|