| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Same machines, very different results |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 13
|
|
| Author |
|
|
depriens
Senior Cruncher The Netherlands Joined: Jul 29, 2005 Post Count: 350 Status: Offline Project Badges: 
|
I have something strange going on with one of my computers, I'll try to explain a little.
----------------------------------------All of the computers I will describe are in the same domain, have the same software installed and are used for the same thing. I'm running BOINC on all of my machines and the deployment has been via Active Directory, so the installation is also identical. Most of them are running 24/7 so you should expect that the same machine gets on average the same result as the other. They are all free of spyware and adware and are checked for virusses on a daily basis. There are also 2 P4's 3.2Ghz and both of them have 512MB RAM. The machines are exactly identical. I've been doing some stats on my machines for the last month or so and I found some confusing results. - P4 3.2GHz (1) Average workunit length: 5.14 hours Average points per hour: 10.41 (BOINC) - 72.87 (WCG) points - P4 3.2GHz (2) Average workunit length: 7.28 hours Average points per hour: 7.48 (BOINC) - 52.36 (WCG) points Strange huh? Does anyone have a possible cause for this difference?? For reference, I have 2 other P4 3.2GHz with 1024MB RAM. They do more or less the same as the first one. So it's definately the second one which is "the strange one".  I have quite some different models from P3 999MHz up to the P4 3.2's and "the strange one" scores less points than a P4 2.4GHz. Of course these points are of no importance compared to the scientific results which are achieved, but as a "tech-guy" I am always curious to find out the cause of this.   |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello depriens,
The machines are exactly identical. If so, then Task Manager might show something different in the software environment. Alternatively, Belarc Advisor or a similar diagnostic might show a different cache size on chip. These are the only possibilities that occur to me. |
|||
|
||||
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
2 cents from a team member.....BOINC automatically runs every 5 days a new benchmark. The benchmark is like on-clock, so if one happens to run nighttime and the other daytime?
----------------------------------------If you run BOINCview, you can actually see for each machine what the Benchmark hourly claim will be till the next benchmark. .....which still does not explain the average WU time part......a few heavies of FAAH that have been cycling at checkpoint hurdles? 512mb at least suggests you've only been doing FAAH off late on these machines....again BOINCview does a great job of tallying all the WU's gone thru and the time taken. It might highlight a few extra long once.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
... and are used for the same thing. They may be used for the same thing but are they used the same way? For example, they might be used by 2 draftsmen running AutoCAD. One draftsmen may have habits that allow BOINC to have more CPU time. For example, when my friend uses a computer she never stops moving the mouse. Even when reading a page of text she constantly moves the mouse around for no reason which takes CPU cycles that could otherwise go to BOINC. |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Maybe a case of Double Dutch, but are we talking clock time or the BOINC CPU time? On identical machines, in CPU time they should be virtually identical.
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
uplinger
Former World Community Grid Tech Joined: May 23, 2005 Post Count: 3952 Status: Offline Project Badges: 
|
p4 3.2...is the one machine that is producing less results running in a different HyperThreading mode than the other machines. Easy way to tell is find out how many wcg_*.exe processes are running (100% CPU) at the same time.
-Uplinger |
||
|
|
depriens
Senior Cruncher The Netherlands Joined: Jul 29, 2005 Post Count: 350 Status: Offline Project Badges:
|
Ah, a whole bunch of replies already!
---------------------------------------- Time for some clarification:The 2 computers are in our sales office and are only running Outlook, Word, Excel and a bit of internet sometimes. No CPU-intensive CAD / ERP programs whatsoever. The processors are of the same type and both processors have HT enabled. In my BOINC profile I have selected to use only 1 (one) CPU, so on both pc's there is only one workunit running at the same time. As far as I know, BOINC only reports actual CPU time, so wall clock time is not applicable. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
There's two things you can do: check the benchmark results on both machines. Identical machines should give very similar benchmarks, assuming nothing else is running.
Secondly, run System Monitor for a while, and graph the CPU usage of the science application process. If you open Task Manager and sort the process list by CPU, you will be able to see immediately what is taking time away from BOINC. |
||
|
|
Jason1478963
Senior Cruncher United States Joined: Sep 18, 2005 Post Count: 295 Status: Offline Project Badges: 
|
It may also be worth checking the temperatures to see if it may be throttling because of heat issues. A temp monitor program like speedfan may work if you don't already have one. Higher temps then normal is a good indication its time to check the fans and clean the CPU heatsink.
----------------------------------------I believe the task manager is a good idea too. When your in task manager under the process tab you can add the view for cpu time to see where your cpu time is going. Good Luck Jason  |
|||
|
||||
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7687 Status: Offline Project Badges: 
|
One other thougt occurs. Possibly run a device profile on both machines and see if all the components are identical. Also check to see if both motherboards are the same. Some whitebox manufacturers will mix and match various components even in "identical" machines. There might also be subtle timing differeences in the machines.
----------------------------------------
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
|