CPU time, claimed credit oddity...
28 Dec 2005 14:22:11 UTC
Topic 190459
(moderation:
One of my machines is seemingly kicking butt... Since I brought it on line the 'credit claimed' is running in the range of 11.1 - 11.3 for the 16 work units it's processed... If I understand correctly, the claimed credit is a function of the amount of time required to complete the work unit...
Well, I have nine other machines identical to this one that are claiming credit in the high 50s to low 60s... These are all P4s running at 2.8 gigs, half a gig of ram, non-HT. All are running WinXP Pro and the same BOINC client...
Any thoughts???
It's easier to beg for forgiveness that it is to ask for permission...
-AFFTC
CPU time, claimed credit oddity...
)
Claimed credit is solely a function of CPU time and benchmarks. Your benchmarks could be low if you were running another program when BOINC ran benchmarks the last time. Also, spyware is often the cause of poor benchmarks.
RE: One of my machines is
)
Perhaps the older machines are still working off Einstein work units, and the new started up right away with Albert 4.37's? In the last couple of days, one of my machines has gotten a string of 4.37's that it completes in a small fraction of its time on the older ones, claiming typically about 6.7 credit, and being awarded about 15.5 credit.
RE: RE: Claimed credit
)
Ummm... nope... nothing running on this machine other than the normal background apps and BOINC... It's one of the few machines that I have that is used only every other day or so...
No spyware, either... the machine is absolutely clean...
It's easier to beg for forgiveness that it is to ask for permission...
-AFFTC
RE: RE: Perhaps the
)
All my machines, 30 or 31 of 'em now, are running 5.2.13 which I believe is the default when you download the app... None of the boxes are running Albert...
Any other ideas??
It's easier to beg for forgiveness that it is to ask for permission...
-AFFTC
RE: Any other ideas??
)
You might want to unhide your computers. Or at least provide links to the
computer(s) in question. Might help in finding the cause.
Michael
Team Linux Users Everywhere
RE: RE: Claimed credit
)
Do you have any similar computers to compare performance with? (Since your computers are hidden, it's hard for us to tell anything.) If you have similar hosts, I would compare benchmarks and CPU time to see which variable is out of whack. If benchmarks look low, try running them again (under the "Commands" menu) to see if there is any change.
RE: RE: RE: Claimed
)
Yeah... I have nine more exactly like this box... I'll give that a shot and report back...
Thanks!
It's easier to beg for forgiveness that it is to ask for permission...
-AFFTC
RE: RE: RE: RE: Clai
)
I looked at four boxes exactly like the one in question and all the results were quite similar; around 1400 DP MIPS and 2200 integer MIPS for the sample boxes and 1401 and 2480 for the oddball...
?????
It's easier to beg for forgiveness that it is to ask for permission...
-AFFTC
Warhawk, This is the BOINC
)
Warhawk,
This is the BOINC Wiki page that explains Claimed Credit. As you can see, if the "oddball" has higher benchmarks, then it must be processing WU's much quicker. Are you sure it hasn't been processing the new Albert units?
Stick
Edit: Or are you sure the WU's are finishing successfully? (i.e. No "Client errors" ???)
RE: Warhawk, This is the
)
All the work units are finishing successfully...
It looks like several of my boxes are processing Albert work units... This is from the results from the oddball box:
5.2.13
2005-12-27 19:39:10.4062 [normal]: Start of BOINC application 'projects/einstein.phys.uwm.edu/albert_4.37_windows_intelx86.exe'.
2005-12-27 19:39:10.4062 [normal]: Started search at lalDebugLevel = 0
2005-12-27 19:39:11.0156 [normal]: Checkpoint-file 'Fstat.out.ckp' not found.
2005-12-27 19:39:11.0156 [normal]: No usable checkpoint found, starting from beginning.
2005-12-27 19:47:29.8906 [normal]: Fstat file reached MaxFileSizeKB ==> compactifying ... done.
2005-12-27 21:09:19.5156 [normal]: Search finished successfully.
It looks like fully a third of my boxes are running the 4.37 app... How does it happen that most got 4.79 and some got 4.37???
Also, out of the ten or so boxes running 4.37 only one is producing claimed credits in the range of 11... all the rest are 30 to 40...
What's the downside of processing Albert work units, other than the low credit claimed?? By the way... that box is out of work units and has been since this morning as it's reached it's quota for the day... I understand that I can force more work units by modifying the # in the 'Connect to network' field, correct???
It's easier to beg for forgiveness that it is to ask for permission...
-AFFTC