Tech and Talk > Computing Clubs > Team Starfire > [BOINC] How to Optimize Boinc and Seti v5.2.13

reply to Sat_Man

Re: [BOINC] How to Optimize Boinc and Seti v5.2.13

Is there any issue with this? I mean what good is the enhancement of finishing the WU's faster if there is no "real" improvement in the final result? Or am I missing something somewhere?

reply to Gremlin_CT

reply to Sat_Man
Gremlin_CT is using the optimized seti client, but he is using the "stock" berkeley boinc project manager.

Just finishing work quicker with the stock boinc manager will actually lower your claimed credit / result because it reports a fixed claimed credit per hour of work, using a formula that considers your computers Measured floating point speed, Measured integer speed, and the actual (wall) clock time.

Trux's calibrating boinc manager comes from the school of thought that one result should approach a fixed value of approximately 32 credits, and will adjust over time the claimed credits to get there. Hence, the "calibrating" part of the name.

Dual Core AMD Opteron(tm) Processor 270 (Number of CPUs 4 )- claims ~15.2 credits/result/hour and finishes a result in 3000 seconds:
»setiathome.berkeley.edu/show_hos···=2021249

With Trux's calibration, your Dual Core AMD Opteron(tm) Processor 270 claimed credit / result / hour should rise from ~15.2 to ~38.4 at the same 3000 seconds wall time / result.

Intel(R) Xeon(TM) CPU 2.80GHz (Number of CPUs 4) - claims ~9.45 credits/result/hour and finishes a result in 6300 seconds:
»setiathome.berkeley.edu/show_hos···d=653418

With Trux's calibration, your Intel(R) Xeon(TM) CPU 2.80GHz
claimed credit / result / hour should rise from ~9.45 to ~18.2 at the same 6300 seconds wall time / result.

Formula: ( claimed credit x 3600 seconds ) / actual wall time in seconds = claimed credit / hour

PS - do you have your preferences set to allow all four processors to run?

PPS - of course your claimed credit may bear slight resemblence to your granted credit since your result must be validated against the quorum of other users claimed credit.
--
Hello, my name is Bill and I'm a crunchaholic...

Proud to be the current host of Crunchenstein #1, #3, #4 and #5

I downloaded Trux's calibrating manager. Copied the files. Will see what happens! Thank you!

I just put it on the dual opteron machine for now, i want to see what happens.

Monotonous, yes, I downloaded Cpuz and installed the proper client as per that website. Unless I missed it there I hadnt heard of the Trux's software until just now.

Yes, my preferences allow 4 processors.

reply to Sat_Man
Here's a question for the group. I am running the Trux client. I have my cache set for 9 days. Whenever I send in a work unit, credit is given right away because 3 other people have already sent in their work.
Does it really help the team when the granted credit has been established already? Or should we be running with a shorter cache in order to help offset the granted credit value?
--
SWIM in a new direction. Check out Space Weather Information Monitor by going to »www.spacew.com/swim/index.htmlDownload your free trial version of Aurora Monitor by going to »www.spacew.com/aurora/index.html

reply to Sat_Man
The short answer is maybe yes.

To help resolve your concern, in your boinc folder, you will find the file truxoft_prefs.xml - add this line to the file

<return_results_immediately/> 

reply to Sat_Man
Hi, Parkut. Thanks for the quick reply.

My machine is sending the work units out as soon as it finishes. The problem is that with a cache of "X" number of days, it can take quite a few days to send a particular work unit back to S@H as it will crunch the oldest work unit that is in my cache first.
In looking at my results, it seems as if quite often people are sending them back within one or two days. Then after several more days my work unit is finally crunched and sent back, getting the average credit that has already been granted. This can be higher or lower, depending on what everyone else has already claimed. That's why I am wondering if we should have a shorter cache so we can help to set the claimed credit early for the team to get maximum credit.
--
SWIM in a new direction. Check out Space Weather Information Monitor by going to »www.spacew.com/swim/index.html
Download your free trial version of Aurora Monitor by going to »www.spacew.com/aurora/index.html

reply to Sat_Man
Well, with the advent of truX's BOINC client adopted as Starfire's supported BOINC optimized client, this thread will be unstickied, and »How to Optimize with Crunch3r and trux clients will be the updated one.

Jeff&Lana, your's is a good question, perhaps worth it's own thread? I had dropped my cache down to three days to get my work units in quicker, downside is I ran out of work across the board with the outage this week.

Paul
--
Distributed Computing.... The Search....
Proud member of the Cajun Crunch Team

reply to Sat_Man
Good idea, Paul. I'll create one to get some more input from the team.

reply to sargeeld

said by sargeeld:

Yepper, optimizing has raised my RAC by about 600 cobbies. I am expecting my upgrade kit this week and hope to have it online by this weekend. Hyper Threaded P4 2.66 GHZ with 1 meg of DDR 333 RAM. With the optimized client it should kick out some fantastic times and put me well over 2000 for my RAC.

reply to Sat_Man
3/27/2006 1:22:28 PM||Listening on port 31416
3/27/2006 1:22:28 PM||truXoft add-on: calibrate_credit = all
3/27/2006 1:22:28 PM||Suspending network activity - running CPU benchmarks
3/27/2006 1:22:28 PM|seti@home|CC calibration failed: credit_correction_factor checksum failed!
3/27/2006 1:22:33 PM||Running CPU benchmarks
3/27/2006 1:23:32 PM||Benchmark results:
3/27/2006 1:23:32 PM|| Number of CPUs: 2
3/27/2006 1:23:32 PM|| 1441 double precision MIPS (Whetstone) per CPU
3/27/2006 1:23:32 PM|| 2934 integer MIPS (Dhrystone) per CPU
3/27/2006 1:23:32 PM||Finished CPU benchmarks
3/27/2006 1:23:33 PM|seti@home|Resuming computation for result 13oc02aa.19035.8016.390890.1.157_1 using setiathome version 418
3/27/2006 1:23:33 PM|seti@home|Resuming computation for result 20dc98aa.26871.25362.298554.1.86_2 using setiathome version 418
3/27/2006 1:23:33 PM||Resuming computation
3/27/2006 1:23:33 PM||Rescheduling CPU: Resuming computation
3/27/2006 1:23:33 PM||Resuming network activity

I got that error message after i did the trux calibration stuff. is that supposed to happen?

*EDIT* and asside from that error the optimized client had dropped my times from 3:55 to about 1:40 and thats without the validators working to send me fresher wus so who knows yet.

*EDIT2* Now its down to 1:15 per and it looks like it may still get even lower on barnburner.

reply to Sat_Man
Trux stores a checksum in the files to prevent fraudluent manipulation of credits. This file sometimes gets clobbered.

Warning - back up your file folder first.

To clear this problem, I have deleted all but one of the seti related xml files in the \program files\boinc directory. The calibration will start over from zero.

You should leave the account_*.xml project files, as that is the idenfifying information between you and Berkeley (or other projects).

If you do delete the account_*.xml file, you will have to re-attach.
--
Hello, my name is Bill and I'm a crunchaholic...

Proud to be the current host of Crunchenstein #1, #3, #4 and #5

reply to Sat_Man
I've had that happen to me when I've installed the Trux client after I've started crunching a WU. After that it works fine.

Edit: This is an old thread that maybe we shouldn't be adding to. The newer one is »How to Optimize with Crunch3r and trux clients and it's stickied
--
Expand your universe... Join BBR Team Starfire

reply to 34140721
Well, BOINC is all we have. However, installed as a service, the whole deal is a snap to get going.

Even optimizing is basically a save-just-in-case, overwrite programs, go.
--
THE SENILITY PRAYER : Grant me the senility to forget the people I never liked anyway, the good fortune to run into the ones I do, and the eyesight to tell the difference.