Blocked [Negative Calibration Limit] w/ Trux 5.3.12.tx36?

Gecko
Gecko
Joined: 5 Jun 06
Posts: 10
Credit: 61,039
RAC: 0
Topic 191369

Hi,

I have a M 1.6 Banias w/ S41.07. Have used Trux's calibrating client on before w/ S@H. Just reinstalled Trux's Boinc.exe in place of Crunch3r's 5.50 (didn't install .dll w/ Trux pkg.) I understand claimed credit can dive initially and it can take 20-30 WUs to recover, but never have seen this before?

6/12/2006 9:27:09 AM|Einstein@Home|CC calibration: blocked [negative calibration limit] 14.02 >> 10.14 (time: 3844s >> 2469s / Gfpops: 1.49 >> 2.28)

You can see the WUs' stderr here:

http://einsteinathome.org/host/657782

Is this behaving correctly? I also notice that I don't see the calibration acknowledgement in each WU stderr report, yet BOINC message tab isn't showing an errors other than the negative calibration message.

What gives? Ideas?
Is Crunch3r's 5.50 a better solution?
Thanks!

archae86
archae86
Joined: 6 Dec 05
Posts: 3,150
Credit: 7,116,464,931
RAC: 573,513

Blocked [Negative Calibration Limit] w/ Trux 5.3.12.tx36?

Quote:

I understand claimed credit can dive initially and it can take 20-30 WUs to recover, but never have seen this before?

6/12/2006 9:27:09 AM|Einstein@Home|CC calibration: blocked [negative calibration limit] 14.02 >> 10.14 (time: 3844s >> 2469s / Gfpops: 1.49 >> 2.28)

What gives? Ideas?
Is Crunch3r's 5.50 a better solution?
Thanks!


tx36 is willing to report a lower credit claim on SETI than benchmark claim, but on other projects it won't.

Just wait a dozen or so work units. You'll see the attempted claim rise by about then, though it won't be anywhere near equilibrium.

Until S4 is done, I think tx36 is your best client for getting approximately fair claim on Einstein using akosf science applications.

paul milton
paul milton
Joined: 16 Sep 05
Posts: 329
Credit: 35,825,044
RAC: 0

i askd a similare question a

i askd a similare question a while back, sure enough, cleard right up after about 1.5 days, so i see

6/12/2006 10:22:44 PM|Einstein@Home|CC calibration: 15.66 >> 36.55 (time: 5092s >> 6047s / Gfpops: 1.07 >> 6.20)

:)

seeing without seeing is something the blind learn to do, and seeing beyond vision can be a gift.

Gecko
Gecko
Joined: 5 Jun 06
Posts: 10
Credit: 61,039
RAC: 0

RE: Until S4 is done, I

Message 37430 in response to message 37428

Quote:
Until S4 is done, I think tx36 is your best client for getting approximately fair claim on Einstein using akosf science applications.

Thanks for the reply and please excuse me for asking...the last thing I want to do is open a can of worms after seeing the Seti forums implode partly over credit issues. I just want to claim "same" credit as if using std ap, nothing more, nothing less. My "reward" is being able to do the job faster (more science) w/ Akos's aps, but I'm happy to claim normal credit.

From looking at some WUs, it looks like BOINC studio claims MUCH higher. I understand possible benefits to the quorum against std clients running opt aps "underclaiming" and that it may help leverage against receiving lower granted credit from time to time. No worries, to each their own.
Heck, I'm not clear on the install process after researching the subject, even if I wanted to.
Running Crunch3r's 5.50 still looked to claim low @ 24-26 per WU.
What is the normal credit claimed when using a std client and ap? This will help give me an idea of where I'm at. Though, if I read correctly, S5 will use FLOPS counting? So this all will be moot in a few days?

archae86
archae86
Joined: 6 Dec 05
Posts: 3,150
Credit: 7,116,464,931
RAC: 573,513

RE: I just want to claim

Message 37431 in response to message 37430

Quote:
I just want to claim "same" credit as if using std ap, nothing more, nothing less.


That is what tx36 seems good at--once it has had the roughly 3 dozen WU's needed.

Quote:

What is the normal credit claimed when using a std client and ap?


We generally talk about two major sorts of WU's "long" and "short", though a while back I saw one batch of "super shorts". Within each major type there is systematic variation of the order of 10-20% perhaps, but since the types differ by about 3.5x, they are still quite distinct. Very roughly, the long ones get about 38 cobblestones, the short ones more like 11.

Quote:

Though, if I read correctly, S5 will use FLOPS counting? So this all will be moot in a few days?

It will be moot, but not by FLOPS counting, but rather a preassigned cobblestone credit determined by the project.

SETI needs computation time adjustment, because some work units error out extremely early, so must be given very little credit, while some others are "noisy" and should get more credit than otherwise comparable ones. The Einstein ones have far less random variation in work content from that predicted, so using end measurement data in the credit process is probably introducing more random credit noise than it is removing. I support this move by Einstein, but it is not a model for other projects unless the project can make an accurate prediction of relative computational effort for each WU sent out.

Gecko
Gecko
Joined: 5 Jun 06
Posts: 10
Credit: 61,039
RAC: 0

RE: We generally talk

Message 37432 in response to message 37431

Quote:

We generally talk about two major sorts of WU's "long" and "short", though a while back I saw one batch of "super shorts". Within each major type there is systematic variation of the order of 10-20% perhaps, but since the types differ by about 3.5x, they are still quite distinct. Very roughly, the long ones get about 38 cobblestones, the short ones more like 11.

Thanks for the info. That's what I was looking for.
Cheers!

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.