HP2 computation errors

[sierra_bound]

Just a warning for those of you who've just started crunching. The Human Proteome 2 project has been known to cause errors. There are numerous threads at the WCG forum on this problem.

http://www.worldcommunitygrid.org/fo...d_thread,26706

Hypernova says he's seen fewer errors on this project using an i7-980X.

http://www.worldcommunitygrid.org/fo...fset,80#277196

Still, my advice to people is that unless you feel a strong need to run this project, uncheck it in your Device Profiles section.

[OldChap]

Thanks Sierra (this works too)

[fallwind]

Yeah this project has many issues, especially with Vista and Win 7. Older OS's were OK but you'll still get the occasional WU that will "hang" at a certain % and stay there until it times out which is annoying as heck, not to mention wasteful.

[NKrader]

yikes hope i havnt wasted much time..

[sierra_bound]

You may be one of the lucky ones who isn't getting errors. If that's the case, there's no reason to stop running HP2.

You can check how many errors you're getting by going to your My Grid page. On the left-hand side, click Results Status. Then, on the second drop-down menu, go to Error and then click Filter.

[NKrader]

You may be one of the lucky ones who isn't getting errors. If that's the case, there's no reason to stop running HP2.

You can check how many errors you're getting by going to your My Grid. On the left-hand side, click Results Status. Then, on to second drop-down menu, go to Error and then click Filter.

ah i saw i had like 253 returned results..


hmmm no errors on QX6700 @ 3.5
and i have like P4 and a Dualcore all running differant OS yay for me time to check that box again

[sierra_bound]

253 returned results or 253 work units with errors?

[NKrader]

253 returned results or 253 work units with errors?

returned results out of the ammount of points in my siggy i only have 3 errors total from all projects.

[=[PULSAR]=]

120 Errors since 4/12 works out to roughly 1400-1500 BOINC ppd.

[xenolith]

My HP2 project participation has returned 119 errors, 25 pending WUs, 257 verified WUs, and 0 invalid WUs in the last 13 days since I've joined the team. That's about a 30% error rate. All of them have returned an error within 0.05 CPU time hours except for one WU which returned an error after 32.64 hours. The following is the claimed/granted BOINC credits for WUs that returned errors: 836.25/47.75 (take out that one bad 32.64 hour WU and it would have been 55.75/47.75).

I may be wrong, but I don't think the amount of wasted time I have had yet justifies discontinuing my participation in the HP2 project. Although, I'll have to keep a close eye on this because if I get another WU error that wastes 32+ hours again, I'll have to reevaluate my decision.

This is on two Windows 7 machines w/ i7-860 and i7-720QM.

[D_A]

HPF2 is ... rather error prone on a lot of machines. I refuse to run it to prevent wasting my time as well as a form of protest. A lot of people, I suspect, never check their results status and so never realize they're getting errors on this project.

[mezcal]

Wow, ugly.
Two pages of HP2 errors for me.
Removed from my projects.

Thanks for the info.

[sierra_bound]

I will sticky this thread for a week or two until everyone has a chance to see it.

[mreuter80]

The guys at WCG made some adjustments to the science application (version 6.17) and were able to significantly reduce the errors with that in the last beta test. The results are currently reviewed by the scientists to ensure everything is OK.

Here are two quotes from uplinger:

http://www.worldcommunitygrid.org/fo...d_thread,28940

Sek, the slow part of the link was me on the beta test. The results on the beta test were not 100% complete so I was waiting on that before sending them the results. They recently have received the results and I have asked them to evaluate them and let us know if the results look good. We need to give them time to evaluate the results to make sure everything science wise is still good.

http://www.worldcommunitygrid.org/fo...8872_offset,20

The beta so far has been a good increase over what we saw in production for the same batch.
There were 57 exit 1's for BETA compared to 609 for production.
We are going to send this information to the researchers later today when a few more results come back in foor them to review.

I can't wait to turn it on again

[Sparky]

What gets me is I currently have 10 errored units listed, all from my main rig, none HP2 as that is unchecked. And my system is stock

[=[PULSAR]=]

Which project are they from? No random crashes or system reboots?

[Sparky]

No system appears rock solid. Never had it do anything strange. They are from random projects that I'm running, cancer and FAAH and such.

[fallwind]

You might want to do some stability testing just to be sure there's no issues with the rig. An hour of OCCT Linpack will usually turn up anything that's amiss. Also, if you use antivirus software, go in and exclude the BOINC folders(there's 2 of them) from scanning. This has been known to corrupt WU's too.

[xenolith]

My understanding is this:

error = most likely WCG WU fault (if they're uncommon or have low reproducibility user system fault could be possible cause)
invalid = 100% user system fault (usually due to system instability)

[sierra_bound]

Invalid means the work unit was completed and returned to WCG, but was not found to be similar to the other identical work units that were turned in. So the result gets tossed out.

Errors can be caused by various things - the user aborted the work unit, the work unit itself is bad, or system instability.

A result could also be called "inconclusive". What that means is that the validation program could not determine which of the results returned are valid. When that happens, the same work unit is sent out to more computers and a second validation will be attempted when they're turned in.

[Sam_oslo]

Good to hear it. I've got totally 8 errors, but 7 of them belongs this project. I thought something on my system didn't like them :noob:. It is good to know it wasn't only me.

[gmat]

I had lot of errors on my main rig, found out by searching they were related to that project and not to my overclocking (which should be stabilized by now at 4.15Ghz stage).

[D_A]

A rig can seem stable for general use and still return errors with WCG. If all your trouble was HPF2, then yes, drop that project. If you were getting errors across the board, drop your clock slightly.

[gmat]

All errors were on HPF2 (0.02 cpu time on units etc.) and were the errors reported on the WCG thread. And my rig has gone under some tough stability tests (not general use by far) like Linpack, etc.. BOINC actually produces a BSOD if something's not right, as i found out
I double checked by going to stock clocks, and still getting the errors on that project... Not disabling it for now as units that produce errors only run for a few seconds, not much of a loss.

[Musho]

I've been crunching almost a week now I think. I only got 1 error so far, which errored out in less than a minute, so I'm leaving it enabled now. Q9650 with Windows 7 x65 pro