Folding Early Unit End

RobOC

RobOC

RobOC

Associate
Joined
11 Nov 2004
Posts
522
Hi chaps,

I've been on boinc for a long time with no problems, but having just moved over to give folding a go, I've just had a WU finish early with an error...

Code: 
[14:51:44] Step 3453694, time 6907.39 (ps)  LINCS WARNING
[14:51:44] relative constraint deviation after LINCS:
[14:51:44] max 0.458214 (between atoms 750 and 751) rms 1.#QNAN0
[14:51:44] 
[14:51:44] Simulation instability has been encountered. The run has entered a
[14:51:44]   state from which no further progress can be made.
[14:51:44] This may be the correct result of the simulation, however if you
[14:51:44]   often see other project units terminating early like this
[14:51:44]   too, you may wish to check the stability of your computer (issues
[14:51:44]   such as high temperature, overclocking, etc.).
[14:51:44] Going to send back what have done.
[14:51:44] logfile size: 226932
[14:51:44] - Writing 227615 bytes of core data to disk...
[14:51:45] Done: 227103 -> 10842 (compressed to 4.7 percent)
[14:51:45]   ... Done.
[14:51:45] 
[14:51:45] Folding@home Core Shutdown: EARLY_UNIT_END
[14:51:48] CoreStatus = 72 (114)
[14:51:48] Sending work to server

What could this be down to? I assume overclocking is the likely suspect, but this machine has been stable for ages in boinc. Any ideas?
 
Just like it says in the message:

This may be the correct result of the simulation, however if you
[14:51:44] often see other project units terminating early like this
[14:51:44] too, you may wish to check the stability of your computer (issues
[14:51:44] such as high temperature, overclocking, etc.).
Click to expand...

I had an EUE just the other day even though my machine has been running stable with all the same settings for months - you should only worry if you see a whole bunch of them particularly if it's on different projects

Quite often with the lincs warnings if it does turn out to be your machine at error then simply backing the ram timings off a little is enough to correct it

I would wait and see if the next 2-3 WUs finish alright or not before you go changing anything. Hopefully you just got a duff WU - you'll still be credited for what portion you did (although it's not always overly fair)
 
It's quite likely that overclocking is to blame. BOINC relies on less error-checking and more redundancy in its calculations. FAH is one of the most sensitive projects becasue the managers prefer not to have to check many WUs against each other becasue it slows down the scientific progress. Whenever a WU is processed it is handed back to the servers which package it back up as a WU again. One number is added to the generation number and it continues on.

Is the machine Prime stable? It is Memtest stable?

Thought OC'ing is the likely casue, there are situation in the simulations where it wil reach an impasse and cause an EUE. These are natural and normal, though exceedingly rare. Usually with a EUE you will get partial credit for what has been returned, the amount of which depends mostly on the amount crunched before it EUE'd.

If you see this error just once, I wouldn't be concerned. However, if you are getting early unit ends frequently, your machine is not stable and is producing gibberish work.
 
Last edited:
Thanks guys. I'll see how it goes... though since that was basically my first WU, it doesn't look promising! But yeah it is prime and memtest stable so :confused:
 
You'll always get the odd one or two EUEs, particularly on the newer projects. You only need to worry if you get more than one in a short period of time.

If you're not sure what's going on, somebody can post on the Folding forums with details of the WU - then the Stanford guys can check to see if anybody else completed the WU successfully (EUE'd WUs are reissued.) That can clear up the cause of the EUE. But I'd wait a few days to see how it goes.
 
It always was at this overclock, but admit haven't run prime for some time so am running it now. I guess I should really run it until at least morning for a proper test. Will post you the outcome.
 
good luck with the testing

at least the good news is that you're now officially part of the team :D

the WU that EUE'd on you was a P2106 worth 202 points when complete - you were awarded 69 points :cool: so it must have got pretty far through before the error

Official Stats - EOC Stats

you can even get a tiffy for the 69 points :p
 
I thought you were psychic for a minute then, Rich! Couldn't figure out how you knew it was a 2106. Then I found that detailed listing thing on the official stats which has somehow eluded me up to now. That's pretty cool!
 
Mattus said:
I thought you were psychic for a minute then, Rich! Couldn't figure out how you knew it was a 2106. Then I found that detailed listing thing on the official stats which has somehow eluded me up to now. That's pretty cool!
Click to expand...
what detailed listing thing? :confused: ;)


yeah It wouldn't work once a few more have been done
the one area where F@H is really lacking is in the detail of what work each of your machines has done - there were plans a while ago to be able to identify what work has come from what rig (since each has it's own unique ID) but I think any work done on the stats system would be a mammoth task and due to how it's been designed may not be possible without a stats reset anyway :eek:
 
Well the prime checked out ok, so I restarted the services and a couple of results have since returned with no problems, so I am keeping an eye on it with my fingers crossed for now ;)
 
You must log in or register to reply here.
Back
Top Bottom