Message boards : Number crunching : Problem with WU
Message board moderation

To post messages, you must log in.

AuthorMessage
vlmath

Send message
Joined: 2 Mar 08
Posts: 4
Credit: 1,052
RAC: 0
Message 20826 - Posted: 27 Nov 2008, 13:51:54 UTC
Last modified: 27 Nov 2008, 14:43:22 UTC

Hello,

I have a WU (n° 16109076) which take 18 houres for 1.000.000 turns using SixTrack 4.67 ... I think it\'s not normal, is it ?

For the other tasks, it take normally not more than 4 hours (for 1.000.000 turns).

Bye
ID: 20826 · Report as offensive     Reply Quote
jrobbio

Send message
Joined: 12 Sep 08
Posts: 10
Credit: 2,747
RAC: 0
Message 20830 - Posted: 28 Nov 2008, 11:08:28 UTC - in response to Message 20826.  

Your floating points scores aren\'t particularly impressive.

Floating Point = 934.86 million ops/sec
MIPS = 2048.08 million ops/sec

A computer with the same task, which did it in a third of the time had 3 times the CPU performance

Floating Point = 2931.25 million ops/sec
MIPS = 7365.98 million ops/sec

The T model Core Duo processors are designed for low wattage efficient performance. Running at ~85W your laptop has a 6th the power of my desktop. Your other computer would have done it twice as fast.

Rob
ID: 20830 · Report as offensive     Reply Quote
vlmath

Send message
Joined: 2 Mar 08
Posts: 4
Credit: 1,052
RAC: 0
Message 20832 - Posted: 28 Nov 2008, 21:19:19 UTC - in response to Message 20830.  

Thank\'s Rob.

But comparing to myself, it was anyway much more longer.

My score is not faramonus because it\'s a portable computer always on \"power save\" :D
ID: 20832 · Report as offensive     Reply Quote
Michael D. Belanger

Send message
Joined: 24 Jan 09
Posts: 3
Credit: 117
RAC: 0
Message 21142 - Posted: 30 Jan 2009, 12:43:09 UTC

I had 2 WU\\\\\\\'s finish so far and they both were \\\"Aborted by project\\\"; why is this? Will I still get whatever credit(s)?
ID: 21142 · Report as offensive     Reply Quote
rembertw

Send message
Joined: 21 Apr 07
Posts: 42
Credit: 409,087
RAC: 0
Message 21146 - Posted: 30 Jan 2009, 18:07:37 UTC - in response to Message 21142.  

\"Aborted by project\"; why is this?


I can not answer with certainty about the credit story, I will leave this for someone else to answer.

The \"why\" question is easier to answer: every work unit is sent out immediately 5 times. Every time is one task. The moment 3 tasks are returned with identical results the other 2 tasks around get canceled. This is what happened with your 2 tasks. 3 others also had your Work Unit but were faster returning the result.

Depending on the speed of your computers you can count on an average of 40% (2 out of 5 tasks) that will get canceled at some time. If you want to be more certain that you actually can crush the tasks that are allotted to you, then the best way is to keep your queue as short as possible. This way you get constantly tasks from \"fresh\" Work Units, with less chance that others are faster. Or with bigger chance that their tasks are somewhere down their queue.
ID: 21146 · Report as offensive     Reply Quote
Mike.Gibson

Send message
Joined: 14 Nov 07
Posts: 11
Credit: 215,982
RAC: 42
Message 21147 - Posted: 31 Jan 2009, 0:17:02 UTC - in response to Message 21142.  

I had 2 WU\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\'s finish so far and they both were \\\\\\\\\\\\\\\"Aborted by project\\\\\\\\\\\\\\\"; why is this? Will I still get whatever credit(s)?


It is not quite as simple as rembertw said.

Each WU is sent out to 5 people and once 3 identical replies have been received, an abort order is sent out to anyone who they don\\\'t know as having started the WU. The idea is so that you can move onto a more productive WU.

Unfortunately they don\\\'t know that a WU has started unless you have had a project update since it was started. You will not get any credits for those WUs. If they know that someone has started the WU, they are allowed to continue and get the credit.

Many people have their computers on permanently and have a high resource share for LHC@home, so they pick up the WUs as soon as they become available and start crunching immediately. Many WUs also finish very quickly so it can be that the Abort message is issued only half an hour after the WU has been issued.
ID: 21147 · Report as offensive     Reply Quote
noderaser
Avatar

Send message
Joined: 4 Oct 05
Posts: 32
Credit: 344,444
RAC: 0
Message 21148 - Posted: 31 Jan 2009, 4:28:40 UTC

If it was aborted by the project, it hadn\'t started running yet. So, there is no credit to lose.
Click here to see My Detailed BOINC Stats
ID: 21148 · Report as offensive     Reply Quote
Mike.Gibson

Send message
Joined: 14 Nov 07
Posts: 11
Credit: 215,982
RAC: 42
Message 21152 - Posted: 31 Jan 2009, 10:29:01 UTC - in response to Message 21148.  

If it was aborted by the project, it hadn\\\'t started running yet. So, there is no credit to lose.


There are some which slip through the net. If a WU has started but there hasn\'t been an update since it started, then it can still abort because the project doesn\'t know it that it has started. Or it may be the BOINC Platform rather than the project. Either way, hit the update and it doesn\'t abort.
ID: 21152 · Report as offensive     Reply Quote

Message boards : Number crunching : Problem with WU


©2024 CERN