| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Help Stop TB
Thread: Ridiculous low credit claim for longer running tasks |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 99
|
|
| Author |
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
What's happening on your machine may be alluring me, but what's on mine has been stopped [for now]. Did some shake, rattle, and Rick Astley rolling [upped the herz to 3Ghz, lowered a locked benchmark down to what it was at 2.5] and got newest to oldest order of completion claims of > 500. Efficiency not a problem... 7 cores running now HST, 1 OET1 left overs.
HST1_000440_000071_AT0018_T325_F00095_S00001_0 249 hst1 Valid 539,706 15,9107 0:15:54:39 16,0494 0 458,77 28,83 99,14% HST1_000440_000063_AT0018_T325_F00087_S00001_0 249 hst1 Pending Validation 571,966 15,6879 0:15:41:16 15,8189 0 0,00 36,46 99,17% HST1_000436_000023_AT0014_T325_F00047_S00001_1 249 hst1 Pending Validation 572,678 15,6895 0:15:41:22 15,8258 0 0,00 36,50 99,14% HST1_000204_000013_MT0010_T350_F00013_S00001_2 249 hst1 Valid 413,946 15,3774 0:15:22:39 15,4966 0 387,50 25,20 99,23% HST1_000409_000083_AC0028_T325_F00007_S00001_0 249 hst1 Valid 168,615 15,8841 0:15:53:03 16,2023 0 324,43 20,42 98,04% HST1_000331_000027_KC0010_T350_F00027_S00001_0 249 hst1 Valid 395,609 17,0036 0:17:00:13 17,1585 0 386,49 22,73 99,10% HST1_000012_001121_AC0007_T350_F00021_S00001_1 249 hst1 Valid 492,341 16,5963 0:16:35:47 16,7815 0 307,18 18,51 98,90% Surely this is going to call for some new silver-iodine spraying on the clouds to make it rain again.  |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1684 Status: Offline Project Badges: 
|
Hi guys,
----------------------------------------since the HST project seems to be specially demanding, I devoted only not hyperthreaded hosts to it. However, I noticed regularly on the same host (Phenom II x6, Ubuntu 14.04 x64) a factor 4 difference between OET (about 60 points / hour / core) and HST (average of 15 points / hour / core, with >99% CPU efficiency). By the way, HST is worst than FA@H projects in terms of granting credit. This huge difference is particularly noticeable for long WUs (5+ hours). Now my recurrent question: Is the poor credit only a credit calculation problem or does it show a poor computation efficiency? Beside the credit trouble demotivating some contributors, my concern is more specific on computation efficiency. If the project is not efficient, we should wait for improvement and optimisation and devote our resources for better projects. Cheers, Yves |
||
|
|
Mathilde2006
Senior Cruncher Germany Joined: Sep 30, 2006 Post Count: 269 Status: Offline Project Badges: 
|
Poor wingmen- just crunching the repair:
----------------------------------------HST1_ 000326_ 000040_ KC0008_ T300_ F00040_ S00001_ 1-- Microsoft Windows 7 Professional x64 Edition, Service Pack 1, (06.01.7601.00) 714 Pending Validation 28.03.16 09:03:13 30.03.16 10:59:20 37.69 176.1 / 0.0  |
||
|
|
Crystal Pellet
Veteran Cruncher Joined: May 21, 2008 Post Count: 1407 Status: Offline Project Badges: 
|
Yves Samson wrote:
----------------------------------------Now my recurrent question: Is the poor credit only a credit calculation problem or does it show a poor computation efficiency? Computation efficiency for HST1 is >98%. On the 'BOINC only AMD Opteron' even 99,56%. The calculation problem is caused by initial too short estimated run times. The tasks run 5 times longer and so the machine's claim is corresponding higher. Probably WCG has an inbuilt anti-cheating mechanism for credit claim and/or run time and so those crunchers are punished with a much lower claim. [Edit 2 times, last edit by Crystal Pellet at Apr 7, 2016 11:37:30 AM] |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1684 Status: Offline Project Badges:
|
Thank you CP for your explanation.
----------------------------------------Until now, I was not aware about how "crucial" is the role of the estimated time. If the credit calculation is as you described, the correction of this credit problem should be easy. Why is it not done? Currently, over 50% of the WUs crunched by my hosts are long (6 - 17 hours) WUs. As already mentioned, the impact of this forecasting mistake is significant. We have now only to hope to better estimated time. Cheers, Yves |
||
|
|
zaschf
Advanced Cruncher New Zealand Joined: Jan 28, 2009 Post Count: 61 Status: Offline Project Badges: 
|
I'm getting a bit brassed off with the awarding of credits.
----------------------------------------As soon as a task goes over 12.5 hours the credits are fixed down to 266.9 while tasks below 12.5 hours are awarded the expected credits. Two examples: Normal credits: HST1_ 000767_ 000059_ AT0008_ T300_ F00083_ S00002_ 0-- homer2 Valid 16/04/16 18:33:33 17/04/16 13:14:31 12.40 / 12.60 562.8 / 481.8 And fixed low credits: HST1_ 000725_ 000056_ AC0014_ T350_ F00056_ S00002_ 0-- homer2 Valid 15/04/16 05:33:07 15/04/16 19:09:23 12.96 / 13.18 266.9 / 365.9 As a result of the fixed lower credit, my wingman gets pulled down too. Something isn't right here and should be sorted out.
Ubuntu 22.04.1 LTS [Linux 5.15.0-47-generic]
CPU: Intel Core i7-9700 @3.00GHz x 8 GPU: NVIDIA GeForce RTX2060 Rev. A plus a mighty Raspberry PI 4 |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
Credit_New (apparently 3rd generation credit system for BOINC), has proven to be more broken than the camel's back. There's a discussion over at Berkeley which also touches on how the Gridcoiners seem to have found holes to further pervert the Swiss cheese ensemble... Remember, the distributed computing is to a subset of 'volunteers' no longer about advancing science, rather about fake how great I look, or how to cheat the 'system', therefor there being [legacy] outlier rules that further mince the corpse it is.
----------------------------------------[Edit 1 times, last edit by SekeRob* at Apr 19, 2016 11:14:13 AM] |
||
|
|
enels
Senior Cruncher Joined: Apr 25, 2008 Post Count: 286 Status: Offline Project Badges: 
|
My Atom Z3740 takes 65 hours and claims the expected 266.9 points. Dragging my wingman down with me. So I've quit that experiment.
|
||
|
|
uplinger
Former World Community Grid Tech Joined: May 23, 2005 Post Count: 3952 Status: Offline Project Badges: 
|
I have it on my list of things to look into, but the maxing out on 266.9 does seem like a bad limitation for this project. I will see what I can do to prevent it.
Thanks, -Uplinger |
||
|
|
vjahn1
Cruncher Joined: Apr 29, 2006 Post Count: 8 Status: Offline Project Badges: 
|
I wrote support over a month ago and in response, they claim they always monitor the Forums and implement corrective action when needed.
So, has anything changed? Who and where are the techs assigned to this project? They should have responded weeks ago to the many volunteers who have already voiced their observations. Lets give project management a few more days to respond before we escalate these issues with upper management. |
||
|
|
|