| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Smash Childhood Cancer
Thread: Points Yield low? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 30
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Why is there an adjustment? Aren't BOINC credits based on the benchmark and run-time? It seems one should get the same credit per unit of time regardless of the length of run-time per WU. If the run-time is relatively fixed per day one should get a fixed number of BOINC credits, give or take a small percent. The number of results per day will vary widely but the points and run-time should stay relatively fixed. The length of the WU could impact the number of validations per day and therefore the amount of runtime credited per day but the ratio of BOINC credit per unit of time should still be fixed. When runtime is relatively flat and point credited drops dramatically (SEE simjoe post above) it would seem something has been changed. I've been running the same one project (SCC1) for over a month and my credit has dropped considerably in the past week or so. I run a relatively large queue so I have just now started running the shorter work units. That means my drop in credit happen during the time the WU length was less variable.
----------------------------------------[Edit 1 times, last edit by Doneske at Mar 26, 2017 2:30:52 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
|
||
|
|
RTS48
Veteran Cruncher Bolivia Joined: Aug 2, 2009 Post Count: 1353 Status: Offline Project Badges: 
|
Thanks SekeRob - I think that I need to sit down in a darkened room and read that article with a calculator and spreadsheet by my side.
----------------------------------------I run benchmarks regularly so I am sure that my device information is up to date. About a month or so ago I ran SCC exclusively. My run time per week is around 135 days (20 cores) and is consistent. I was maintaining a user average (actually rising) of around 10,500 (~70,000 points / day). This is (and still is) plummeting to an average of below 8,800 (~35,000 points / day) for the same CPU / Elapsed time. While I am happy to continue crunching in a good cause I am still competitive enough to chop and change projects if there is a points advantage. I think that desppite all of the fancy mathamatics there is something that is fundamentally wrong if crunchers are rewarded inconsistently for a constant run time.
Rod Peel
Santa Cruz Bolivia South America  ,  |
||
|
|
RTS48
Veteran Cruncher Bolivia Joined: Aug 2, 2009 Post Count: 1353 Status: Offline Project Badges:
|
Here are my daily results for the past 30 days to illustrate the above
----------------------------------------Statistics Date Total Run Tim Points Generated Results Returned 03/27/2017 0:009:23:35:50 19,628 66 03/26/2017 0:019:06:24:19 35,671 131 03/25/2017 0:018:10:55:41 35,183 121 03/24/2017 0:018:22:46:45 34,403 115 03/23/2017 0:019:05:47:28 47,755 116 03/22/2017 0:019:02:30:48 69,070 128 03/21/2017 0:019:02:25:56 68,202 127 03/20/2017 0:019:08:23:24 72,366 126 03/19/2017 0:019:12:06:11 75,629 129 03/18/2017 0:019:21:32:07 81,478 136 03/17/2017 0:019:02:42:37 80,179 133 03/16/2017 0:019:21:48:41 83,481 154 03/15/2017 0:020:05:20:22 91,241 187 03/14/2017 0:019:09:59:23 87,485 255 03/13/2017 0:017:07:58:45 69,537 277 03/12/2017 0:020:00:45:07 62,053 202 03/11/2017 0:018:18:10:20 59,677 191 03/10/2017 0:018:02:37:30 63,525 197 03/09/2017 0:018:13:12:03 73,417 204 03/08/2017 0:019:08:14:10 79,875 202 03/07/2017 0:020:04:16:28 87,029 234 03/06/2017 0:017:23:10:22 77,952 210 03/05/2017 0:017:19:45:05 81,296 215 03/04/2017 0:018:04:24:37 84,390 233 03/03/2017 0:018:19:05:13 84,843 248 03/02/2017 0:020:00:38:58 104,799 323 03/01/2017 0:018:20:20:55 89,673 348 02/28/2017 0:018:19:46:14 66,031 316 02/27/2017 0:019:04:08:15 64,458 276 02/26/2017 0:018:11:39:46 69,869 233
Rod Peel
Santa Cruz Bolivia South America , |
||
|
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18667 Status: Offline Project Badges: 
|
For now, I've been crunching SCC only. Had a big drop in points for the same number of cores running the same amount of time. That seems to have lasted for just a few days and now the numbers for the past couple of days have swung to a bit higher than usual. It's probably due to something specific to the batches being crunched and how the science app reacts to that.
---------------------------------------- |
||
|
|
RTS48
Veteran Cruncher Bolivia Joined: Aug 2, 2009 Post Count: 1353 Status: Offline Project Badges:
|
Once again the points yield for this project plummets. It seems ridiculous that WUs estimated at ~ 21,000 GFlops take ~1h 30m to run whereas WUs estimated as ~17,000 GFlops are taking ~3h to run. The points awarded seem to be proportional to the estimated GFlops rather than the elapsed CPU time. Can this be sorted out please!
----------------------------------------
Rod Peel
Santa Cruz Bolivia South America , |
||
|
|
andgra
Senior Cruncher Sweden Joined: Mar 15, 2014 Post Count: 195 Status: Offline Project Badges:
|
Once again the points yield for this project plummets. It seems ridiculous that WUs estimated at ~ 21,000 GFlops take ~1h 30m to run whereas WUs estimated as ~17,000 GFlops are taking ~3h to run. The points awarded seem to be proportional to the estimated GFlops rather than the elapsed CPU time. Can this be sorted out please! +1
/andgra
 |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
Seems the darkened room had no light, so punch a small hole in the wall, and you'll get a classic Camera Obscura
 (Whatever ridiculousness you see in the execution of the algorithm used, it is what it is, no hope in heck this is going to change, no hope in heck WCG will go it's own way of fixing this i.e. no points [pun], but to resign from the topic, or go somewhere else where they do have ear for the problem... join the Credit_New #4 discussion on the Berkeley forums, which too thoroughly sanded simply because no one in power is interested... the system works on regular runtimes, NOT on variable) |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7844 Status: Offline Project Badges:
|
SekeRob* +1
----------------------------------------Discussed to death already Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
JimWork
Cruncher Canada Joined: Oct 11, 2005 Post Count: 35 Status: Offline Project Badges: 
|
SekeRob* and Sgt.Joe +1
ps Where is KLik? He'll fix it |
||
|
|
|