Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
World Community Grid Forums
Category: Completed Research Forum: Nutritious Rice for the World Thread: Machines granted different BOINC credit/points for same unit |
No member browsing this thread |
Thread Status: Active Total posts in this thread: 21
|
Author |
|
marky1124
Cruncher Joined: Jan 10, 2005 Post Count: 29 Status: Offline Project Badges: |
Hi,
I am noticing that this project seems to grant BOINC credits differently from others. My understanding is that at the usual way is to average the claimed points across the contributing machines, ignoring the highest and lowest, and then grant that number of points to each machine that contributed. However with this project I'm seeing each machine be granted a different number of points which varies widely, so my machine was granted 20 points whilst others were granted upto 196. That's a big difference. Each work unit I've looked at has shown this type of difference. Here's an example Workunit Status Project Name: Nutritious Rice for the World Created: 05/15/2008 13:16:52 Name: R00003_197b03639b10295035bea93b0cf04ed1_08 Minimum Quorum: 14 Initial Replication: 19 The large number of copies sent out for this workunit is due to the unique nature of this project. We encourage you to read the FAQs about this project for more information. Result Name Status Sent Time Time Due / Return Time CPU Time (hours) Claimed/ Granted BOINC Credit R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 19-- Valid 05/15/2008 19:53:50 05/16/2008 18:02:47 8.00 185.3 / 190.0 R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 12-- Valid 05/15/2008 18:55:54 05/16/2008 11:20:49 8.04 51.0 / 43.4 R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 18-- Valid 05/15/2008 18:50:37 05/17/2008 10:14:35 8.01 51.2 / 39.3 R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 0-- Valid 05/15/2008 18:45:54 05/16/2008 19:27:44 8.01 144.4 / 196.5 R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 14-- Valid 05/15/2008 18:22:45 05/17/2008 14:56:47 8.03 47.0 / 48.4 R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 10-- Valid 05/15/2008 18:17:23 05/18/2008 08:47:26 8.03 57.5 / 65.5 R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 4-- Valid 05/15/2008 18:14:30 05/17/2008 22:27:57 8.05 40.7 / 37.7 R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 8-- Valid 05/15/2008 18:05:08 05/17/2008 17:31:59 8.00 119.1 / 127.6 R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 6-- Valid 05/15/2008 18:04:06 05/18/2008 23:38:02 8.02 56.6 / 57.7 R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 16-- Error 05/15/2008 18:01:11 05/22/2008 15:38:22 0.00 0.0 / 0.0 R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 2-- Valid 05/15/2008 17:59:56 05/17/2008 15:24:02 8.07 20.1 / 20.9 R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 17-- Valid 05/15/2008 17:53:50 05/16/2008 15:37:43 8.00 97.7 / 107.6 R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 5-- Valid 05/15/2008 17:32:28 05/19/2008 15:52:53 8.02 61.3 / 58.0 R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 15-- Valid 05/15/2008 17:31:43 05/17/2008 09:07:49 8.02 65.8 / 60.2 R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 13-- Valid 05/15/2008 17:30:49 05/21/2008 08:06:43 8.00 76.7 / 86.4 R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 11-- Valid 05/15/2008 17:27:15 05/16/2008 13:43:34 8.02 122.1 / 140.7 R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 3-- Valid 05/15/2008 17:25:22 05/20/2008 13:09:11 8.01 75.2 / 77.1 R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 9-- Error 05/15/2008 17:22:57 05/19/2008 23:23:57 0.00 0.0 / 0.0 R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 7-- Valid 05/15/2008 17:17:47 05/16/2008 12:46:37 8.02 119.0 / 101.1 R00003_ 197b03639b10295035bea93b0cf04ed1_ 08_ 1-- Error 05/15/2008 17:16:58 05/15/2008 19:47:57 0.00 0.0 / 0.0 I picked a very old unit so that every machine had finished, the same type of result can be seen with the latest units as well. Please can someone confirm whether this is wrong? I know that contributing is the most important thing, but it's nice to see ones progress up the ranks as well. Cheers, Mark |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
No, this is all perfectly correct.
Nutritious Rice for the World is slightly unusual because work units run for the same amount of time on fast machines and slow machines. In that time, each computer given the work unit predicts as many structures as it can. The validator works out an average credit like normal, but per structure, not for the whole work unit. Then the credit is multiplied by the number of structures. And that's the credit you get (it should be fairly close to your claimed credit, and as you can see from your example, that is usually the case). Each Nutritious Rice for the World work unit runs for exactly eight hours. |
||
|
marky1124
Cruncher Joined: Jan 10, 2005 Post Count: 29 Status: Offline Project Badges: |
Ah interesting. Thanks very much for the explanation.
Cheers, Mark |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
So what you mean is that if you have a faster machine she will do more structures in 8 hours so then she gets more points, versus a slower machine which in the same time will produce a lower amount of structures during the 8hr period.
Is this correct? in other words the WU will benefit the ones with a fast pc versus anyone which puts its time to work them. So say I got say a 1.2mhz pc and she finds 10m structures in the same WU and you have a 3.6 mhz and you find 50 then you get more credits!! What about if out of the 10 machines only 4 go thru the full WU and find 50 , but the slower machine also goes thru the WU but find only 10, they only get credit for the 10, what about confirmation on the structures found by the faster machine, versus the other structures which I supposed must be confirm to be elegible? Its not fair that if everyone will have to put the same amount of time 8hrs as per the project say's. then everyone should get the same amount of credit as long as you finish your 8 hrs regarless. We are talking about time, not structures or otherthings. I like to have my machine work, but I don't think it should be this way, cuz 8 hr is a long time and you can have lots of production in other projects as well as in other studies and get a decent credit for your time. I know that this is important, but time is time and if the project will run for 8 hrs regarles then all should get the same credit regarless!!!! Thanks Dr Iggy |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Rather lots of words for the simple answer that even the slowest machine can attack the rice project and finish a job in time, albeit with less seeds completed, thus less points.
Many contributors would love to have all WCG projects do this or be able to choose run time. Somewhere in the BOINC future WCG will be able to send the lighter work to the lesser shouldered devices ending the moan about the long jobs. |
||
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3715 Status: Offline Project Badges: |
Dr Iggy, that's why WCG has three classifications, one by points, one by returned results and one by time.
----------------------------------------The classification by time is the one you are interested with and it measures exactly what you say, the machine time that participants have offered to research. You will notice that it is very often this measurement that WCG puts forward, which confirms that the time you offer matters at least as much as the computation you produce. The classification by points measures the actual work produced and we (WCG and participants) need that too, for quickly evaluating if the whole system, a team or a given machine is producing what one could expect from it. Surprising variations of this point measurement has often helped to detect anomalies at individual or more global level. The classification by results is more "for fun" for participants since it adds up very dissimilar things like the short WUs of Genome comparison and the monster ones of HPF2. But when looking at a given project they greatly help to figure out where we stand regarding the progression of this particular project. And anyway, you are not crunching for points but for helping Research, no? Cheers. Jean. |
||
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3715 Status: Offline Project Badges: |
I know that this is important, but time is time and if the project will run for 8 hrs regarles then all should get the same credit regarless!!!! There is no project that I know of that will credit the same 8 hours of an 8-core clover and 8 hours of an older PIII. The difference in Rice is that in usual projects this is achieved through a different number of WUs. But the result is the same. For Rice what the scientists are interested with is structures, then it is easy to achieve this same-duration scheme which is convenient for all machines and the servers, and avoids deadlines problems. And, as usual, faster machines return more results (structures in that case) and are granted more points than slower machines. Cheers. Jean. |
||
|
trevorcolby00
Advanced Cruncher Joined: Aug 11, 2005 Post Count: 101 Status: Offline Project Badges: |
Hey Jean,
----------------------------------------very well explained - thank you very much !!! Now things became much clearer according to this project. |
||
|
mreuter80
Advanced Cruncher Joined: Oct 2, 2006 Post Count: 82 Status: Offline Project Badges: |
yeah man,
----------------------------------------now I understand ... if you have an old computer use this project ... if you have a superduperheckmachine computer don't even think about the project .. you're ****......... [Edit 2 times, last edit by TKH at Jul 7, 2008 11:43:01 AM] |
||
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
mreuter80,
----------------------------------------seems you (still) have not understood the explanations. First off, all tasks are unique and are only pooled for the purpose to compute the average claim per seed computed in a job and use that as the factor. A superduperheckmachine does enormously more seeds in 8 hours than a middle of the road or crawler. Narrowing down the pool to just 2 machines you get e.g. P4: 8 hours doing 100 seeds claiming 65 QX9750: 8 hours doing 1000 seeds claiming 650 The total pool generated 1100 seeds and 715 credit claim which gives: P4: 100 * 715 / 1100 = 65 credit granted QX9750: 1000 * 715 / 1100 = 650 credit granted That was a perfect world example but almost without exception is my superduperheckbendmachine proportionally getting 5-15% more than claimed as award. It does more seeds per unit of time than the average of the first 10 results in the pool that are used to determine the claim per seed (some refer to here as structure). All this is why you get the GRANTED CREDIT VARIATION PER RESULT as shown in the quorum of the opening post. cheers
WCG Global & Research > Make Proposal Help: Start Here!
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 1 times, last edit by Sekerob at Jul 4, 2008 6:47:32 AM] |
||
|
|