Message boards :
Number crunching :
Differences in claimed credits?
Message board moderation
Author | Message |
---|---|
jeffusa Send message Joined: 21 Aug 02 Posts: 224 Credit: 1,809,275 RAC: 0 |
I am wondering if there is a way to optimize your system to get more claimed credits. It is weird but I have noticed faster machines claim more credit than my machine and then I have noticed faster machines claiming less credit? I don't get it. What is going on here? |
PhonAcq Send message Joined: 14 Apr 01 Posts: 1656 Credit: 30,658,217 RAC: 1 |
Look at the seti version. versions 4.11 are optimized and faster computers claim less credit. a faster computer using 4.09 will claim more. Then the stupid credit calculator throws out the high and low and credits the midpoint, irrespective of the seti version being used. I sure hope the caculation results are being used efficiently, because the algorithms seem pretty ad hoc. May this Farce be with You |
mikey Send message Joined: 17 Dec 99 Posts: 4215 Credit: 3,474,603 RAC: 0 |
I am wondering if there is a way to optimize your system to get more claimed credits. It is weird but I have noticed faster machines claim more credit than my machine and then I have noticed faster machines claiming less credit? I don't get it. What is going on here? Everyone that crunches a unit gets the same credit, your asking for more is not going to affect that. It COULD affect the amount IF you were consistently higher than all the other people crunching the same unit AND ALWAYS returned the unit BEFORE the 4th person. Then you would be depending on the person with the middle amount of credit to also be high so you could get a good amount. Too many coulds and ifs and ands for me. The system is not perfect, some very smart people are beginning to look at it, but as it is it works the same for everyone. |
PhonAcq Send message Joined: 14 Apr 01 Posts: 1656 Credit: 30,658,217 RAC: 1 |
yeah, that 4th person gotcha really stinks too. I don't like seeing my electrons flow for no credit. why bother with the 4th person, who is likely to be a looser and a waste of resources. I guess that the third person may never return a result, and so the 4th person will. But I fail to see why there is a time limit ( at least a short one like two weeks ). Since the seti wu's don't take all that much time to compute, say 24h on a slow or busy machine, why not just wait for the two weeks and if the wu is not closed, then issue the 4th wu to someone. Then if the slow poke returns the 4th unit, give him at least some credit for the effort. Maybe issue that 4th unit to the fastest turn around time clients so that it will return more quickly than average. Jeez, each wu gets computed at least 4 times and by plan the 4th (or more) result get absolutely no credit. Again, it doesn't take an Einstein to design a better system, I hope. May this Farce be with You |
Jord Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 |
why bother with the 4th person, who is likely to be a looser and a waste of resources. If you'd like to designate yourself just that... One of your own units, you were the 4th returning it. :) Jeez, each wu gets computed at least 4 times and by plan the 4th (or more) result get absolutely no credit. Says who? Have you even checked out all the 4th results? Which didn't get credit? Again, it doesn't take an Einstein to design a better system, I hope. Hmm, yes, might be time they disabled the credits altogether. Then no-one would have to complain again, since there's nothing to complain about... Or is there? :) |
W-K 666 Send message Joined: 18 May 99 Posts: 19085 Credit: 40,757,560 RAC: 67 |
The reason for the credits varying is because the benchmarks used: a. not really suitable for any one particular project, The powers that be are thinking about this problem as it affects the calculated processing time and therefore the performance of the scheduler. b. do not test the computer system fully, they basically only test the cpu and a small bit of cache memory. They send out four units, but only need three successful results that roughly agree to verify the result, as its probably more efficient for the data server, as if they only sent three and one was not returned before the deadline then it would have to be kept on the server for another 14 days, taking up valuble server disc space (if you've got a few thousand spare $ they could use it for more server capacity). The benefit to those worried about credits is that they are granted earlier also. If you are the fourth to return a successful result you still get credits granted, you just don't take part in the calculation, normally, the main exception being just after communication outages. At least under Boinc the seti units only get processed 4 times not 400 times as happened with Seti Classic Andy |
PhonAcq Send message Joined: 14 Apr 01 Posts: 1656 Credit: 30,658,217 RAC: 1 |
Ageless: my bad. My impression is dated. I looked and indeed most of the time all 4 credits are granted. However, take a look at these, they seem to be counter examples. 1 2 3 Yes, my theme this month is that SETI doesn't seem to use the compute resources very efficiently/wisely. I don't care if I don't get credit for the 4th duplicate calculation. I assume there are other wu's I can actually contribute doing. Yes. Do away with the credits. It would be good if one could use them to understand the computational science of it all, but I'm beginning to think the credits can't help if they aren't well planned. Oh, well. Nobody would volunteer FPU hours unless they get a score, I suppose. Cheers May this Farce be with You |
Jord Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 |
1 1 and 2. These were in outside the deadline (30th and 31st of May respectfully). It says in the Result ID's Validate state: Result was reported too late to validate. 3. This one is outside the quorum, because that PC thought the unit was one filled with garbage etc. SETI@Home Informational message -9 result_overflow NOTE: The number of results detected exceeds the storage space allocated. Such units run for mere seconds, then return their outcome to the server as they cannot be continued. Nothing wrong with that, it happens to us all. :) |
PhonAcq Send message Joined: 14 Apr 01 Posts: 1656 Credit: 30,658,217 RAC: 1 |
Ok, I didn't know where to look for that information. On the face of it, the table is misleading, at least until you drill down aways. That is the table says the computation was a success but received no credit. Can you take a look at this? I don't understand it. Something is invalid after 8000 sec of work. Can't debug the problem without more information. May this Farce be with You |
mikey Send message Joined: 17 Dec 99 Posts: 4215 Credit: 3,474,603 RAC: 0 |
Ok, I didn't know where to look for that information. On the face of it, the table is misleading, at least until you drill down aways. That is the table says the computation was a success but received no credit. When you get to the chart with 2 sets of numbers in the left column, click on the far left one and you will see more about the status of the computer that has returned the result. |
TPR_Mojo Send message Joined: 18 Apr 00 Posts: 323 Credit: 7,001,052 RAC: 0 |
If you click on the associated "Result" tag you will see validate state is "invalid", which as I understand it means your returned result did not match those submitted by the other two machines in the quorum. HTH |
mikey Send message Joined: 17 Dec 99 Posts: 4215 Credit: 3,474,603 RAC: 0 |
[quote]
Which in turn means that you could be overclocking too much and the computer is too close to the top edge. If you are not overclocking it could be a whole range of things from system running too hot, bad ram, etc. Basically it means that if it continues you need to do a test of your system because it is producing bad results. |
jeffusa Send message Joined: 21 Aug 02 Posts: 224 Credit: 1,809,275 RAC: 0 |
So is version 4.09 the best version for claiming the most credit? Where can I download this? Any bad bugs compared to 4.43? |
PhonAcq Send message Joined: 14 Apr 01 Posts: 1656 Credit: 30,658,217 RAC: 1 |
I'm running two boxes on seti/boinc and they are stock systems from Dell; nothing special (centrino and P4). However, they are both running optimized clients. I'll keep a look out for repeated issues, thanks. May this Farce be with You |
Pascal, K G Send message Joined: 3 Apr 99 Posts: 2343 Credit: 150,491 RAC: 0 |
So is version 4.09 the best version for claiming the most credit? Where can I download this? Any bad bugs compared to 4.43? 4.09 is the application version and 4.43 is the CC version. Basically the app processes the WU and the CC controls up and download of the WU...Credit depends on cpu speed and benchmark Semper Eadem So long Paul, it has been a hell of a ride. Park your ego's, fire up the computers, Science YES, Credits No. |
Professor Desty Nova Send message Joined: 17 May 99 Posts: 59 Credit: 579,918 RAC: 0 |
Another thing contributing to the diference in credits is people still using pre-BOINC 4.25 CC. Remember that the Windows BOINC until 4.19 had inflated benchmarks that resulted in higher credit. This was due to the MS compiler "optimizing" the benchmark code (it was throwing away parts of it). Only in 4.25 and above was this corrected. So people using these old BOINC versions always claim more credit. SETI@home classic workunits: 1,985 CPU time: 24,567 hours Professor Desty Nova Researching Karma the Hard Way |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.