But, one comment you made is incorrect. The benchmark is NOT to be optimized to a particular system. If it is, it is no longer an independent measurement. Which is the whole point. Sorry ... :)
When I said "optimized" I obviously mean a benchmark able to do a "good work" even on Linux. Not a good "score".
If a benchmark say "on this pc a wu will take 2 days" and after only 5 hours the wu is completed then the benchmark is not doing a really good work.
A benchmark able to do a good work is a benchmark that actually measures what is designed to measure. The most accurate method is in my opinion a benchmark that actually compute the same kind of data it has to measure.
The idea of small reference work units is not bad... I'll take a look to the developer discussions archive.
If a benchmark say "on this pc a wu will take 2 days" and after only 5 hours the wu is completed then the benchmark is not doing a really good work.
This is what the "duration correction factor" was created to address. At the moment, it is being used to adjust the estimated completion time, and it is doing a very good job of that. Using it to "adjust" claimed credit in the same way has been discussed, and there is even a very simple formula for doing it. What we don't know is if the BOINC folks will use this "simple quick hack" approach, or go ahead and do the full calibrated-host approach, or what...
Some days have passed, I had installed all the new Releases of Boinc and Einstein recommended by all of you and nothing has changed. My new PC works fine with Debian Linux, a WU need some seven (7) hours to complete, same time as before and the credit is round about 42 claimed and mostly !!! also granted - other Linux-User have apparently the same problem (The interesting thing is that a WU is send to PCs with similar power, may it less or high).
My other 3 PCs with much lower perfomace (10 to 12 hours for a WU) but running on MS-Win claimed 60 to 85 cobblestones and have an average of granted credit 75. I think a fair value for computing a WU on the Debian PC was 70 to 90 (claimed). This supposed, I loose an average credit of 90 to 150 cobblestones a day! This is an average of 4,000 a month - not bad for a single PC.
I don't wont to be a nitpicker, but this is not fair to all Linux users who have this problem and I become a little angry because it is only a _heavy bug_ affecting a serious part of the donating community.
I agree highly with basilicofresco and quote one of his statements:
"In my humble opinion a good credit system is important as well the science algorithms for a simple reason: a great number of users (majority) keep heavily crunching for competitive reasons. Even the opinion of "less competitive spirit" users about the fairness of the system is _very_ important.
If the fairness is "felt" weak, than the computing efficiency will be lower."
I compiled the latest boinc-client v.4.72 on SuSE-9.3 64bit. Crunching was working well and the benchmark showed a much better result ....
So, he built his own BOINC executable based on 4.72. I'm guessing he was able to optimise the build using particular complier flags so as to improve the benchmarks in much the same way that folks over at Seti have been building optimised BOINC clients and science apps. I don't think his 100+ claims per result could only come from an optimised BOINC. I think there has to be another factor at work there for the claim to be that high.
I didn't use any special compiler flags, did it right away. And I did nothing else to "optimize" the claimed credits of this CPU. Should say this AMD 64 3000+ is running @2500MHz instead of 1800. :-)
Still this host takes a lot longer to crunch a WU under Linux than under Win.
Maybe that's the reason, why the calculations take so much longer and therefore the claimed credits are around 105.
A better einstein-application, or maybe just one for the many 64Bit CPUs out there, would probably be a big progress.
Quote:
Please realise there are no negative connotations in the above commentary. Good luck to both of them. If I were running linux and was being penalised by low benchmarks around 40-50 then I'd be finding an optimised BOINC as well in order to try to get my claim on a par with what others are claiming.
;-)))
I'm shure I get this right. And yes, I'm in a team and beside the enthusiasm about the science it is a kind of sports to get best credits. But for shure there is no motivation at all to use any "tricks" to get better results. All I do is optimal overclocking and recompiling the BOINC core client with system default compiler switches. ;-)
Done so on an old celeron 433MHz increased the claimed credits from 55 to 61. So the effect was only about 10%, but better than nothing. ;-)
[...]My new PC works fine with Debian Linux, a WU need some seven (7) hours to complete, same time as before and the credit is round about 42 claimed and mostly !!! also granted - other Linux-User have apparently the same problem [...]
Hi Flysoft, can you contact me privately? (why does not exist a private message service on this forum?)
My email address is dave6nopanic.it. (replace the 6 with @)
RE: But, one comment you
)
When I said "optimized" I obviously mean a benchmark able to do a "good work" even on Linux. Not a good "score".
If a benchmark say "on this pc a wu will take 2 days" and after only 5 hours the wu is completed then the benchmark is not doing a really good work.
A benchmark able to do a good work is a benchmark that actually measures what is designed to measure. The most accurate method is in my opinion a benchmark that actually compute the same kind of data it has to measure.
The idea of small reference work units is not bad... I'll take a look to the developer discussions archive.
Thank you for the replies.
RE: If a benchmark say "on
)
This is what the "duration correction factor" was created to address. At the moment, it is being used to adjust the estimated completion time, and it is doing a very good job of that. Using it to "adjust" claimed credit in the same way has been discussed, and there is even a very simple formula for doing it. What we don't know is if the BOINC folks will use this "simple quick hack" approach, or go ahead and do the full calibrated-host approach, or what...
Some days have passed, I had
)
Some days have passed, I had installed all the new Releases of Boinc and Einstein recommended by all of you and nothing has changed. My new PC works fine with Debian Linux, a WU need some seven (7) hours to complete, same time as before and the credit is round about 42 claimed and mostly !!! also granted - other Linux-User have apparently the same problem (The interesting thing is that a WU is send to PCs with similar power, may it less or high).
My other 3 PCs with much lower perfomace (10 to 12 hours for a WU) but running on MS-Win claimed 60 to 85 cobblestones and have an average of granted credit 75. I think a fair value for computing a WU on the Debian PC was 70 to 90 (claimed). This supposed, I loose an average credit of 90 to 150 cobblestones a day! This is an average of 4,000 a month - not bad for a single PC.
I don't wont to be a nitpicker, but this is not fair to all Linux users who have this problem and I become a little angry because it is only a _heavy bug_ affecting a serious part of the donating community.
I agree highly with basilicofresco and quote one of his statements:
"In my humble opinion a good credit system is important as well the science algorithms for a simple reason: a great number of users (majority) keep heavily crunching for competitive reasons. Even the opinion of "less competitive spirit" users about the fairness of the system is _very_ important.
If the fairness is "felt" weak, than the computing efficiency will be lower."
That's it !
Ulrich
Hi Gary, RE: Hi
)
Hi Gary,
I didn't use any special compiler flags, did it right away. And I did nothing else to "optimize" the claimed credits of this CPU. Should say this AMD 64 3000+ is running @2500MHz instead of 1800. :-)
Still this host takes a lot longer to crunch a WU under Linux than under Win.
Maybe that's the reason, why the calculations take so much longer and therefore the claimed credits are around 105.
A better einstein-application, or maybe just one for the many 64Bit CPUs out there, would probably be a big progress.
;-)))
I'm shure I get this right. And yes, I'm in a team and beside the enthusiasm about the science it is a kind of sports to get best credits. But for shure there is no motivation at all to use any "tricks" to get better results. All I do is optimal overclocking and recompiling the BOINC core client with system default compiler switches. ;-)
Done so on an old celeron 433MHz increased the claimed credits from 55 to 61. So the effect was only about 10%, but better than nothing. ;-)
cu,
Michael (ziegenmelker)
RE: [...]My new PC works
)
Hi Flysoft, can you contact me privately? (why does not exist a private message service on this forum?)
My email address is dave6nopanic.it. (replace the 6 with @)