Credit for April Fools
1 Apr 2005 5:01:40 UTC
Topic 188832
(moderation:
This morning (actually the morning of March 31) the credit claimed by one of my computers ("addtron-ga")dropped abruptly from about 68 to 43 points. It has stayed at 43 since then. I am not aware of any changes on my end, and the time to complete each WU is almost exactly the same as before.
Is this an April Fools Day joke by the programming staff?
ADDMP.
Credit for April Fools
)
Have you updated from 4.1x to 4.2x inbetween and run windoze?
The old benchmark cheated, and claimed more credit than it rightful should.
Grüße vom Sänger
> This morning (actually the
)
> This morning (actually the morning of March 31) the credit claimed by one of
> my computers ("addtron-ga")dropped abruptly from about 68 to 43 points. It has
> stayed at 43 since then. I am not aware of any changes on my end, and the time
> to complete each WU is almost exactly the same as before.
>
> Is this an April Fools Day joke by the programming staff?
This should only occur if your benchmark results have changed.
To find this out, you can search in the "stdoutdae.txt" file when the last benchmark has occured and compare the values of that one with the earlier benchmark results.
If the latest benchmark is indeed lower compared to the previous one, you can run the benchmark once again and then do an "Update".
Greetings from Belgium
Thierry
> To find this out, you can
)
> To find this out, you can search in the "stdoutdae.txt" file when the last
> benchmark has occured and compare the values of that one with the earlier
> benchmark results.
Thanks.
There is a file "stdoutdae.txt". It has only two lines, both from March 11. The first says there is no work & the second says it is deferring communication for one minute.
ADDMP
> Have you updated from 4.1x
)
> Have you updated from 4.1x to 4.2x inbetween and run windoze?
> The old benchmark cheated, and claimed more credit than it rightful should.
Thanks,
I'm not sure what you mean by "in between". I have made no changes recently. Before March 11, this computer was running E@H with a different computer name under linux. That seemed very inefficient, so I changed to WIN2000 on March 11 & I changed the computer name at the same time. On March 11, I did install BOINC 4.25. Before March 11, I was using BOINC 4.19 bacause I was not abls to get version 4.19 to run under wine-linux.
ADDMP.
As each WU result is returned
)
As each WU result is returned to the server, the benchmark score of that machine at the time of completion is also uploaded to the server (in the same message).
And so whatever your benchmark was then was creating your other claims...and what it is now apparently causes your current claims.
You had another host where the older numbers were in the 43 range, but the newer ones are in the 68 range (the AMD Athlon running "Win98se"). That changeover occured approx. 3-28-05.
2520296 648774 28 Mar 2005 11:28:36 UTC 29 Mar 2005 3:53:58 UTC Over Success Done 29,637.33 68.22 68.31
2511778 646825 28 Mar 2005 3:09:54 UTC 28 Mar 2005 19:40:38 UTC Over Success Done 29,838.24 45.22 76.73
> As each WU result is
)
> As each WU result is returned to the server, the benchmark score of that
> machine at the time of completion is also uploaded to the server (in the same
> message).
>
> And so whatever your benchmark was then was creating your other claims...and
> what it is now apparently causes your current claims.
Thanks,
Yes, I've now checked the BOINC messages, & what happened was that new a benchmark calculation was run that reduced the claimed credits from about 68 to about 43. I think the original benchmark that was producing credits of 68 was something that had been carried over from previous incarnations of this computer. The computer had been started running native linux E@H & then switched to wine-linux & finally to win2000. So the most reasonable claimed credits of 68 were produced by obsolete & inappropriate benchmarks & the most unreasonable claimed credits of 43 were produced by up-to-date benchmarks.
That seems to mean a problem in BOINC. Or perhaps there is still some sort of contamination by the previous benchmarks that aren't fixed by re-running the benchmark calculation & that I don't know how to wipe out.
> You had another host where the older numbers were in the 43 range, but the
> newer ones are in the 68 range (the AMD Athlon running "Win98se"). That
> changeover occured approx. 3-28-05.
>
> 2520296 648774 28 Mar 2005 11:28:36 UTC 29 Mar 2005 3:53:58 UTC Over
> Success Done 29,637.33 68.22 68.31
> 2511778 646825 28 Mar 2005 3:09:54 UTC 28 Mar 2005 19:40:38 UTC Over
> Success Done 29,838.24 45.22 76.73
Right. I tried to check the BOINC messages tab for that computer too, but they don't go back far enough to see what might have happened on Mar 28. My guess is that there waa also a benchmark calculation run by BOINC around Mar 28. On this computer, the corrected benchmarks made the claimed credits MORE reasonable. Note also that the CPU seconds barely change whereas the claimed credits change quite significantly, so this glitch cannot plausibly be attributed to any change in the difficulty of the units being processed. It is a BOINC change & not an E@H change.
ADDMP