Automatic Benchmarking
31 Aug 2005 16:47:18 UTC
Topic 189809
(moderation:
Hello, all
Does anyone know if automatic benchmarking can be disabled? If so, how?
The auto-benchmark runs every 5 days or so, and often at inconvenient times, i.e. while ny comp is multitasking or otherwise distracted from Einstein, resulting in considerably lower benchmark #s and thus lower credit.
And yes, numbers DO matter! Normal human nature desires gratification, and the #s are all we get, no tax write-off for donating considerable valuable computer time to a non-profit organization.
microcraft
"The arc of history is long, but it bends toward justice" - MLK
Automatic Benchmarking
)
From my experience, you can force the core client not to benchmark. It's a command option that you can add, I forget the exact syntax. Be careful, however, for when the system reaches the point where it should trigger an automatic benchmark test and finds your system restricting this, it immediately sets your Whetstone and Dhrystone performance figures to 0. This gives you 0 claimed credit for any of your results submitted at that point. You may be able to manually run a benchmark to resolve this, but you'll still suffer from a gap there from when you notice the auto benchmark should have run and when you manually execute it.
From my experience, you can
)
From my experience, you can force the core client not to benchmark. It's a command option that you can add, I forget the exact syntax. Be careful, however, for when the system reaches the point where it should trigger an automatic benchmark test and finds your system restricting this, it immediately sets your Whetstone and Dhrystone performance figures to 0. This gives you 0 claimed credit for any of your results submitted at that point. You may be able to manually run a benchmark to resolve this, but you'll still suffer from a gap there from when you notice the auto benchmark should have run and when you manually execute it.
Thanks for your attention, David. Sounds like a risky proposition, perhaps I'll just have to continue to force the benchmarks to run manually before the automatic kicks in, if I can remember to do so.
microcraft
"The arc of history is long, but it bends toward justice" - MLK
What I did was force the
)
What I did was force the benchmark to run at a time when I wouldn't normally use the computer. From then on it seemed to run at about that same time every time it ran automatically. However, I haven't checked it recently so don't know if that still holds true.
Team MacNN - The best Macintosh team ever.
RE: What I did was force
)
Thanks, Shaktai. Well, ahem, what I've been doing is overclocking it more than long-term stable (like 5% more than usual) for the benchmarks, and then resetting it to the usual 2424Mhz for WUs. Can't seem to get any more out of it on aircooling in this 85 degree ambient these days.
microcraft
"The arc of history is long, but it bends toward justice" - MLK
RE: Well, ahem, what I've
)
So you're cheating...
You know what you're doing really isn't doing much except making your local numbers (claimed credit) appear better? Since your granted credit is derived from a quorum of users, your bogus credit numbers are more or less moot. The system was designed for just such hijinks as this. ;)
But if you want to continue going through the overclock only for the benchmarks rigmarole, that's your prerogative. :)
RE: RE: Well, ahem, what
)
Thanks for your input. I prefer to call it "Tweaking", and since I am often the mean of the first 3 results, it brings all 4 of us up about 3 points. Maybe helps in some small way to make up for those who use "optimized" clients;-), and consistantly claim credit in the 110s on a 20000-second Einstein WU, while my claimed credit is usually 71-72 for a 19000-second result, both on Athlons. My little Athlon 2600 Mobile is working it's little heart out on this project, and is consistantly among the most productive single-processors, same as it was when it was running SETI. So, you may can the cheating talk, OK. Let them level the playing field first. Cut the baloney and give us ALL equal credit for each Einstein WU; they're all the same amount of work, anyway, they don't deviate by more than 1-2%.
microcraft
"The arc of history is long, but it bends toward justice" - MLK
RE: Thanks for your input.
)
Michael, Tweaking would be, if you run OC all the time, not only for benchmarking :-)
For original issue: as Shaktai said, you can run benchmarks manually any time
when your computer is not busy with other CPU greedy tasks.
RE: RE: Thanks for your
)
OK, i'll admit to being on the gray side of a fine line. I DO overclock all the time (Athlon 2600 Bartons running WinXP2 are incapable of 18800-19500 sec WUs at default speed, believe me), and OCing it further for benchmarks IS defying an unfair system. I'm not boohooing here, just railing against what I consider to be an arbitrary credit-allotting. I repeat (see previous), since virtually all WUs so far are the same size/require the same amount of crunching, why not level the field by granting equal credit for equal crunching, for each validated WU, for each user, regardless of the platform/OS, at least until such time as different size WUs are distributed? Perhaps this should be the topic of another thread, on another message board? I don't know, I'm new to communicating with the group. Thank you for your valid opinion, as well as that of Heffed.
microcraft
"The arc of history is long, but it bends toward justice" - MLK
RE: Maybe helps in some
)
I don't understand... How are you making up for the optimized clients? You are doing much the same thing, but on a lesser scale. You are requesting higher credit through overclocking for the benchmarks, thus giving inflated benchmark results. The optimized clients are requesting higher credits by 'optimizing' the actual benchmarking code, thus giving inflated benchmark results. Both scenarios being false representations of your actual processing conditions.
In your situation, the assumption is that you will also have the same overclock while processing a WU. In the optimized client situation, it is assuming you are also using an optimized science application to give the inflated benchmarks relevance. Both assumptions are incorrect in this instance.
If your inflated numbers are that important to you, why not bypass the overclock dance and just install an optimized client? Higher benchmarks, no muss, no fuss. :)
And it you're really feeling your oats, run an optimized client, then overclock for the benchmarks using it. WOOHOO! ;)
RE: RE: Maybe helps in
)
Heffed, I posted asking for information, and you are unable to give me any apropos to my question. Your point is valid, your righteous indignation superfluous, your sarcasm noted. I see you care not to address the secondary issue of the tilted playing field, the benchmarking imbalance, an artificial foundation for credit, without which my artifice would not be motivated. Crunch well, my brother, and Thank you.
microcraft
"The arc of history is long, but it bends toward justice" - MLK