The search thing returns nothing so I'm guessing this question hasn't been asked before - apologies if it has.

Can automatic benchmarking be disabled? What's the option?


TIA

--
john

No.

> No.
>

Well that was short :D

I'll have a go at the developers :P


--
john

why do you want it disabled? it only runs once. this is determine the size of wu's your computer gets so what the deal?
ttyl
Jeff (Nightowl)
All your answers in one spot:
http://setiweb.ssl.berkeley.edu/transition.php
=====
If you dont like Boinc, then go back to classic seti.

> why do you want it disabled? it only runs once. this is determine the size
> of wu's your computer gets so what the deal?

It doesn't run once - it runs randomly. Sometimes BOINC starts and the benchmarks run, other times they don't. At boot-up the CPU is busy so the benchmarks are always wrong. Post boot-up benchmarks are about a 1/3 better. So if I didn't re-run the benchmarks my CPU would be assigned WUs that are less than it's capable of processing.

--
john

Benchmarks are set to run every five days.

They are done automatically because people have a tendency not to keep their benchmarks up to date, skewing the results of others.

The benchmarking during boot up it known, and I'm sure they're working on a fix.

<a> [/url]

> They are done automatically because people have a tendency not to keep their
> benchmarks up to date, skewing the results of others.
>

And that's worse than having benchmarks that are totally inaccurate because they were taken when the CPU was busy.

> The benchmarking during boot up it known, and I'm sure they're working on a
> fix.
>

The list of things to fix is growing fast. God I love BOINC.


--
john

> Do any of the current BOINC projects actually issue different sized WU based
> on anyone's benchmarks?

I don't know, but my main concern is, that the credits a client will claim are dependent on the benchmark => as long as the benchmarks aren't accurate, the credit system will not be fully fair.

Raphael Schweizer
Schweizer Informatik

> > Do any of the current BOINC projects actually issue different sized WU
> based
> > on anyone's benchmarks?
>
> I don't know, but my main concern is, that the credits a client will claim are
> dependent on the benchmark => as long as the benchmarks aren't accurate,
> the credit system will not be fully fair.
>

Credit claims are based on the benchmarks and the CPU time. The developers did not realize just how bacly the benchmarks would be affected by doing them during startup. I have suggested that doing benchmarks as a "real time" process would solve the problem, admitedly at the expense of any other program that was running.

Yes, if you have a slow enough machine, certain classes of WUs will not be downloaded to your machine. For example when Predictor sets its deadline to 24 hours, that effectively cuts 4 of my machines off from mfold WUs as the estimates for these machines for an mfold WU is greater than 24 hours. Another project (not yet announced, and I will not until they are ready) will only download to a machine if it has 600Mb free hard disk allocated as usable by BOINC.