??benchmark doesn't work??

Message boards : Number crunching : ??benchmark doesn't work??
Message board moderation

To post messages, you must log in.

AuthorMessage
bjacke
Volunteer tester
Avatar

Send message
Joined: 14 Apr 02
Posts: 346
Credit: 13,761
RAC: 0
Germany
Message 31418 - Posted: 30 Sep 2004, 12:38:12 UTC

Hi there.
I've run a benchmark yesterday. In conclusion a S@h wu will last 3h45min. But in real it is finished in 2h30min.
Any ideas?
<img border="0">
The whole is more then the sum of its particles.
Aristoteles
Best wishes from Berlin(52°35'N,13°23'O), Basti
S@h Berkeley's Staff Friends Club © member
ID: 31418 · Report as offensive
Profile Mchl
Volunteer tester
Avatar

Send message
Joined: 17 Feb 00
Posts: 35
Credit: 231,170
RAC: 0
Poland
Message 31421 - Posted: 30 Sep 2004, 12:43:36 UTC - in response to Message 31418.  

Benchmarks seem to underestimate my pc's computing power sa well. I run S@H,LHC and CPDN and I finish wu's for S@H and LHC in less time than estimated. Still computing my first CPDN unit ;-)
This probably mean my pc claim less pending credit...but who cares...Glad I get that credit at all ;-)
BOINC@Poland: Polish crossproject team.

http://mchl.republika.pl/boincatpoland/

- Join -
ID: 31421 · Report as offensive
Profile Paul D. Buck
Volunteer tester

Send message
Joined: 19 Jul 00
Posts: 3898
Credit: 1,158,042
RAC: 0
United States
Message 31445 - Posted: 30 Sep 2004, 13:46:37 UTC - in response to Message 31418.  

> Hi there.
> I've run a benchmark yesterday. In conclusion a S@h wu will last 3h45min. But
> in real it is finished in 2h30min.
> Any ideas?

The estimate is just that ... a guess ... and the reason that it seems high is that they cannot predict your load.

In my case, most of the machines are just doing BOINC, and just about nothing else. What does that mean, it means that the "slop" is always on the high side. If they under predicted people will be mad also.

:)

At the moment, we will likely not see much change in these guesses. We have to finish with the fundamentals and then the developers can start to look at how to make the estimates more accurate. The numbers are there in the Benchmark, and the amount of time on BOINC ... but I don't think that they need to be that much more accurate ..
<p>
For BOINC Documentaion: Click Me!


ID: 31445 · Report as offensive
Profile MP3
Volunteer tester

Send message
Joined: 1 Sep 00
Posts: 20
Credit: 3,752,328
RAC: 0
Malaysia
Message 31488 - Posted: 30 Sep 2004, 16:46:14 UTC

Here something for you to think about
the time you work on the wu can actually slow down if your pc is proccesing something else if you set boinc to work non-stop

example, a wu said it taken 2 hour to complete but in real-life it taken 2 hours and 15 minutes.

thats what happening to me. try it


ID: 31488 · Report as offensive
Profile Mchl
Volunteer tester
Avatar

Send message
Joined: 17 Feb 00
Posts: 35
Credit: 231,170
RAC: 0
Poland
Message 31489 - Posted: 30 Sep 2004, 17:03:41 UTC - in response to Message 31488.  

> Here something for you to think about
> the time you work on the wu can actually slow down if your pc is proccesing
> something else if you set boinc to work non-stop
>
> example, a wu said it taken 2 hour to complete but in real-life it taken 2
> hours and 15 minutes.
>
> thats what happening to me. try it
>

My PC is being used in an office. Working on huge spreadsheets and so huge databases. Yet the time units are processed is still lower than estimated. And I don't complain ;-)
BOINC@Poland: Polish crossproject team.

http://mchl.republika.pl/boincatpoland/

- Join -
ID: 31489 · Report as offensive
bjacke
Volunteer tester
Avatar

Send message
Joined: 14 Apr 02
Posts: 346
Credit: 13,761
RAC: 0
Germany
Message 31490 - Posted: 30 Sep 2004, 17:05:56 UTC - in response to Message 31488.  

> Here something for you to think about
> the time you work on the wu can actually slow down if your pc is proccesing
> something else if you set boinc to work non-stop
>
> example, a wu said it taken 2 hour to complete but in real-life it taken 2
> hours and 15 minutes.
>
> thats what happening to me. try it
>
Ya thats right, but infact, the time shown on BOINC displais the time based on the benchmark. If I run the benchmark when BOINC is the only application realy running it shws me the preformence it will need to crunch when the system isn't realy in use. So if I use the sytem strongly with other it has to be higher time, like you said.



The whole is more then the sum of its particles.
Aristoteles
Best wishes from Berlin(52°35'N,13°23'O), Basti
S@h Berkeley's Staff Friends Club © member
ID: 31490 · Report as offensive
Profile Benher
Volunteer developer
Volunteer tester

Send message
Joined: 25 Jul 99
Posts: 517
Credit: 465,152
RAC: 0
United States
Message 31569 - Posted: 30 Sep 2004, 21:44:23 UTC

If the programmers fix the estimate /\ benchmark, they will still probably "over" estimate the time by a little bit...this will make end users feel better about their systems capabilities.
(Hah! My system did it in less...)

;)
ID: 31569 · Report as offensive
Profile Marco Berkhout

Send message
Joined: 3 Jul 04
Posts: 14
Credit: 4,059,793
RAC: 0
Netherlands
Message 31579 - Posted: 30 Sep 2004, 21:54:48 UTC - in response to Message 31488.  

> Here something for you to think about
> the time you work on the wu can actually slow down if your pc is proccesing
> something else if you set boinc to work non-stop
>
> example, a wu said it taken 2 hour to complete but in real-life it taken 2
> hours and 15 minutes.
>
> thats what happening to me. try it
>
>
>
>

The time displayed is the CPU time. The time the CPU was active on the WU.
In realtime the CPU is also doing other tasks, but still does some crunching. Only the time it is working on the WU is displayed/measured.

So if you are playing a game while crunching your pc it will take a whole lot longer to crunch a WU but still displays roughly the same CPU time because the CPU has spent the same amount of time on that WU. The rest was spent on the game.


ID: 31579 · Report as offensive

Message boards : Number crunching : ??benchmark doesn't work??


 
©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.