climateprediction.net home page
Time to complete 1 wu

Time to complete 1 wu

Questions and Answers : Windows : Time to complete 1 wu
Message board moderation

To post messages, you must log in.

AuthorMessage
old_user538

Send message
Joined: 7 Aug 04
Posts: 2
Credit: 5,161
RAC: 0
Message 373 - Posted: 8 Aug 2004, 1:56:21 UTC

I have just started my first wu and have noticed it is going to take over 730 hrs to complete, is ths normal?

I have the following :
AMD ~1916 Mhz
506.26 MB

ID: 373 · Report as offensive     Reply Quote
Jord
Avatar

Send message
Joined: 5 Aug 04
Posts: 250
Credit: 93,274
RAC: 0
Message 374 - Posted: 8 Aug 2004, 2:03:50 UTC

Expect at some time even more. Haven't you read your email?

On a 2.6GHz machine it can take 4 weeks or more to crunch a unit. CPDN isn't about fast results, fast units. If you are here for fast credits (just asking ;)), then you are at the wrong address.
--------------------
Jordâ„¢
ID: 374 · Report as offensive     Reply Quote
belgix

Send message
Joined: 5 Aug 04
Posts: 85
Credit: 2,924,043
RAC: 0
Message 375 - Posted: 8 Aug 2004, 2:06:13 UTC

You're right. CPDN WU are very long to process. From the e-mail ...

<i> ... which typically takes a month on a fairly good computer (i.e. G5 Mac or P4/2.6GHz PC) running "24/7." ...</i>

This project is a paradise for dial-up users!

ID: 375 · Report as offensive     Reply Quote
old_user538

Send message
Joined: 7 Aug 04
Posts: 2
Credit: 5,161
RAC: 0
Message 377 - Posted: 8 Aug 2004, 2:10:44 UTC - in response to Message 374.  

&gt; Expect at some time even more. Haven't you read your email?
&gt;
&gt; On a 2.6GHz machine it can take 4 weeks or more to crunch a unit. CPDN isn't
&gt; about fast results, fast units. If you are here for fast credits (just asking
&gt; ;)), then you are at the wrong address.
&gt; --------------------
&gt; Jordâ„¢
&gt;
Sorry about that. I did read my email but must of missed it. I don't mind how long it takes I was just use to seti@home where I could complete up to 8 a day.
ID: 377 · Report as offensive     Reply Quote
Profile astroWX
Volunteer moderator

Send message
Joined: 5 Aug 04
Posts: 1496
Credit: 95,522,203
RAC: 0
Message 378 - Posted: 8 Aug 2004, 2:12:55 UTC

~30 days -- that seems on ~ the right part of the Pitch.

It is big model and does a lot of work:
Three Phases of 259248 Time Steps each (15 years each Phase, every 30 minutes of the Model) for 19 levels of atmosphere, on a grid of 96 Longitude by 73 Latitude -- and interaction among the grid cells. Boundary layer (earth surface, including slab ocean -- upwelling, etc, not considered -- yet). Plenty of parameters -- pressure, heat flux, temperature, moisture.... Your machine will be well used, to a good end.

HTH.
________________________________________________
We have met the enemy and he is us -- Pogo
ID: 378 · Report as offensive     Reply Quote
Profile astroWX
Volunteer moderator

Send message
Joined: 5 Aug 04
Posts: 1496
Credit: 95,522,203
RAC: 0
Message 382 - Posted: 8 Aug 2004, 4:27:26 UTC - in response to Message 374.  

&gt; Expect at some time even more. Haven't you read your email?
&gt;
&gt; On a 2.6GHz machine it can take 4 weeks or more to crunch a unit. CPDN isn't
&gt; about fast results, fast units. If you are here for fast credits (just asking
&gt; ;)), then you are at the wrong address.
&gt; --------------------
&gt; Jordâ„¢

Greetings, Jord,

You are right-on with regard to fast results. But -- is it not true that, when all is done, a massive pile of Cobbles will accrue to those who persevere?

I'm not here chasing rewards but appreciate that some folks are. I think we should encourage them to participate, too, because a month of CPDN crunching should reap big rewards.

And, as belgix noted, this project should reap large rewards for dial-up users -- no need to be connected multi- times per day.

The BIG reward is, IMO, the satisfaction of contributing to the greater good. However, if recognition brings in participants -- I'll drink to that.

(OT I remember reading your posts from boinc/beta/AP. Good to see you here. /OT)
________________________________________________
We have met the enemy and he is us -- Pogo
ID: 382 · Report as offensive     Reply Quote
old_user355

Send message
Joined: 7 Aug 04
Posts: 187
Credit: 44,163
RAC: 0
Message 388 - Posted: 8 Aug 2004, 7:56:22 UTC - in response to Message 382.  

&gt; You are right-on with regard to fast results. But -- is it not true that,
&gt; when all is done, a massive pile of Cobbles will accrue to those who
&gt; persevere?

I would assume so, after all, credit is granted by the amount of CPU time spent on the WU. A month of processing should amount to a fairly hefty amount of credit.

I don't know if the project will ever appeal to the credit chasers though, because of the long payoff. They generally aren't very patient and want to see results right away.
ID: 388 · Report as offensive     Reply Quote
old_user1
Avatar

Send message
Joined: 5 Aug 04
Posts: 907
Credit: 299,864
RAC: 0
Message 395 - Posted: 8 Aug 2004, 9:39:29 UTC
Last modified: 8 Aug 2004, 10:02:22 UTC

I haven't actually worked out "credits" yet, I was thinking trickles should be "incremental" (i.e. partial, unverified) credit until the end of a run where it's "validated"?

PS for estimated completion times for a run you can peek at my new cpu-timing page (preliminary, based on 1000 trickles):

<A HREF="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/cpu.html">CPDN/BOINC Stats/Timings By CPU/OS/etc</A>
ID: 395 · Report as offensive     Reply Quote
old_user355

Send message
Joined: 7 Aug 04
Posts: 187
Credit: 44,163
RAC: 0
Message 406 - Posted: 8 Aug 2004, 10:33:36 UTC - in response to Message 395.  

&gt; I haven't actually worked out "credits" yet, I was thinking trickles should be
&gt; "incremental" (i.e. partial, unverified) credit until the end of a run where
&gt; it's "validated"?

That sounds like a good concept. The more similar the project can appear to other projects running under the BOINC framework, I think will prove to be good for the project.

&gt; PS for estimated completion times for a run you can peek at my new cpu-timing
&gt; page (preliminary, based on 1000 trickles):

Thanks for the info! It's interesting to look at. Looks like there's still some funky benchmarking going on with this version of the CC.
ID: 406 · Report as offensive     Reply Quote
Profile old_user156
Avatar

Send message
Joined: 5 Aug 04
Posts: 186
Credit: 1,612,182
RAC: 0
Message 446 - Posted: 8 Aug 2004, 17:05:13 UTC
Last modified: 8 Aug 2004, 17:23:00 UTC

Summat does not compute about those trickle calcs Carl - You list "Avg_TS_per_sec" &amp; "Est_WU_Comp_Days" and they don't appear to match up. ie. low "Avg_TS_per_sec" doesn't match a low "Est_WU_Comp_Days" in reality does it. :? Mebbe that oughta read "Avg_secs_per_TS" instead....?

The top few look ridiculously fast anyways - like they're running fast processing iceballs..? First reasonable figure when gauged against benchmark figures is the Athlon FX53. Although, do P4s run CP-boinc a lot faster 'cause of using a different compiler..?

Wish I knew how to get rid of "Unknown CPU Type Pentium" :lol: Should be "2.6GHz AMD Athlon XP-M" or something like that. 1.949 s/ts and 17.542 days to completion is about right though... ;-)
ID: 446 · Report as offensive     Reply Quote
Profile geophi
Volunteer moderator

Send message
Joined: 7 Aug 04
Posts: 2167
Credit: 64,403,322
RAC: 5,085
Message 449 - Posted: 8 Aug 2004, 17:24:42 UTC - in response to Message 446.  

&gt; The top few look ridiculously fast anyways - like they're running fast
&gt; processing iceballs..? First reasonable figure when gauged against benchmark
&gt; figures is the Athlon FX53. Although, do P4s run CP-boinc a lot faster 'cause
&gt; of using a different compiler..?
&gt;
May be part of it, but more likely it's the way it's figuring out times when running two models at once on dual or hyperthreaded CPUs. Certainly the time to compute one workunit doesn't quite make sense though, unless one work unit is 45 years total instead of one model run.
ID: 449 · Report as offensive     Reply Quote
Profile Thyme Lawn
Volunteer moderator

Send message
Joined: 5 Aug 04
Posts: 1283
Credit: 15,824,334
RAC: 0
Message 506 - Posted: 9 Aug 2004, 7:59:16 UTC - in response to Message 449.  

&gt; &gt; The top few look ridiculously fast anyways - like they're running fast
&gt; &gt; processing iceballs..? First reasonable figure when gauged against
&gt; benchmark
&gt; &gt; figures is the Athlon FX53. Although, do P4s run CP-boinc a lot faster
&gt; 'cause
&gt; &gt; of using a different compiler..?
&gt; &gt;
&gt; May be part of it, but more likely it's the way it's figuring out times when
&gt; running two models at once on dual or hyperthreaded CPUs. Certainly the time
&gt; to compute one workunit doesn't quite make sense though, unless one work unit
&gt; is 45 years total instead of one model run.

Timings are spot on for my HT P4s (they stand out because I'm running Windows 2000 and there are only three system running 2 models on that combination).


<a href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/team_display.php?teamid=3"><img src="http://www.teampicard.net/templates/fisubice/images/phpbb2_logo.jpg"></a>
ID: 506 · Report as offensive     Reply Quote

Questions and Answers : Windows : Time to complete 1 wu

©2024 climateprediction.net