climateprediction.net home page
[topic]HadCM3L

[topic]HadCM3L

Message boards : Number crunching : [topic]HadCM3L
Message board moderation

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
old_user92639

Send message
Joined: 13 Aug 05
Posts: 54
Credit: 117,227
RAC: 0
Message 41028 - Posted: 13 Nov 2010, 23:55:54 UTC

hello,

BOINC - Tâches
=> 0H - 0% => in 1446 hours (2 months)
=> 6H - 0.37% => in 1443 hours

http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=12018154

2.90 sec/TS

TimeStep 7575 / 2073960

:)
ID: 41028 · Report as offensive     Reply Quote
Les Bayliss
Volunteer moderator

Send message
Joined: 5 Sep 04
Posts: 7629
Credit: 24,240,330
RAC: 0
Message 41029 - Posted: 14 Nov 2010, 0:23:20 UTC

hadcm3igeo_w15e_2000_80_06746675_1

825 hours
2.06 s/TS


Backups: Here
ID: 41029 · Report as offensive     Reply Quote
Helmer Bryd

Send message
Joined: 16 Aug 04
Posts: 147
Credit: 7,748,561
RAC: 8,366
Message 41030 - Posted: 14 Nov 2010, 9:25:19 UTC
Last modified: 14 Nov 2010, 10:09:08 UTC

Got 4 of them and all are running well after about 20h :)

No graphic showing though after first trickle

1.80-1.89 s/TS on Linux, much faster than the last batch on the Beta site
ID: 41030 · Report as offensive     Reply Quote
Profile JIM

Send message
Joined: 31 Dec 07
Posts: 1152
Credit: 22,053,321
RAC: 4,417
Message 41033 - Posted: 15 Nov 2010, 0:39:39 UTC

I just downloaded 2 of the new 80 year CM models to my 2.2 GHz machine with 4 GB of RAM and was astonished to see the “to completion” time at 5069 hours. I am hoping that this is a major over estimation. What gives? Why so high an estimate?

I remember running the old 160 year CM’s on a 1.5 GHz machine that had only 1 GB of RAM and it took only about 3500 hours to complete them. Good thing that I didn’t download them to my slower machine as I would like to finish them before the sun burns out. ;-)

ID: 41033 · Report as offensive     Reply Quote
Les Bayliss
Volunteer moderator

Send message
Joined: 5 Sep 04
Posts: 7629
Credit: 24,240,330
RAC: 0
Message 41035 - Posted: 15 Nov 2010, 7:32:04 UTC - in response to Message 41033.  

There's an older thread here that talks about the length of these models.

As to why you have an estimate of over 5,000 hours, that may depend on what other work BOINC has in it's To Do list, and also this project's resource share.


Backups: Here
ID: 41035 · Report as offensive     Reply Quote
old_user92639

Send message
Joined: 13 Aug 05
Posts: 54
Credit: 117,227
RAC: 0
Message 41039 - Posted: 15 Nov 2010, 10:14:51 UTC

Trickle 1 = 326.59 Credit

326.59 x 80 = 26127.2 credit

:)
ID: 41039 · Report as offensive     Reply Quote
Profile JIM

Send message
Joined: 31 Dec 07
Posts: 1152
Credit: 22,053,321
RAC: 4,417
Message 41045 - Posted: 15 Nov 2010, 15:02:56 UTC - in response to Message 41039.  

Trickle 1 = 326.59 Credit

326.59 x 80 = 26127.2 credit


26.127 credits should be about tight of this type of model. If I remember correctly, the old 160 year CM’s produced about 44,000 credits for a finished model.

[/quote]
ID: 41045 · Report as offensive     Reply Quote
Helmer Bryd

Send message
Joined: 16 Aug 04
Posts: 147
Credit: 7,748,561
RAC: 8,366
Message 41094 - Posted: 19 Nov 2010, 15:07:18 UTC

My four models haves past the first decade and uploaded the 15Mb 1.zip

I thought they would show some graph like in the former hadcm variants I ran, no?
ID: 41094 · 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 41096 - Posted: 19 Nov 2010, 16:07:51 UTC - in response to Message 41030.  

1.80-1.89 s/TS on Linux, much faster than the last batch on the Beta site

Thanks for that note cwhyl. The applications build date suggests it was compiled back in 2008, and I would swear that that version was slow on Linux. Perhaps a rebuild that isn't advertised on the applications page?
ID: 41096 · Report as offensive     Reply Quote
Profile Milo Thurston
Volunteer moderator
Volunteer developer

Send message
Joined: 2 Mar 06
Posts: 253
Credit: 363,646
RAC: 0
Message 41097 - Posted: 19 Nov 2010, 16:27:34 UTC - in response to Message 41096.  

I didn't rebuild the application before generating the most recent workunits, as it's for the same sort of experiment that was previously done with that model and we're a bit short-staffed at the moment.
ID: 41097 · Report as offensive     Reply Quote
Helmer Bryd

Send message
Joined: 16 Aug 04
Posts: 147
Credit: 7,748,561
RAC: 8,366
Message 41098 - Posted: 19 Nov 2010, 17:53:49 UTC - in response to Message 41096.  
Last modified: 19 Nov 2010, 18:12:45 UTC

1.80-1.89 s/TS on Linux, much faster than the last batch on the Beta site

Thanks for that note cwhyl. The applications build date suggests it was compiled back in 2008, and I would swear that that version was slow on Linux. Perhaps a rebuild that isn't advertised on the applications page?

Ah, I compared to the latest highly optimised 6.05/Beta batch that was 2.7!! times slower than Jims' Q9550/2.8 GHz Win7

My boxen is Q6600 at 3.2-3.4 GHz

I don't remember the speed of the latest hadcm I ran here some years ago and the record seem to have been wiped out :-(

Here I've seen 1.11 and 1.50 s/Ts on newer windows i-series machines
ID: 41098 · Report as offensive     Reply Quote
Les Bayliss
Volunteer moderator

Send message
Joined: 5 Sep 04
Posts: 7629
Credit: 24,240,330
RAC: 0
Message 41099 - Posted: 19 Nov 2010, 18:07:46 UTC - in response to Message 41098.  

... the record seem to have been wiped out ...

Not 'wiped out', archived.

This is mentioned on people's Account pages, not far from the top.

ID: 41099 · Report as offensive     Reply Quote
Helmer Bryd

Send message
Joined: 16 Aug 04
Posts: 147
Credit: 7,748,561
RAC: 8,366
Message 41102 - Posted: 19 Nov 2010, 19:20:37 UTC - in response to Message 41099.  

... the record seem to have been wiped out ...

Not 'wiped out', archived.

This is mentioned on people's Account pages, not far from the top.


oh,I understand that Les..
ID: 41102 · Report as offensive     Reply Quote
Darmok

Send message
Joined: 29 Dec 09
Posts: 34
Credit: 18,395,130
RAC: 0
Message 41156 - Posted: 25 Nov 2010, 12:30:11 UTC - in response to Message 41094.  

"26.127 credits should be about tight of this type of model. If I remember correctly, the old 160 year CM’s produced about 44,000 credits for a finished model."

44,000 perhaps should be the amount, as I am noticing so far a drop of up to 20-40% from FAMOUS. The models, every 24 timestep hours, revert back and redo the same timestep albeit at a much faster speed. They will likely go over 1000 hours to complete on my M/C.
ID: 41156 · Report as offensive     Reply Quote
Les Bayliss
Volunteer moderator

Send message
Joined: 5 Sep 04
Posts: 7629
Credit: 24,240,330
RAC: 0
Message 41157 - Posted: 25 Nov 2010, 13:03:50 UTC - in response to Message 41156.  

The models, every 24 timestep hours, revert back and redo the same timestep albeit at a much faster speed.

Not quite.
For the Coupled Ocean models, there are two different phases in the calcs: the "slow" 24 hours is the Atmospheric phase. The "fast" 24 hours is the Ocean phase. It's faster because the conditions deep in the ocean change very slowly.

The FAMOUS models use a different strategy. This can be read about here on the relevant page of the Experiments section near the front of the project's web site.


Backups: Here
ID: 41157 · Report as offensive     Reply Quote
Helmer Bryd

Send message
Joined: 16 Aug 04
Posts: 147
Credit: 7,748,561
RAC: 8,366
Message 41160 - Posted: 25 Nov 2010, 19:34:13 UTC
Last modified: 25 Nov 2010, 20:09:25 UTC

Why the heck is there no graphic showing on these models?
ID: 41160 · Report as offensive     Reply Quote
old_user92639

Send message
Joined: 13 Aug 05
Posts: 54
Credit: 117,227
RAC: 0
Message 41162 - Posted: 25 Nov 2010, 19:48:27 UTC

w1aj-2000-80_1.zip 15.70 MB

15.70 MB x 8 = 125.6 MB

:)
ID: 41162 · Report as offensive     Reply Quote
Darmok

Send message
Joined: 29 Dec 09
Posts: 34
Credit: 18,395,130
RAC: 0
Message 41163 - Posted: 26 Nov 2010, 0:18:45 UTC - in response to Message 41157.  

Thanks Les. It is an interesting behavior for the model though the Ocean Phase is as fast as the dip in my RAC. Oh well.
:)

ID: 41163 · Report as offensive     Reply Quote
Les Bayliss
Volunteer moderator

Send message
Joined: 5 Sep 04
Posts: 7629
Credit: 24,240,330
RAC: 0
Message 41165 - Posted: 26 Nov 2010, 19:13:13 UTC - in response to Message 41160.  

Why the heck is there no graphic showing on these models?

They'll be along later. There hasn't been time to write the graphs program yet.

Backups: Here
ID: 41165 · Report as offensive     Reply Quote
[B^S] mavau

Send message
Joined: 30 Aug 04
Posts: 142
Credit: 9,936,132
RAC: 0
Message 41166 - Posted: 26 Nov 2010, 19:33:42 UTC

I'm going to second Darmok about the dip in RAC.
Is any adjustment possible?
It's an old story, but if a 1000+ hours model is useful, is there a way to encourage crunchers? ;-)
Jim, did you mean "right"? Although "tight" may be to the point :-)

Still, they're going along nicely.

Forum search Site search
ID: 41166 · Report as offensive     Reply Quote
1 · 2 · Next

Message boards : Number crunching : [topic]HadCM3L

©2024 climateprediction.net