Posts by old_user202885

1) Message boards : Number crunching : Core 2 Duo E6600 speed (Message 26130) Posted 17 Jan 2007 by old_user202885 Post: we\'re beta testing a new version of the hadcm3 model, using the latest Intel Fortran compiler (9.1) and some optimizations which seem to work. People are getting \"sub-1-second\" per timesteps on Core2 Duos (these CPUs seem to be the best; I guess Intel has made their compiler work best on their latest & greatest chip! :-)
2) Message boards : Cafe CPDN : can a task really take this long? (Message 25727) Posted 24 Dec 2006 by old_user202885 Post: your computer seems to be running a bit slow (8.9 seconds per timestep) considering it\'s a P4/3.2GHz (I have a P4/3.2GHz at home and it gets 2.1 seconds per timestep). I\'m not sure quite why it\'s so slow, but if it\'s running at 8.9 secs per timestep it could indeed take over a year to finish a complete 160 model-year run!
3) Message boards : Number crunching : 20,000,000,000 Credits across ALL BOINC Projects (Message 25628) Posted 16 Dec 2006 by old_user202885 Post: well I\'ve been accused of CPDN being \"inflated\" credits, so maybe it all balances out! ;-) having credits that compare reliably across projects is still a very big concern of BOINC developers & users all the way to the top (David Anderson). The \"benchmarking credits\" can give odd scores sometimes; a guy in our department has the top MacBook Pro (Intel) and it\'s getting less than a second per timestep! But his benchmarks are lower than my home PC which gets 2.15 secs/ts. So I think the CPDN method of \"credit per trickle\" probably gives a fairer share of credit. Although there is probably something like a +/- 10% variation possible between workunit (i.e. parameters used in different experiments may \"turn on\" a routine that will take more CPU time). But anyway, thanks to Zain for posting about the credits benchmark. To me it\'s like celebrating the millenium, i.e. even though people from different background may not agree that it was \"year 2000\" it was still fun to get together around the world and watch the fireworks! :-)
4) Questions and Answers : Macintosh : * Sticky - PowerPC Mac users (Message 25401) Posted 1 Dec 2006 by old_user202885 Post: Tolu has done amazing things with the Intel Mac and will hopefully be making an announcement soon for those interested in a beta tests. Early results look promising -- 1.72 seconds/timestep on the coupled model running on our new ibook (my home PC, a pretty fast P4/3.2GHz, gets 2.15s/ts).
5) Message boards : Number crunching : Preferences not transferring from here to my m/c (Message 25018) Posted 7 Nov 2006 by old_user202885 Post: I tried it out on my work PC (I run SETI@home as well as CPDN right now), and although the red \"update preferences\" doesn\'t show up in messages, the global_prefs.xml file has been updated.
6) Questions and Answers : Windows : New: Throttling CPU Usage, Simple BOINC GUI (Message 24986) Posted 4 Nov 2006 by old_user202885 Post: I have recently done some updates so that the newer settings in your \"Global Preferences\" web page, to limit CPU Usage, should work on the models. For example, the usual default of using your CPU as much as it\'s idle can now be \"throttled\" back to a different value. So if you find your CPU or computer is too hot, you can change it from 100% to perhaps 75%. Of course the model will take longer to get done, but you may prefer keeping your CPU \"cooler.\" You will need a newer BOINC client such as the new 5.4.11 to use this functionality. you can download the latest boinc at http://boinc.berkeley.edu/download.php Also, I have added some functionality for those of you using the simple BOINC GUI (currently in beta testing) -- it should get the appropriate CPDN icons & screensaver snapshots downloaded (see image below).
7) Message boards : Number crunching : I know, there\'s something wrong with my clientstats.xml, should I reset the WU? (Message 24927) Posted 30 Oct 2006 by old_user202885 Post: this is an odd problem, can you email me your client_state.xml file which is in your BOINC installation directory? Send it to carlc at atm dot ox dot ac dot uk
8) Message boards : Number crunching : Wrong default encoding for CPDN/BOINC website pages. (Message 24858) Posted 23 Oct 2006 by old_user202885 Post: can you please post the exact link because everything looks fine for me here: http://climateprediction.net/versions/RU/index.php
9) Message boards : Number crunching : Server Upgrade Completed (Message 24789) Posted 19 Oct 2006 by old_user202885 Post: Sorry for the server downtime today, I ended up redoing the database server to speed things up, so hopefully things are up now and back to normal if not faster!
10) Message boards : Number crunching : Run Length - a suggestion (Message 24785) Posted 19 Oct 2006 by old_user202885 Post: Thanks for your comments, but I\'m afraid you greatly underestimate the work involved in doing what you described. There\'s nothing regarding climate models that can be done in \"a dozen lines of code!\" :-) What is really needed, and would take weeks of work full-time (as if we have it); is a system for recovering and automatically processing uploaded start dumps, really checking that they are OK for resending. That would make full use of our 40-year workunits (which we\'re capable of now on the client side; but haven\'t put into practice other than beta tests).
11) Message boards : Number crunching : Server Upgrade Completed (Message 24784) Posted 19 Oct 2006 by old_user202885 Post: Okay, so it\'s normal that I have to wait. No problem then, thanks for explaining. it could be a problem in that you seem to have received a workunit when there was a BOINC server bug that caused you to download hundreds of \"blank\" workunits (which you can see on your results page). can you type in what workunit name you are running, i.e. hadcm3oh_0j50_00002342 or similar.
12) Message boards : Number crunching : Server Upgrade Completed (Message 24773) Posted 18 Oct 2006 by old_user202885 Post: I hope! If anybody finds anything wrong with the upgrade today, please post in this thread, thanks!