ClimatePrediction.net on the BBC...

NickK · 14 Feb 2006 at 12:43

http://news.bbc.co.uk/1/hi/sci/tech/4702636.stm

Including a new dynamic sea model... I wonder if that's going to be slower than the sulphur model

NickK · 14 Feb 2006 at 15:22

Enfield said:
I just tried this and it will takes like years to complete one thing (unit?). Trying rosetta instead.

There is a deadline of a year on the work units.

My lappy is currently 18.07% through a sulphur model after 304:13:19 CPU time... only 1657:23:20 to go....

NickK · 15 Feb 2006 at 08:19

BillytheImpaler said:
What's the reason for having such long WUs? Is there not a generational system whereby a simulation is begun by one user, turned in, and finished by another or others. A system like that would make it similar to FAH's work distribution. How large in terms of bytes are the WUs? Is the long processing time meant to compensate for large download size?

The WU isn't a section of data to be analysed by the project code in the same way that SETI, folding etc is. Instead a CP.net WU is a complete full scale meterological model run with specific parameters. It's related (ie so close it's written in FORTRAN) to the met office's weather model code that runs on their super computers.

Hence the processing power required per WU for CP.net is far higher than the other data analysis projects. This is why the disc space used to record the model state for each commit is larger than a simple analysis WU.

My current sulphur WU is up to 1.5Gb of disc used which is expected as the old hd3sm slab model used ~750Mb IIRC.

Distributing a run over computers isn't feasable due to the maths needing to access data from all over the model (even with a cellular model).

NickK · 15 Feb 2006 at 08:33

Fillado said:
Just installed this

Looking reasonably good, and AFAIK it's the first BOINC project to actually use the graphics card to display the graphics (if you're that way inclined).

2150hrs to go

Using the graphic or the graphical screensaver CP.net will kill the crunching performance as there's quite a bit of maths and processing to visualise the model.

NickK · 15 Feb 2006 at 16:02

Whitestar said:
Do you take part using boinc or do you have to use the download from the bbc website?

I use the BOINC one as there's an OcUK team too

NickK · 16 Feb 2006 at 17:43

Yup it appears hat the BBC one is a seperate model.

NickK · 17 Feb 2006 at 08:24

Good stuff

208 is the team number for OcUK for BOINC so In theory, any Cimate Change or Climate Prediction effort gets added to the team score.

Climate Prediction : http://www.boincstats.com/stats/project_graph.php?pr=cpdn
Climate Change: http://www.boincstats.com/stats/project_graph.php?pr=cce

Just put in "OcUK" into the search user/team and you'll see all of the OcUK effort over all of the BOINC projects.

NickK · 17 Feb 2006 at 10:09

Shhh ....

my plans are being uncovered :eek:

NickK · 17 Feb 2006 at 13:13

208 if your doing the standard Climate Prediction with ClimatePrediction.net
280 if your doing the BBC Climbate Change with ClimatePrediction.net

You will neet to attach to the correct project within BOINC.

NickK · 21 Feb 2006 at 13:20

Berkley is only the distributed system component. The actual software and the model it distributes that does the hard work is actually Oxford uni.