Climate models use a huge supercomputer program (1 million plus lines of code), to run, and tests showed little improvement using 64 bit systems. They were also unstable, making the results useless.

Come back next year. But bring a VERY powerfull computer with you. Some models being planned make the present requirements look like toys.

How is the situation today? Why CPDN don\'t have 64bit applications? ABC and RieselSieve are have a lot better performance on LINUX 64bit than on 32bit. So the power user crunch rather ABC or RieselSieve than CPDN. I think, some users (me too) will come back with a bunch of PCs to CPDN when 64bit applications are available.

The increase of the performance depends very strongly on the application and of the architecture and the OS.
A very good increasement is about 20 %. Mostly there will a perfomance benefit from about 5 to 10 %. But, it could also be, that the 64 bit application runs slower than the 32 bit application.

---

It\'s just a shame they didn\'t wrote BOINC in Java and the projects too. Our server Java applications run considerably faster on 64 bit without changing the code - just the Virtual Machine to a 64 Bit version. Zero problems with multiple platforms or upgrades like 32 Bit to 64 Bit.

If the 64 Bit version \"is not stable\", this is just a bad prove of incompetence of the software debelopers - and not a problem of 64 Bit.

---

64 Bit BOINC on 64 Bit Linux

dentaku

As has been posted all over these boards, some of it in a post just before yours, the climate programs are the ones used by the UK Met Office for weather forecasting and climate modelling.
They have been written and evolved over many decades, by many software engineers and climate scientists, to run on their supercomputers, the latest of which can be seen here.

The one converted to desktop computers for use here, is over a million lines of code, and written in Fortran. All of this was started way before Java was thought of, and Fortran is far more suitable for the purpose than Java.

Climate models are, by their very nature, unstable, because they are modelling complex, chaotic systems made up of many variables. Getting a program to produce a model that mimics the weather/climate is no easy task, and definitely not one for something as slow as Java. Even the supercomputers of the Met Office take a day or so to produce just one model.
The problem is getting the program to not discard values for very tiny variables, such as sulphates, while at the same time using quite large values for other variables. When the programmers ran a test with 64 bit models a year ago which proved to be unstable, they just didn\'t have the time to investigate further; their mandate was to produce a 32 bit model for the ten\'s of thousands of people with 32 bit computers, not the thousand or so with 64 bit systems.

In a couple of years, when the majority of people may well be using 64 bit systems, the models will be there.
For now, the research is proceeding quite nicely on those computers that are currently suitable.

As I stay with my opinion, that it wouldn\'t be probably a problem to offer CPDN for the 64 Bit BOINC users (as other 32 bit projects already do without special 64 Bit versions of their applications ...), I understand that the project maintainers don\'t have time for this. For this reason, I quit CPDN and continue projects that run with 64 Bit BOINC (no matter if 64 or 32 bit client applications ... like SETI@home).

---

64 Bit BOINC on 64 Bit Linux

Dentaku\'s aborted model had an interesting graph:



I agree that 64-bit applications would be an improvement, but as this model took a year minus a day to reach this point and still had about 73% of the task to crunch, I\'m afraid I fail to understand why he was in such a rush for a new application.

---

Cpdn news

Easy: because this task was due within a few days and it still had more than 1000 hours to complete. So I aborted this task as it would fail anyway. As this \"x86_64-pc-linux-gnu not found\" still occured, the aborted tasks still hung in my boinc and didn\'t got send back.

But the surprise is: at 17th October, suddenly the aborted task vanished! And I got credit for the last work. It seems , that the server now is fixed to accept results from 64 bit clients! :-)

Then I can continue running my last climate task. 1969 hours to go ... :-/

---

64 Bit BOINC on 64 Bit Linux

Another person who hasn\'t read the hundreds of posts saying that THERE IS NO DEADLINE ON CLIMATE MODELS.

Anyone who can\'t finish a model in a year is just not trying.
Taking several years to complete a model runs the risk of the data being obsolete.

Another person who hasn\'t read the hundreds of posts saying that THERE IS NO DEADLINE ON CLIMATE MODELS.

Anyone who can\'t finish a model in a year is just not trying.
Taking several years to complete a model runs the risk of the data being obsolete.

Well, BOINC manager shows a deadline for the climate task: 18.03.2008.And 1968 hours to go. That roughly means, I need to calculate climate 12 hours a day. But I don\'t have my computer on all days and not for 12 hours. So what happens at 18.03.2008 .... ????

---

64 Bit BOINC on 64 Bit Linux

So what happens at 18.03.2008 .... ????

I\'ve no idea.
Earthquakes?
Floods?
Tornadoes?
Politicians turn honest?


Your climate model, however, will continue on it\'s way, slowly crunching towards 2080. (Or whatever date it\'s due to finish, depending on model type.)

BOINC will get huffy and start telling you (perhaps in red text), that it\'s past it\'s deadline. Ignore BOINC. This project has it\'s own ideas about when a model finishes, which is when it crashes, or when it gets to the end of it\'s task.
And you also continue to get credits.

---

Backups: Here

Dentaku, when you finish this model and get a new one, look first in your project preferences to see what choice of new models you have. The current HADSM models are shorter than the HADCM. The HADAM models from SAP are also relatively short, but you need 1Gb RAM per model.

The choice we have may not be the same in 3 or 6 months.

I think it\'s time for a reminder in the News thread about the deadline not applying to CPDN. It\'s a pity that this boinc message can\'t be eliminated for CPDN models.

---

Cpdn news

I think it\'s time for a reminder in the News thread about the deadline not applying to CPDN. It\'s a pity that this boinc message can\'t be eliminated for CPDN models.

Well, why don\'t you write this information as a message that is displayed in the \"Messages\" tab? E.g. all characters CAPITALIZED:

BOINC DEADLINE DOES NOT APPLY TO THIS PROJECT!!!

:-)

1963 hours to go ... (waiting for Phenom to tripple performance ...)

---

64 Bit BOINC on 64 Bit Linux

As far as I know it isn\'t possible for project-specific notices like that to be included in boinc messages. I\'ve asked about this before and will ask again.

---

Cpdn news

Hi all

First post so be gentle Ok. :)

Long story short; I am one of the geezers that would much appreciate being able to fully utilize my AMD64 processor on this project. I am a BOINC/CPDN rookie, and I have tried reading all over this site to find out if full support for 64bit is available. I just saw that most of this thread\'s posts are quite old now, and I also noted that there are 64 bit applications (don\'t fully know what an \"application\" is in this sense) now.

Why am I so interested in utilizing my computer to the full? Because I\'ll be competing with a friend and he don\'t have 64 bit OS, so I\'d have an advantage here if it works. So.

Does 64 bit (i.e., 64bit proccessor, operating system, and BOINC client) work now days?

Hi Jimmy,

I\'ve got good news and bad news. Firstly, the answer to your question is \'yes\' - select \'HadSM3\' from Your Account / climateprediction.net preferences.

On the other hand, it\'ll give you no advantage at all (the actual climate model is 32-bit code running within a 64-bit framework).

---

I'm a volunteer and my views are my own.
News and Announcements and FAQ

... On the other hand, it\'ll give you no advantage at all (the actual climate model is 32-bit code running within a 64-bit framework).

Ohh nooo!!! :) I\'ll crunch 32-bit style then i suppose. It\'s just that it will be so much harder for me to beat my mate now. Thanks for the quick and prompt response nevertheless. Much appreciated. ;)

---

On the applications site (http://climateapps2.oucs.ox.ac.uk/cpdnboinc/apps.php), there is a

Linux running on an AMD x86_64 or Intel EM64T CPU

But when I refresh my climateprediction.net project it just
says \"got 0 new tasks\".

My config: 64 Bit Ubuntu 7.10 Linux with 64 Bit BOINC 5.10.28.

---

64 Bit BOINC on 64 Bit Linux

The project has run out of slab models.
See server status page.

Hi Jimmy,

I\'ve got good news and bad news. Firstly, the answer to your question is \'yes\' - select \'HadSM3\' from Your Account / climateprediction.net preferences.

Does \"HADSM3 mid-Holocene\" work that easily? I guess I\'ll find out soon enough, I just updated my Application Preferences to prefer both those types of WUs. If you already know, cool, & if not I\'ll just reply to myself here when I find out.

On the other hand, it\'ll give you no advantage at all (the actual climate model is 32-bit code running within a 64-bit framework).

It gives me the advantage that I don\'t have to install a 32-bit operating system over the 64-bit operating system I\'ve already installed, and I can run CPDN work. I understand, you mean the 64-bit code isn\'t faster than the 32-bit code, for the same hardware, but 64 is much faster on some other projects. So as long as it is at least equal on this project, it is a net gain.

---