My current task has been running for over 1248 hours, but since it hit 92.318% done the time to completion keeps increasing proportionally, so it always stays at that 92.318% (give or take a few thousandths).

The question is, am I just throwing away all the processor time I am giving climateprediction? Because if so, I shall unsubscribe and run more user-friendly projects like SETI and WCG.

I\'m not too worried if it really is a bug; I\'m happy to provide all the time needed, as long as it\'s not being wasted.

---

Hi Tom
It sounds as though your model has turned into a \"slow processing iceball\".
More about it here.

As for \"wasting time\", the point of this project is to find out what happens to a model when the values used for the many starting parameters are varied slightly. (You can see the parameters and their values that are used for each model, on the page for the model.)
Because of the chaotic nature of weather, the longer a model is run so that it \'becomes climate\', the more likely it is that the model will start to become unstable.

Most of the models will run for the full period, but some won\'t because they are unstable over a long period.
So the physicists that use these results need to know which values produce a model that\'s not realistic over a long period, as well those that produce a long running climate.

---

Backups: Here

Hi Tom, welcome to the forum.

Here are the models you\'ve had:

http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=7388278

If when you look at the model graphics you see monochrome blue for temperature and different monochrom colours for pressure and precipitation, it will have become an \'ice-world\'. This happens to a small proportion of these HADSM slab models. If this is what you see, abort it and get another model.

What I\'m going to say now is unrelated to your unfortunate slab model. One of your previous models crashed with a -107 error code which is usually caused by something graphics-related. If you go to the CPDN README collections using the link in my signature, open the collection about crashes and problems. Link #6 there is a post by MikeMars who deals with the usual advice people need to run climate models including how to avoid these graphics-related crashes.

The next link, #7 by Thyme Lawn, explains in detail how to update a computer\'s graphics card (free update from the web).

If I were you, from now on I\'d have a quick look at your model\'s graphics every couple of days. Press Z then 8 on the keyboard and you\'ll see lots more details including how the model\'s progressing through its timesteps. If you ever think the progress doesn\'t look normal, post again for advice.

We know that running climate models is more tricky than tasks from other projects. I crashed a few models before I learned how to succeed with them.

If you look at the README collection about backups, you could select a method to back up your BOINC folder regularly. Then if you crash a model you can restore your last backup and continue the same model. I use Les\'s easy manual backup & restore methods which are quick and fail-safe.

Restoring a backup would have rescued your previous two crashed models. Unfortunately a restore not have rescued your current iceworld which will be due to the initial parameter values. Two other models from the same workunit look as if they ran into the same problem at the same point. The researchers try a very wide range of possible but plausible values and some combinations just don\'t work.

If the people running those other tasks from the same workunit had reported the problem on the forum we\'d have been able to warn you in advance.

Hope that helps.

---

Cpdn news

Dear Tom

Please don’t give up on climateprediction just because of one iceball model. I had one SM model go iceball on me (at 97%) several months ago, but, I was able to run 2 others to completion and am now 75% complete on a CM model. The work is important. Global climate change is the gravest threat facing our civilization in the 21 Century.

As Les pointed out even iceball models provide the Climate Scientists with important data.

My current task has been running for over 1248 hours, but since it hit 92.318% done the time to completion keeps increasing proportionally, so it always stays at that 92.318% (give or take a few thousandths).

The question is, am I just throwing away all the processor time I am giving climateprediction? Because if so, I shall unsubscribe and run more user-friendly projects like SETI and WCG.

I\'m not too worried if it really is a bug; I\'m happy to provide all the time needed, as long as it\'s not being wasted.

---

Thanks for all the detailed help! I really appreciate it, as I was beginning to feel pretty isolated. Like many users, I just select likely-looking projects and set them going. Apart from checking their overall progress (which led to my original post) I don\'t take much interest.

First things first. I brought up the graphics by clicking the button on the Tasks tab. The graphics window came up in temperature mode, and I set it rotating to get an all-round view. It\'s plain blue with white outlines for the continents - no temperature contours or fill whatsoever. R, P, and C all give similar results, with different colour schemes of course. Looks as though nothing at all is happening - it\'s dead.

Of course I\'ll apply the several other things you and the others suggested, but does this clinch the matter on its own?

does this clinch the matter on its own?

In a word, yes.

---

"The ultimate test of a moral society is the kind of world that it leaves to its children." - Dietrich Bonhoeffer

The \"fub\" in the model name does not necessarily mean that it\'s fubared btw., I finished one of those \"fub\"s not long ago ;-)

I think the monochrome displays must show the default colours. They don\'t indicate real developments in the climate.

One member, John Hunt, had a slab model that turned into an iceball right near the end. He very kindly let it battle on for ages so we\'d see what sort of graphs it generated. This was his model. The phase 3 precipitation graph shows that from the point where the problem occurred, this part of the data just wasn\'t being processed. The model must have been making desperate efforts to process all the data, spending ages trying again and again.

Lots of models recover from processing errors. They recognise the error and recalculate from the last checkpoint, then if necessary the last day, the last month and will finally try from the start of the last model year.

So considering how long the models are, in general they\'re pretty robust.

Tom, as your computer has masses of RAM, in your CPDN project preferences (in your account) you could if you want select the HADAM type before downloading a new model. HADAMs don\'t as far as I know suffer from the iceworld syndrome.

---

Cpdn news

I brought up the graphics by clicking the button on the Tasks tab. The graphics window came up in temperature mode, and I set it rotating to get an all-round view. It\'s plain blue with white outlines for the continents - no temperature contours or fill whatsoever. R, P, and C all give similar results, with different colour schemes of course. Looks as though nothing at all is happening - it\'s dead.

OK, today I looked at the graphics again - and now it\'s perfectly normal, showing a rather hot world in about 2060. Not a spot of blue anywhere this time! Pressure, clouds, precipitation are also varying dynamically. Yet my progress is still jammed at 92.318%. So it looks as though I have not entered an \"ice world\".

What should I check next? And am I adding any value to the experiment by continuing? Please spell things out clearly for me, as I am a complete dabbler (if even that) in this field.

And now it\'s gone back to featureless!

That\'s interesting; it must be having periods of normal processing interspersed with being stuck. It could be looping and going back endlessly. It hasn\'t trickled for 6 weeks. I think you\'re going to have to bite the bullet and make it bite the dust. Two other computers tried and failed with the same model. This one\'s a loser. Abort it.

---

Cpdn news

This one\'s a loser. Abort it.

Many thanks for such a clean, definite, understandable reply! No sooner said than done...

And once again thanks to everyone who responded - one of my best support experiences, I\'m glad to say. Next time I have a problem I will have no hesitation returning... You have been warned.

8-)