Only if you have the needed data files, or a backup so that you can re-run the model to get the files.

---

Backups: Here

Hi Palooka

I wanted to check the workunit your iceworld comes from to see whether any other members crunching a model from the same WU need a project email in case they haven't noticed and are wasting computer time. But your computer's hidden.

Could you please either unhide your computer for a day or two or tell us the task or WU number?

---

Cpdn news

The HADSM3 slab model has now been retired, so any further investigation of iceworlds is rather academic - but, hey, what's wrong with 'academic' on a distributed computing project? So here are my last four iceworlds.

There's one new coastal point, west of South America off the coast of Peru.


The final array of freeze points is therefore:


This investigation started from a concern that volunteers' efforts might be wasted on slow-processing iceworlds or, worse, looping iceworlds. Since no new application version has been issued, no wasted computation time has actually been saved, except by participants reading one of a number of threads on this subject and realising that a wayward model should be stopped. Though a considerable amount has been found out about iceworlds, the investigation is in that limited sense a failure. Nonetheless, as I understand it, later models have been improved as a result: given the high attrition rate of FAMOUS models, an out-of-bounds model that terminates tidily but early may not seem to be an improvement, but it is surely better than a model that doesn't. Our efforts should be to some purpose.

Great thanks are due to the following for contributing models and ideas to this analysis:

belfry
dave peachey
david glogau
dibb fosdyke
hagar
iansm
james
JIM
Les Bayliss
lockleys
martinnz
mo.v
peterfilla
thyme lawn

The slab is dead! Long live the slab!

Keep 'em coming.

The slab is dead! Long live the slab!

Keep 'em coming.

Three cheers ... but they are not being delivered here.
One machine has one core free and two will be within next 12-18 hours so tried to grab 3 slabs. 2 days buffer. No joy.
Don't wish to try for a Famous as I want a slab!!!

The usual...

17/09/2010 13:08:16	climateprediction.net	Message from server: No work sent
17/09/2010 13:08:16	climateprediction.net	Message from server: No work available for the applications you have selected.  Please check your settings on the web site.

The server maintains a queue of 100 available models. There were probably no slabs in this queue. Milo has, I think, run the transitioner and the problem of no available slabs should now be fixed.

---

Cpdn news

The server maintains a queue of 100 available models. There were probably no slabs in this queue. Milo has, I think, run the transitioner and the problem of no available slabs should now be fixed.

Good show. Now got a batch. Thanks!

Have to ask Iain if he will accept new iceworlds? There will be a few with over 120k available :)

Have to ask Iain if he will accept new iceworlds? There will be a few with over 120k available :)

Yes - I've been processing a backlog of slabs and have been implementing a cunning plan to produce more from backups by switching processors. It would be nice to 'complete' the distribution of freezes within a checkpoint interval - just to exclude that as a possibility.

Hello!

May be my work computer have a one of "Slowdown model" - task 11000318 for workunit 6797232. Value in column sec / TS increased with growing number of computed steps. I only now read info about "icewords" and do not precise remember temperature in this model, but colors in "Temperature mode" is orange, yellow and near to north - green and blue.

Number of step continuously increased (in Friday it was ~ 187300 - 188000). Other results for this unit have a usual sequences of sec / TS ratio. Model now computing on Intel Core 2 E8500 @ 3.16 GHz and DDR3 memory. Without any overclocking. May be one of part of this model computed on Intel Core 2 e4300.

It's a "fail unit"? Or not? Whether there a "scientific sense" throughout calculations? (Computaion time and credits is not a problem for me. Science is more important).

Thank you!

---

It's difficult to tell what's going on. The large increase in s/TS would seem to indicate a problem. On the other hand, iceworlds have a solid light blue temperature graphic.

This task, also on an Intel+Windows PC, is just about at the point where it should start slowing down (should know within 24 hours). If it does, it confirms this work unit is a slowdown one on Intel+Windows. The completions in this work unit have been on AMD and/or Linux.

Even if it turns out that the linked task in the previous paragraph continues as normal, there's obviously something wrong with your task. I would probably abort it. It will take more than a month to finish the final 7 trickles.

Check the model. Solid blue color and low tempereature. (Different colors may be from other computer))).

Computing must go on. :)

---

Did anyone ever figure out what the cause of this was?

I just read an interesting educational article on a weather scenario called the "Marine Layer". At first glance, this appears to be what's actually happening!

As far as I know nobody ever found the cause. But the geographical distribution of these 'iceworld' crashes in a narrow band of latitude and near the coast gives the impression that they reflect (or reproduce) some real-world physical weather condition.

That's an interesting link.

---

Cpdn news

[DJStarfox wrote:] Did anyone ever figure out what the cause of this was?

As Mo says, no - but I did continue to look. An earlier post in this thread (here) showed a chart of the step number within the checkpoint cycle at which iceworlds had started. Here's an update of that chart, in which the number of analysed models has increased from 57 to 97 (i.e. an additional 40 models):

It's noticeable that the two empty bins are still empty. One more model takes the significance of that result over 95%; to get to 99% would take 119 models in total. I have another 17 iceworld candidates, which almost gets there - but have to figure how to get Windows models running on a Mac to get close.

A very crude analysis of the data suggests that as well as the 'nulls' there may be an excess of high values (e.g. 7 is the current peak value). In other words the iceworlds are attracted to some regions of the checkpoint region and avoid others - the data does seem roughly periodic with the peaks between the nulls (perhaps three per checkpoint cycle). On the other hand it may be wishful thinking projected onto small numbers.

In that earlier post I said that "there's no harm (though some effort) in excluding the obvious": at the moment the obvious is refusing to be eliminated despite the effort.