Hi,

I had a job unit who was reaching 1929 and worked fine.
Few days ago, it suddently (perhaps after a bluescreen) go back to the XIXth century (now: 11/06/1813, 5.6%).
However I am not sure if this is the same job or if my old one has been cancelled and a new one started on the past.
What makes me think it may be an error is the way the graphical display looks:
- There is no longuer clouds
- In each display mode, the earth has only one color (ie, in Temperature mode, all the globe seems to be at -12°, pressure is at 950, Rain and snow at 0, etc...).

Should I terminate this job or is it normal?

Let me know if you need more information.

Thanks in advance,
Eric

It may have restarted, which is a problem caused by stopping the model just after a phase change, or it may have become an \"iceworld\".
Please read this description and post back about if it describes your model.
There is also a discussion thread about iceworlds here.

---

Backups: Here

Bonjour LeFauve

These are the models you have had (you can also click on your name here to see this list):

http://climateapps2.oucs.ox.ac.uk/cpdnboinc/results.php?hostid=843567

In reverse order, model #4 crashed in March.
#3 was reaching 1929 and crashed 2 days ago. If you have no backup, you cannot continue this model.
#2 is a HADSM slab model sent to you in March and your computer is now processing it. This is probably the model with the abnormal graphics.
#1 is a HADCM model sent to you on 13 April to replace the crashed model. This model is probably normal.

---

Cpdn news

Thanks for your replies but I am a little lost here :o)
I am not sure my \"strange\" unit is one of the blue ice world cases described in the link of Les Bayliss post, and I am not even sure it is slower than usual since I didn\'t join this program for a long time (sorry, I am not very helpfull here :o( ).
Anyways, my computer does have stabilities issues lattely.
If you think this may have effect on this particular BIONC project, I may switch to other ones (too bad I kind of liked this one).

The job is now at 7.26%, so about 313 hours from completion according to the BOINC estimation.
Should I let it a chance or just terminate it?

Eric

Is this the model with the strange graphics? Please look at the model name/number and check it against the name of the problematic model in the Tasks tab of BOINC manager.

http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=7354419
This model trickled again this afternoon and I don\'t think it is processing slowly - in fact its speed (sec/timestep) is a little bit faster than when it made its last trickle.

LeFauve, are you sure this model\'s graphics are abnormal? Sometimes when we first open the graphics window, or we restart a model, the graphics take 10 or 20 seconds to begin to display normally.

Don\'t abort this model at the moment - it may be processing normally.

---

Cpdn news

With an 1813 date, it\'s possible the motherboard battery died. That leaves the CMOS unable to maintain date and time. (You can read about this battery in your computer\'s User Manual.)

---

"We have met the enemy and he is us." -- Pogo
Greetings from coastal Washington state, the scenic US Pacific Northwest.

The 1813 date is from a different model. The original HADCM model that LeFauve was running and had reached 1928 crashed. He then got this new HADSM slab model which has now produced about 10 trickles:

http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=7354419
This is the model with abnormal graphics. The sec/timestep are getting faster and faster. I think this is one of the rare models where the model stops processing some of the data, so it advances faster than normal. These models also have abnormal graphics.

The CPU seconds between trickles have been

16315
12806
12155
12618
13224
12289
12488
13169
12416
12179

I think the model probably stopped processing normally soon after the first trickle, but I\'m surprised that the CPU time figures aren\'t almost identical. But there could be many reasons for the variations.

LeFauve, could you please continue this model until it reaches the end of Phase 1 (1825) and sends its zip file to the server. That will take 2 or 3 more days. When the server has received the Phase 1 zip file we will see the temperature and precipitation graphs for Phase 1 and will know definitely whether the model is normal or not.

If the Phase 1 graphs are abnormal we will ask you to abort the model. You will receive your credits for processing it, that\'s not a problem.

---

Cpdn news

Hi again LeFauve

Abort the model! Look at its graphs for temperature and pricipitation for Phase 1; they\'re totally abnormal:

http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?field=Precipitation&resultid=7354419&phase=1#graph

It stopped processing precipitation after one or two years and I think it\'s only processing temperatures for three seasons of the year. This is why it got faster - it wasn\'t processing all the data. The scientists cannot use these results. Just abort it and get another model. You will receive your credits for processing it.

Thank you for continuing it to the end of Phase 1 and allowing us to make a definite diagnosis.

I\'m going to send a private message to LeFauve to be sure he sees this post.

---

Cpdn news

Hi and thanks for your advices.

I have been away from the computer but I\'ve canceled the unit this morning.
I was used to the idea it was a kind of \"ice world evolution\" so the uniform icy temperature and the lack of other variations could fit with a \"dead world\" but it seems I was wrong :o)

I hope I\'ll be more lucky next time !

A new unit has just started (from 1920).
Btw, (sorry for the \"off topic\"), I was wandering what is the point of predicting past events? Don\'t we have historical data on those periods?
(or perhaps it is to validate the computation model accuracy?).

Eric

A lot of the basics are described in the Climate Science pages (blue menu on the left), in particular, Experiment Strategy.

The first phase of the models is for calabration.
Roughly, it\'s to see if the \"what if\" model created by the given starting values are realistic by comparing with know events.