Hi,

this morning at utc I completed with <a href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/show_host_detail.php?hostid=86867">this machine</a> pase 1 of current model. It just took 2.42xx sec per timestep to compute. Now this host is running phase 2 of this model with current speed of 2.35xx sec/TS.

There were no changes to hardware or software setup of this host. I could'nt saw this at any other host of my farm yet.

Any minds?

Ciao, Tom

&gt; It just took 2.42xx sec per timestep to compute. Now this host
&gt; is running phase 2 of this model with current speed of 2.35xx sec/TS.

Tom,
Model phases can vary +/- 15% in their computational complexity, as can individual models. The change you are seeing is in the expected range. From experience I think you may well find that the model will slow a little as it gets further into the Phase

---

Andrew

<a href="http://cpdnforum.info">CPDNforum<a>

Taking a closer look at the first trickle of phase 2 one can see that this host only needed 6908 sec of CPU Time. Normaly it needs about 26000 seconds. The resulting speed of the first trickle with its 6908 sec for its 10802 TS lies about<b> 0,64 sec / TS</b>. There seems to be something wrong! I don't know what but it looks like.

Ciao, Tom

&gt; Taking a closer look at the first trickle of phase 2 one can see that this
&gt; host only needed 6908 sec of CPU Time. Normaly it needs about 26000 seconds.
&gt; The resulting speed of the first trickle with its 6908 sec for its 10802 TS
&gt; lies about<b> 0,64 sec / TS</b>. There seems to be something wrong! I don't
&gt; know what but it looks like.
&gt;
&gt; Ciao, Tom

Hi, Tom,

Please check the viz. Is the temperature globe all blue? There were some "fast-processing ice-balls" in Classic CPDN and, presumably, some are here as well. (Also some *extremely* slow ice-balls.)

Cheers,
Jim

---

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

Processing has normalized. The second trickle of phase 2 has needed the normal 26000 seconds. Now the globe looks also normal. How it looked during the abnormal trickle phase I don't know.

So now all seems to be normal but the fast trickle remains mystery.

Ciao, Tom

Perhaps is started to be a fast iceball, but the model caught it and did a roll back.

---