I don\'t usually check the graphics of my models, but I just checked the trickles.

My fastest model is a type HadSM3MH running at 0.4959 sec/TS. Anyone have one of these models running (stable) faster than that? Post your latest trickle and phase number.

Both types of hadsm3 models can become \"fast processing iceballs\" under some circumstances.
Keep an eye on the graphics.

It may be that the best indication of this is after an \'end-of-phase\' zip has been uploaded. If it has gone \'fpi\', then the server graphs will show abnormal data lines. (Data points dropped down below the bottom / climbed up above the top of the chart or graph line climbing / dropping at sharp angle.)

---

Backups: Here

DJStarfox

I have 16 of these models running on i7 boxes and they all run ~1 second per T/S.
I have 4 of these models running on a Phenom quad which run at ~.95 - .97 seconds per T/S.

I wonder if your model is on a AMD based machine, in which case it could well be a fast running Iceball world.
Click on the model concerned then on the \"Show graphics\" button on the left side of the BOINC Manager and have a look at your world. If it is a uniform colour, then you have an Iceworld.

---

I\'m not able to tell if it\'s an iceball or not (can\'t look at graphics). I did check the P1 & P2 graphs: They appear normal temperature and pressure.

I have an i7 running on Fedora linux x86_64. Why do I have such good sec/TS? Could be because I run a mix of projects?

All my boxes run Windows. i did have a box running Linux for a while but the climate models ran slightly slower over time so I abandoned it. Even then I could run the graphics mode.

Your computers are hidden so I can\'t look at your models directly and see what is happening to them.

I\'m not able to tell if it\'s an iceball or not (can\'t look at graphics). I did check the P1 & P2 graphs: They appear normal temperature and pressure.

I have an i7 running on Fedora linux x86_64. Why do I have such good sec/TS? Could be because I run a mix of projects?

---

Sorry, I forgot the links. Here are the two in progress right now.

http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=10489487
http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=10257455

The fastest I have run one, was on an E8600 in Linux at .47 s/TS. It was running no other tasks at the time. My Phenom II X2 550 ran one by itself in Linux at 0.49 s/TS.

My i7 920 in Linux when running 4 at a time with hyperthreading turned off, they ranged from 0.49 to 0.54 s/TS. Running 8 with hyperthreading on, they average about 0.88 s/TS. I have not tried just one on the i7 920.

The only overclocking was on the Phenom II which I had running at 3.4 GHz, up from the 3.1 GHz stock.

The two models in your links look fine, DJStarfox. No speeding up in the sec/TS column. As you can\'t see the graphics that\'s what you\'d look for (speedup on AMD, slowdown on Intel).

On my C2D 6600 with XP a HadSM is doing 1.31. The i7 is clearly much faster.

It\'s the HadAM3P that you shouldn\'t fully load a multicore with. All model types will slow each other down a bit but with HadSM I don\'t think this is worth worrying about.

---

Cpdn news

Linux-64 is very fast for HADSM, regardless whether MidHol or not.
If yiou only run one model besides other projects on multicores it\'s even faster.
My fastest one up to now was this: 10527937 with 0.5048 sec/TS average, with some Trickles considerably faster.

My puter is a C2Q9450 @ 3.2GHZ running ubuntu64.

---

Grüße vom Sänger