I downloaded a new 80 year model 5.44 a few days ago to run with an existing 160 year model 5.15 on the other core.

When it started running it reported an s/TS of 3.9 rather than the 2.5 the other one is doing and this has been the norm for all 6 previous 160 year models. At first I thought maybe it needed to settle down but after 75 hours running it still has this very high s/TS. Comparing with the other model it is tricklng every 29 hours rather than 24 so it is definitely running a lot slower.

This model calls itself \'optimised file i/o\' which to me means it should, if anything, be running faster, accessing the drive less often. Task manager reports it is giving the same amount of time to each model so this can\'t be the cause.

I am running BOINC on an almost empty NTFS partition on a large buffered SATA drive which I regularly defrag. I also have 2GB of memory which most of the time is mainly unused. You would think these would be ideal conditions to run these models.

Do these models do the same amount of work? They seem to be doing the same number of calculations each simulated day.

---

Hi John

See the post about this in the News thread on 4 Sep (it\'s at the top of this Number Crunching section).

4 Sep 2007
The current optimised version of cpdn HADCM3 models is advantageous in that these models are more likely to complete. But they run about 20% more slowly. Tolu has now readjusted by +20% the credits we are awarded for running optimised HADCM3 models. This applies to Windows, Linux and Mac. Previous crunching of these models will be taken into account.

Credit awarded for crunching pre-optimised HADCM3 models and HADSM3 slab models remains unchanged. This means there will be no adjustment to BBC credit.

So the optimisation in this case includes adjustments to free them of fatal bugs, and this has slowed them down! (Quite a few models had been crashing due to internal flaws.) Otherwise the models are the same as regards calculations done.

One person did report that the optimised HADCM models do not run slower for him in Linux, but that could be some peculiarity of his particular Linux setup or this particular computer. He still gets the extra credits of course!

I don\'t think we\'ve had any reports of a slowdown greater than 20%.

---

Cpdn news

Thanks mo.v, I must have missed this one. There are so many forums to check for news!!

The slow down does seem to be about 20% so I am still on course to reach my 500RAC winter target (using the computer 24hours a day as an office heater!!)

There\'s a news thread on all 5 climate forums now, sometimes with slight differences in the posts according to what we think the members of each project need. The most complete record is on the independent forum where old news items go into an archive, whereas on the boinc forums old items have to be hidden.

Some old items that are of no further use to anyone or could later be considered misleading get deleted, not archived.

---

Cpdn news