My sec/TS value had been pretty consistently in the 2.7-2.8 range on my Windows machine running either 4.05 or 4.13 However, since upgrading to 4.19 in late January - the value has climbed up to the 2.95 - 3.05 range. Since I upgraded shortly after completing a run, it could be that I just have a new model with a set of parameters that require a bit more computation. Has anyone else seen any change in their speed under 4.19?

If so, do we know whether that is a "good" hit (e.g. an increase in computation caused by a more reliable error checking procedure or some such), or a "bad" hit (e.g. an increase caused because the BOINC developers don't do as much work optimizing a Windows version for speed.)

For what is worth, I don't seem to notice any comparable performance hit since upgrading my (noticeably slower) Linux system to 4.19 - so if there is anything going on, it may only affect Windows machines.

---

Looking at your trickles the difference seems to be 2.79 to 2.93. I think that uk_Nick posted that he was seeing up to 5% difference in speed between different phases of the same model and up to 10% difference with different models. Your difference is well within that.

I haven't seen any explanation of different speed of calculation but hotter model do seem to run slower. (Don't think that will help explain in this case)

---

Visit BOINC WIKI for help

And join BOINC Synergy for all the news in one place.

Additional to what crandles was mentioning - which is correct - Windows may be affected by disk fragmentation as well. Note that CPDN is quite stressful to HD so you may regain your original performance by degradmention HD.

---

<i>phpBB forum for CPDN, all are </i><a href="http://www.climateprediction.net/board">invited</a>

FWIW -

I believe I saw a small difference in slightly longer S/TS occuring around a month, or so, ago, several weeks before I upgraded 4.13 ---&gt;4.19. It was a very small difference, and fell w/i the oft quoted +/- 5% margin of difference. I had 12 models running at the time, &amp; it seemed all the new ones were running ~5%, or so, slower, than the preceeding model, on any specific machine. I thought about posting this, but it didn't really seem important, or easy quantifiable, either. It was just was something I casually "thought" I noticed, but didn't have "hard" numbers. I didn't think of it again, 'till I saw this thread.

I'm pretty sure it started before my 4.19 upgrade.

I wish I kept more exact records, or could be more specific. I just recall thinking "Hmmm...I've run into a series of slightly slower models".

Like I said, FWIW. My opinion is free - You get exactly what you've paid for! ;-)

Strat

---