Running with the noted 600 second write to disk setting, laim on. The efficiencies vary quite a bit, which may be related to the sizes of the compared sequences and the output file. They vary greatly.

Can't believe a wtd is not adhered to in line with the checkpoint interval at most setting. If there'd be one every 'internally unlogged checkpoint', i'd be fuming, but no, the 600 seconds has certainly raised the lowest efficiency well above 95 percent, in fact now running all cores ugm and the range is 96 to 98.6.

The 3 hours average runtime, in my dreams. This not pitiful device gets 3:28 hours at best. The project average is just a fraction under 4 hours.

Back on efficiencies, see the ugm effecting mcm. The more ugm run the lower too for mcm. With mcm only, 99+. Just wonder why all cores under ubuntu gives 99.8, one even hitting 99.97 percent efficiency, upon completion. Does size matter? It seems.

I am getting 100 % efficiency on Windows Vista on a stock q6600. On a Linux machine I was getting 99%+ efficiency on a core 2 Duo at 2.67ghz. I see efficiency as good or better than the beta runs.
(Got to get the bronze and then back to MCM1)
Cheers

Did some tweaks, to include excluding the data-dir from av scanning, and switching os priority from background to foreground processes and getting 99 percent or better on ugm, but it keeps degrading the more are run, and keeps impacting mcm too. Coded up an app_config to set for 2 / 6 ugm / mcm run ratio, which are the allowed sciences at this time for the node's associated profile. Slightly risky, as the ratio of ugm/mcm is getting 1 in 9, so it could happen a ugm core goes idling. Surely the technicians will change this once they've figured out the storage needs with a few weeks live and no assimilated results, at lowest feeder preference.

<app_config>
<app>
<name>cep2</name>
<max_concurrent>1</max_concurrent>
</app>
<app>
<name>mcm1</name>
<max_concurrent>6</max_concurrent>
</app>
<app>
<name>fahv</name>
<max_concurrent>7</max_concurrent>
</app>
<app>
<name>faah</name>
<max_concurrent>1</max_concurrent>
</app>
<app>
<name>ugm1</name>
<max_concurrent>2</max_concurrent>
</app>
</app_config>

To do still is create a swap file specific partition, stopping fragmentation, per http://www.online-tech-tips.com/computer-tips...figuring-the-paging-file/ . Even a few tenth of a percent improvement counts up to lots of hours extra computing in a year.

The ratio seemed to have bettered a bit. Now about 18 hours further, got 5 ugm and 30 mcm in ready to start for a 0.5 day buffer, making it about a 1:6 for this project combination. The ready to start dropped as low as 3, but so far the re-activated simap -a zero weight project- to fill any idle cores, has not had to be called on.

Efficiency continues at just 99 percent with a 2-6 computing share. Wish there was a way to log the file sizes coming down. This way maybe could correlate performance to the size of the txt files. The latest were 6.1mb for 2 tasks out of batch 574.

Looking in the project folder, noticed the time stamps of the slideshow image files are being updated to that of the last task downloaded or updated. For instance the ugm slides have a creation date of 10.14 but a modification date/time of just few minutes ago on oct.23. This is with all project slide files, faah/fahv/cep2/mcm/ugm. Date modified being constantly updated may be something else impacting performance. Need to find why and if this can be switched off.

Yuk, just happened again, timestamps moved 11 minutes forward on all the slide files, some 40, and one mcm txt file, which is the one just started at that moment. Is that with all the projects the case? Testing, suspended a running ugm, the app_config forcing another ugm to start and, and timestamp modding. Forced another ugm to start and it happens, next mcm starts, all timestamps renewed again.

@ technicians: Why are the slide file time/dates modified when a wcg task launches, all slides for all projects, this on windows.

Hmmm, the search indexer would not like that and get hyperactive. Need to exclude this boinc datadir structure from index inclusion, if it can be, and it can be http://www.sevenforums.com/general-discussion...lude-folder-indexing.html and http://www.sevenforums.com/tutorials/36098-search-options-change-restore.html and done.

Interesting point; I always disable the Windows Search service when installing Win7 to eliminate unnecessary disk accesses, since it isn't really needed with a fast SSD anyway.