News:

If you are a member of the Team on BOINC you still need to register on this forum to see the member posts.  The posts available for visitors are not posted to much by members.
 Remember to answer the questions when Registering and also you must be a active member of Team BOINC@AUSTRALIA on BOINC.

Main Menu

Target CPU Run Time for Rosetta

Started by WikiWill, April 27, 2009, 10:02:24 AM

Previous topic - Next topic

WikiWill

For those who run Rosetta@Home, have you noticed any difference in credit returned if you change Target CPU run time in your account preferences?

Target run time can be selected from 1 hour to 1 day, and the note against the selection list says the default is 3 hours.  I've just changed mine from 3 hours to 8 hours, since according to their FAQ this helps run more models and return more scientifically useful data.

The FAQ about this is on the Rosetta forum:
http://boinc.bakerlab.org/rosetta/forum_thread.php?id=669&nowrap=true#10375
(the FAQ lists 8 hours as the default, but I guess it must be the old default, since most of my WUs have been running about 3 hrs up to this point)

I will post here in a few weeks if I notice any big change.

Note: I forgot about this section when doing my change so still have full work queues.  I hope that doesn't cause me too many issues.  Might have to hold BOINC's hand for a few days:
QuoteIt is best to allow your work queue to drain completely before making the adjustment and then allow Rosetta to resume work and download fresh work units under the new setting. This will allow the BOINC client to accommodate the new setting for project time management purposes.

Wang Solutions

I did some testing on this but it was a long time ago and I use a setting of 4 hours, as I found that gave more credit than the 3 hour setting and allowed most models to be fully evaluated. I will be interested to see your results.

WikiWill

So far I can confirm that:

  • The predicted WU run time on all of my clients has increased from about 2:53 hours to 7:53 hours.
  • WUs in progress are now running longer and I expect will either have validation errors or will go to 8 hrs as planned.
  • This has not caused problems on most systems because my BOINC cache is 1 day or less, except on one machine where it panicked into high priority mode and I have aborted 10 unstarted WUs so far.
  • :rocks

It will take a coupla days to get reasonable credit averages which I will post for the different types of CPUs I run.

Mike Mitchell

+1 for a very fortuitous posting. I have run Rosetta down on all my boxes as I near a new (semi) major marker. I will have a look at this thread again when it is time to restart Rosetta again.
AA's > 1-Malaria 2-Tanpaku 3-Riesl Siev 4-Seti 5-ABC 6-Einstein 7-WCG 8-Seti 9-QMC 10-WCG 11-Cosmo 12-ABC 13-MilkyWay 14-3x+1 15-Rosetta 16-ABC 17-MilkyWay 18-Einstein 19-WCG 20-WCG 21-Poem 22-Rosetta 23-Docking 24-Spinhenge 25-Alternate 26-Simap 27-Alternate 28-Constellation 29-WCG 30-Edges 31-Alternate 32-Pogs 33-WCG 34-Seti 35-Pogs 36-Poem 37-Pogs 38-Asteroids 39-Pogs 40-Simap 41-Pogs 42-Seti


veebee

just echoing Wang's comments ... I remember "testing" various length WU's a fair while ago, and also ran at 4 hr WU setting, as it paid slightly higher than shorter or longer WU's ...... still pretty ordinary credit though.

WikiWill

I've run long enough now to be fairly confident of the results I'm seeing, and the Rosetta app version has just jumped to 1.64 so it's a suitable time to review.

Compared to the default 3hr run time, the 8hr run time paid 5%-15% less credit per hour on all my machines except Ubuntu 64bit on E8400 CPU.  On that single machine, credit was about 10% better, however that was averaged from only 5 WU results compared to approx 20 for other machines, so not as statistically reliable :)

I've now set back to 4hr run time, and I'll see if the results confirm what veebee saw a while ago.

Oh and just for comparison, the E8400 returns very similar Rosetta credit running Windows or Linux, but it's only 1/6th of the credit the same machine was getting running the Milkyway optimised app!  No wonder my output has dropped recently :(

Dingo

I have attached a few PC's to Rosetta, that are also running primegrid.   :oz:


Radioactive@home graph
Have a look at the BOINC@AUSTRALIA Facebook Page and join and also the Twitter Page.

Proud Founder and member of BOINC@AUSTRALIA

My Luck Prime 1,056,356 digits.
Have a look at my  Web Cam of Parliament House Ottawa, CANADA

WikiWill

After a few days of 4 hour runtime, I can see that the credit paid has come up again in most cases and is equal to or 5% better than the default 3 hour runtime per WU.

There are some posts in the Rosetta forum about team RAC falling off a cliff (http://boinc.bakerlab.org/rosetta/forum_thread.php?id=4863) in the last few days, but I don't know if this would have affected my results.

Anyway I'm now changing to a 2 hour run time, and will report the exciting news next week  :crazy

WikiWill

More work on various platforms has shown that the 4 hour run time is a good choice for Rosetta.  (You can change this in your computer settings on the Rosetta site)

Of the run times tested, compared to default 3 hour run time credit, it looks something like:

Runtime|Credit rate
2 hours|+ 5-10%
3 hours|default
4 hours|+ 5-20%
8 hours|- 5-15%

Run time can be selected in hourly units up to 12 hours, so there may be a better choice than 4 hours, but this is all I'm going to test.  I hope this has been a useful little exercise  biggrin

Wang Solutions

+1 for the effort WikiWill. It confirms the testing that I did previously is still valid.  :thanks1: