Thread: Universe@home
View Single Post
Old 11th December 2015, 07:49 PM
Coleslaw Coleslaw is offline
Join Date: Oct 2013
Location: IL
Posts: 606
I'm guessing they are using CreditNew, which means credit will always fluctuate widely. I've not seen a CreditNew project that has truly leveled out as designed without the project manually interjecting in some way. If the credit fluctuation is because of CreditNew, then there really isn't much to consider here unless you kick out other Vault projects like WCG which also sees credit swings.

Estimation time is a problem at many projects and is more of a BOINC Manager issue 9 times out of 10. If there are multiple apps or the batches change in size regularly, there isn't much you can do.

The other issues you mention, I would have to look more into in order to have more insight as I have not been running it lately due to other testing, WCG challenge we are in, and personal goals. Since you experienced the issues, can you shed more light on what was going on?


Thank you for feedback
The points problems arrive with short series of WU, where different sets of start parameters gives in results very different computation time.
I had trying to use built in "credit_new" system, but as some of you seem it work... randomly, so I decide to come back to previously used points_per_wu system as it is more predictable.

I think even for series where computation time is as twice bigger then to others current points are good... Let me know if you don't thing so.

Anyway, the problem is that single batch have only 2000WU's and you all take it faster than my machine can finish it to estimate points... (we all doing more than 30'000 WU's per day)

Anyway, very soon we probably again will start longer workunits where I can more accurate predict computation time and set points more accurate.
[link= mId=BP5XNJBR9N1][/link]

Last edited by Coleslaw; 11th December 2015 at 07:52 PM.
Reply With Quote