UT theoretical chemistry code forum

For 13568 secs the points awarded are 5.20 and this has been happening for some time - using v4.02 on Windows XP - wide variation in award points is causing concern. Is there a problem with your allocation method? Please advise

The boinc point system is clearly a bit erratic -- there are quite a number of posts about this. The problem is that we (as a project) do not control this. The boinc system assigns points based upon the time the a job runs, it's CPU benchmark, and some history dependent parameters (which we don't fully understand). In the short term, credit/time can fluctuate a lot. This is particularly the case when the job time changes. Over days, however, I have always seen the system adjust and give sensible credit for work done.

We are not ignoring the issue; it is just outside of our immediate control.

Thank you for the quick reply - this difficulty has previously appeared on the QMC project (http://qah.uni-muenster.de/projectinfos.php ) which was able to work around the BOINC problem; they say "...Credits: Because of several complaints about cheating-attempts, we STOPPED to give out credits according to the standard BOINC system (see here for details). We assign every workunit series a FIXED CREDIT value according to a calculation on our reference system...." Hopefully, you might give consideration to basing value awarded on your own reference system, not that of BOINC. I tend to take this work too seriously sometimes.

WU 1391271348_1440_0 Computer ID 28285 Run time 27,283.44 CPU time 27,283.44 Credit 5.20

Previous WU on same machine - 378334476_20_11450885_0 Run time 428.21 CPU time 428.21 Credit 2.27


647 times longer to run for 2.3 times the credit. I'm sorry but I'm aborting these very long WUs as this is ridiculous.

graeme wrote:The boinc point system is clearly a bit erratic -- there are quite a number of posts about this. The problem is that we (as a project) do not control this. The boinc system assigns points based upon the time the a job runs, it's CPU benchmark, and some history dependent parameters (which we don't fully understand). In the short term, credit/time can fluctuate a lot. This is particularly the case when the job time changes. Over days, however, I have always seen the system adjust and give sensible credit for work done.

We are not ignoring the issue; it is just outside of our immediate control.

Does eOn2 use CreditNew? Almost every project (except Seti) dumped that because it is FUBAR.

I have to agree with this. I do work for lots of projects. I would like to do more work for EON. The science looks interesting.

But I am not even getting 1 credit of work per completed work unit in many cases. This is really horrible credit awards! I have as an experiment left EON running for weeks at a time allowing up to one core for EON work. I can't accomplish even 100 points per day although I complete numerous workunits. This is such poor credit that I usually just shut off EON for periods of time and allow projects that actually reward their users for work. While the science is the most important part, users want to have some mark of what they have accomplished. This project is currently about 1/10th the amount of credit it should be.

BOINC does not control the points awarded. Thats up to project administrators and how they do credit. While it's true that every project has to make up their minds on how this would be best done for their individual apps, it's not at all a BOINC required method. I'm guessing that the administrators have chosen the "Credit New" system which has been deemed a total failure and has been abandoned (or never utilized) by most projects. Most project managers opt for either a flat credit per work unit or a modified time based one appropriate to their own work units. The credit new system is not a resonable award system and was designed to be skewed to SETIs method of awards which was already dismal at best.

EON looks like it could be a good science project but isn't going to ever have much of a user base with such horrible credit awards. I recommend the administrators go to a site like BOINCSTATS and compare their project with others running to get a sense of what credit should be awarded. Right now, EON is far below the curve and one of the worst projects for credit awarded. Too bad considering the science looks good.

I agree with all posts on this subject.

The average points awarded in this project are around 2.0 for about 10-12 mins CPU time.

Based on other projects it should be around 10-12 for this amount of work.

A resonable guide would be 1 point per minute of CPU I think.

I would rather do a lot more for this project as run times are easily managed especially in a multi core environment.

Is there still something wrong with the way points are awarded for this project ?

Hi guys, i registered just yesterday, but im sad from credit ration per task.

If you want keep your existing crunchers + motivate new members -- you need to change credit politics ASAP.

The same issue was in Asteroids@home project and it has been solved by constant credit per task -- 120 points/task
-- one task is cca 2,5h on FX8@3,8Ghz.
- discuss thread here http://asteroidsathome.net/boinc/forum_thread.php?id=15

I fully agree with Debian2 --- 1credit/1minute of CPU would be great, so cmon guys!

wish2 -- this is probably first BOINC project, where is necessary to have another registration just for Message Board :-(
- just merge it with BOINC accouint on your main project page pls.

wish3 - what about GPU app?

p.

So if one's just up for loads of credits, let your CPU/GPU test the Collatz conjecture ;-)


Actually, I think we can compute some 4.5 billion years without falsifying the conjecture, at least we know the answer isn't 42.

peeticek wrote:Hi guys, i registered just yesterday, but im sad from credit ration per task.

If you want keep your existing crunchers + motivate new members -- you need to change credit politics ASAP.

Nonsense. Only a few crunchers turn into total credit whores. Most of us don't care at all about the credits. Never mind telling me ridiculous stories about how this project or that project had crunchers leaving faster than rats leaving a sinking ship due to credits because all that really happens is a handful of crunchers post that they are leaving and that somehow gets exaggerated into thousands quitting.

The same issue was in Asteroids@home project and it has been solved by constant credit per task -- 120 points/task
-- one task is cca 2,5h on FX8@3,8Ghz.
- discuss thread here http://asteroidsathome.net/boinc/forum_thread.php?id=15

I fully agree with Debian2 --- 1credit/1minute of CPU would be great, so cmon guys!

This projects tasks are pretty close to all the same size/length so fixed credits would work.

wish2 -- this is probably first BOINC project, where is necessary to have another registration just for Message Board :-(
- just merge it with BOINC accouint on your main project page pls.

This is a phpBB based forum. Several BOINC projects use this kind of forum and several others use vBulletin. This is hardly the first project to not use the stock BOINC forum software. Another login is not a lot of work for anybody. There is no way they can "just merge it with BOINC account on main page".