You can tell by the various issues that crop up and the lack of a mac app where we are at...as the sign says BETA (middle)

Then to follow up on what Zombie asked, do failed tasks get included in the exported stat? I would tend to think that it doesn\'t. If it doesn\'t, then what is the current failed task rate project-wide?

---

You can tell by the various issues that crop up and the lack of a mac app where we are at...as the sign says BETA (middle)

Then to follow up on what Zombie asked, do failed tasks get included in the exported stat? I would tend to think that it doesn\'t. If it doesn\'t, then what is the current failed task rate project-wide?

As far as I know it does not on any project and I don\'t have that info but I suspect is less than 2%.

You can tell by the various issues that crop up and the lack of a mac app where we are at...as the sign says BETA (middle)

Then to follow up on what Zombie asked, do failed tasks get included in the exported stat? I would tend to think that it doesn\'t. If it doesn\'t, then what is the current failed task rate project-wide?

As far as I know it does not on any project and I don\'t have that info but I suspect is less than 2%.

That is an important stat to know though. In my opinion, the failure rate in combination with the initial design sensitivity / project goals should be the primary determination for whether a project is at alpha, beta, or production. Not having an app for a platform is a bit lower on the justification scale, IMO. I would suppose it would depend on if that app was in testing or not, at least to me. If one were to not even be attempting to come up with a Mac app, you could have a stable project otherwise, but still call yourselves \"beta\" because of it. I think alpha/beta projects should reward more than the baseline due to the generosity of the participants in realizing that they may encounter problems and are still willing to try. However, you can\'t run a perpetual \"beta\"...IMO.

---

You can tell by the various issues that crop up and the lack of a mac app where we are at...as the sign says BETA (middle)

Then to follow up on what Zombie asked, do failed tasks get included in the exported stat? I would tend to think that it doesn\'t. If it doesn\'t, then what is the current failed task rate project-wide?

Remember it is the credits \"per CPU second\" that DA is counting. So it\'s not the failure *rate* that maters, but instead the number of CPU seconds that get crunched, but not counted because the WU fails, or fails to validate. For some projects/apps/WUs, the WUs fail (if they fail) in the first second. For others, it may go 23 hours before failing, or even complete but not validate.

I believe the stat sites count only validated points, and therefore miss all the un-validated seconds.

And I believe that is unfair to projects with higher failure rates.

---

Dublin, CA
Team SETI.USA

Remember it is the credits \"per CPU second\" that DA is counting. So it\'s not the failure *rate* that maters, but instead the number of CPU seconds that get crunched, but not counted because the WU fails, or fails to validate.

Yes, I realized that my wording was not good enough, but I didn\'t feel motivated enough to go back to correct it. Still, I don\'t think it would get included, as I think it only includes something into the averaging if it is granted credit. Also, it\'s not the \"stat sites\" that might \"miss\" this information. They only rely on the exported stat from the projects. It would be a project thing to get those tasks weighted properly and used to calibrate the cps value, because as you pointed out, 0 for 2 seconds and 0 for 8000 seconds are both \"0\" in the credit column, but the 8000 seconds is obviously more effort and should impact the cps value more than the 2 second effort.

Along similar lines, LHC actually awards \"half-credit\", so standardization would need to be established in this area too...

And I believe that is unfair to projects with higher failure rates.

Yes, that\'s why it is important to know what the failure rate is across the board here.

---

My credit per cpu second for my AMD system here at Cosmology has increased, while the same stat for my Intel has remained roughly the same...

Average credit per CPU second 0.011226 (AMD)
Average credit per CPU second 0.008523 (Intel)

Revisiting this, I have since increased my Intel hosts\' participation up to 90% Cosmology and only 10% SETI. Guess what\'s happened???

Average credit per CPU second 0.004776 (Cosmology)
Average credit per CPU second 0.003616 (SETI)

Looks like a stream of longer-running tasks came through for my Intel system, thus the spread in cps is now down to only 32.1% (or 1.321) instead of over 2X like it was earlier...

So, let\'s ask again: Is this the natural runtime variation that has been in effect since December, or was there any change over the past few days in reaction to this discussion?

---

So, let\'s ask again: Is this the natural runtime variation that has been in effect since December, or was there any change over the past few days in reaction to this discussion?

This is the natural run-time variation. It\'s interesting to see that your tests seem to confirm a high level of credit for C@H.

Ben

---

Creator of Cosmology@Home

So, let\'s ask again: Is this the natural runtime variation that has been in effect since December, or was there any change over the past few days in reaction to this discussion?

This is the natural run-time variation. It\'s interesting to see that your tests seem to confirm a high level of credit for C@H.

Ben

Yes it does, but that was already a given when looking at the statistics sites. That said, the level of \"betaness\" (perhaps not a word there, but oh well) is important. If you and Scott feel that the project is still \"beta\", and can quantify and qualify that with David Anderson, then I think some leeway should be alloted to you. If, however, the project is closer to being \"production\" instead of \"beta\", then I can see where David has a point, although I don\'t really think that all that many BOINC participants are behaving the way he appears to think that they are. For example, look at the high numbers of people working on WCG and Rosetta. The high participation in those projects would tend to refute the notion that there are strong indications that projects with lower credits have difficulty attracting participants.

---

Actually, I was a little surprised that the average runtime for the WU\'s has not changed too much over the last several months (sample of 12,573 WU\'s):

---

Boinc Button Abuser In Training >My Shrubbers<

Actually, I was a little surprised that the average runtime for the WU\'s has not changed too much over the last several months (sample of 12,573 WU\'s):

AHA! GREAT GRAPH. This clearly demonstrates the phenomenon I am seeing with the drastic change in my Pentium 4\'s credit per second... Apparently I was real lucky for some time as the Pentium 4 was doing extraordinarily well. Now it seems I\'m getting the slower tasks coming in...

Now, expanding on that theory, SETIs host distribution both in number of hosts and in the work done by hosts is heavily weighted with Pentium 4 processors in the top 25 as viewed at BOINCstats. This project has a higher concentration of Core/Core 2...

Not sure where to go with that for a theorum, but your graph at least does show the volitility of runtimes with Pentium 4...

---

I wonder if the great variation in run time for the P4 is due to HT. Since the FPU must be shared between the two threads, the performance of the cosmo task could vary greatly depending on what kind of task is running on the other thread.

---

Dublin, CA
Team SETI.USA

I wonder if the great variation in run time for the P4 is due to HT.

That\'s a good thought, but my 2.4 is a Northwood, which only had HT once you got to the 3.06, and only then with BIOS support...

---

I wonder if the great variation in run time for the P4 is due to HT.

That\'s a good thought, but my 2.4 is a Northwood, which only had HT once you got to the 3.06, and only then with BIOS support...

Ah, showing my ignorance here, or more so just not clarifying enough. My 2.4 is a Northwood \"B\" core (533FSB) and not the \"C\" (800FSB) variety. The B cores did not get HT until the 3.06GHz model...

So ends my clarification :-)

---

I wonder if the great variation in run time for the P4 is due to HT. Since the FPU must be shared between the two threads, the performance of the cosmo task could vary greatly depending on what kind of task is running on the other thread.

That is with Cosmo running on both threads. Both of my P4\'s show the same trend. It looks like some of the WU\'s do some calcs that the P4 doesn\'t like (P4 D is fine).

---

Boinc Button Abuser In Training >My Shrubbers<

I wonder if the great variation in run time for the P4 is due to HT. Since the FPU must be shared between the two threads, the performance of the cosmo task could vary greatly depending on what kind of task is running on the other thread.

That is with Cosmo running on both threads. Both of my P4\'s show the same trend. It looks like some of the WU\'s do some calcs that the P4 doesn\'t like (P4 D is fine).

Got any stats on cache hit / miss rates?

---

Just joined C@H yesterday. I have been with seti since 2003 and Einstein since last year.When I started seti I had no concept,or notion,of credits. It was just something I believed in.
Last year I heard about Einstein and joined that as well,by this time a member of a team.
As a 60 year-old trucker[HGV 1] {18 wheeler on your side of the pond} I have seen the imensity of the night sky, and the beauty of it, and this project is worthy of support.
I have an AMD Athlon x2 4200 and seti runs between 1.5 hours and 6 hrs., Einstein between 15 hrs and 18 hrs., and Cosmo.,so far,between 4 hrs. and 6 hrs.
I don\'t care about credits. I care about science.

---

Just joined C@H yesterday. I have been with seti since 2003 and Einstein since last year.When I started seti I had no concept,or notion,of credits. It was just something I believed in.
Last year I heard about Einstein and joined that as well,by this time a member of a team.
As a 60 year-old trucker[HGV 1] {18 wheeler on your side of the pond} I have seen the imensity of the night sky, and the beauty of it, and this project is worthy of support.
I have an AMD Athlon x2 4200 and seti runs between 1.5 hours and 6 hrs., Einstein between 15 hrs and 18 hrs., and Cosmo.,so far,between 4 hrs. and 6 hrs.
I don\'t care about credits. I care about science.

Couldn\'t agree more. It\'s about helping advance our knowledge of our universe and not about how many credits are scored. I donate my computer time not for kudos but for my own personal satisfaction in knowing that in a small way I am contributing what I can to the advancement of science.

Just joined C@H yesterday. I have been with seti since 2003 and Einstein since last year.When I started seti I had no concept,or notion,of credits. It was just something I believed in.
Last year I heard about Einstein and joined that as well,by this time a member of a team.
As a 60 year-old trucker[HGV 1] {18 wheeler on your side of the pond} I have seen the imensity of the night sky, and the beauty of it, and this project is worthy of support.
I have an AMD Athlon x2 4200 and seti runs between 1.5 hours and 6 hrs., Einstein between 15 hrs and 18 hrs., and Cosmo.,so far,between 4 hrs. and 6 hrs.
I don\'t care about credits. I care about science.

Couldn\'t agree more. It\'s about helping advance our knowledge of our universe and not about how many credits are scored. I donate my computer time not for kudos but for my own personal satisfaction in knowing that in a small way I am contributing what I can to the advancement of science.

@dougdoug
@Michael James

I second Your statements. Credits are secondary, the main goal is the science behind the projects and the fun we have with our little hobby !

---

I whole heartedly agree with the science. I wounldn\'t be here if I wasn\'t. I understand Mr. Anderson\'s concerns reguarding credits. However I do believe we will discover a model for our universe long before ever finding little green men. That is my choice. Harsh as that may sound, getting credit for the work I do is nice. As far as the calculated credit. Seti w/u\'s don\'t have multiple results per w/u. Cosmology does. My calculations are alot closer (1 to 1) than his third party stats fact finding mission. I\'d like to see more proof to back his proposterous claim. Rather than a \"oh by the way\". Secondly the box I\'m currently using can \'choke\' down a cosmology, simap, yoyo, w/u easier. It was the way I built it, and tuned. Getting credits here, or the other wonderful projects I participate, makes me feel better about the 40-60 dollars per box, I spend every month to keep my hobby flurishing 24/7. Perhaps it is SETI that doesn\'t grant enough credit, that is what I believe. Mr. Anderson, don\'t step on the feet of the people you depend on. We are the all singing, all dancing, crunchers of the world.

Stefan

I read over on the BS shout box that Cosmo, RS and QMC are all lowering credits. As far as I can tell, only QMC has made any sort of official statement confirming this. Is this true for Cosmo? If so, how exactly will the credits be changing?

---

Dublin, CA
Team SETI.USA