I figure about 20000 sec per R4 task divided by about 200 credits and that is around 100 sec per credit roughly. So I stand by the 107 sec per R4 credit for now. Am I missing something big???
I don't know, but I'm scratching my head! ;-)
I just scanned through what you had showing for R4 with credit granted and picked the one which looked closest to the median in runtime.
Then I drilled in further and did the same for R3, and turned the crank for the ballpark rates I mentioned. :-?
That seemed to agree with what I've been seeing in my local task database for my hosts' rates, when I selected power app tasks from R3 and the stock R4.
I know I've said it before, but after the last 2 days, I am finally and truly DONE with the subject. Mostly because I would rather remain ignorant of the depths of deplorable insults that some will resort to over such an utterly meaningless subject. I'm sure my curiosity will get the better of me and I'll forget and click into some credit discussion in the future, but I'm also equally sure that my sense of disgust will make me leave it just as quickly.
Do points make prizes? - can volunteer projects afford not to award credit/s - how many will leave that project because of credit issues, however you define the issues - how many will join because of the credit value/s?
How may volunteers do E@home need to move the science onwards at a rate they hope for/can cope with?
How much time is devoted to the science, how much to the credits etc
Will E@home take that gamble, I'm sure some SQL query will be able to tell.
Do points make prizes? - can volunteer projects afford not to award credit/s - how many will leave that project because of credit issues, however you define the issues - how many will join because of the credit value/s?
How may volunteers do E@home need to move the science onwards at a rate they hope for/can cope with?
How much time is devoted to the science, how much to the credits etc
Will E@home take that gamble, I'm sure some SQL query will be able to tell.
just a few ponderings/observations/thoughts
There's no database, SQL or otherwise, to interrogate for those questions. We have a summary page of current overall project output here.
I don't know of anyone who is proposing abolition of credit. However adjustments of credit 'worth' between projects is a running topic. In that regard I suppose that's a matter for management judgment. As I've indicated before, personally I think there is a watershed being approached by project managements ( existing and potential ) as regards the cross project credit issue. My guess is that it is hard enough to cope with the variety of issues within a single project framework without having to extends one's assessments beyond financial, legal, technical and administrative borders. One will break away and the rest will follow. Mainly because for some projects that may be the only way up, that or getting out of the DC idiom entirely. For others the rigamarole is getting too cumbersome.
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
I agree with Mike, as the additional burdens imposed on DC Project managements to develop an equitable cross BOINC projects reference is significant and irksome.
I do think that an internal credit system for a project makes things more interesting from a number of view points -
1. Gives a reference on how the individual is and how things change with time.
2. Can give comparisons for different hardware make up, assuming the crunching mix is not to different.
3. Gives a reference for inter-team and extra-team rivalry to happen. It also gives some measure for "inter team competitions, etc,
Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!
I agree with Mike, as the additional burdens imposed on DC Project managements to develop an equitable cross BOINC projects reference is significant and irksome.
I do think that an internal credit system for a project makes things more interesting from a number of view points -
1. Gives a reference on how the individual is and how things change with time.
2. Can give comparisons for different hardware make up, assuming the crunching mix is not to different.
3. Gives a reference for inter-team and extra-team rivalry to happen. It also gives some measure for "inter team competitions, etc,
Yep, I agree in principle with everything you and Mike have mentioned here.
As it stands now, the only viable way to even get close to parity at any level I can see would be if the projects took a plain vanilla math version of their to be run publicly application, and then ran it on a host calibrated to the reference computer specifications to set the basis. Then they would have to implement some sort of server side determined work to credit conversion or fixed scoring system depending on the characteristics of the work flow, ie whether it was deterministic or not.
Once that was done, then it doesn't matter how much the app is improved later by optimization, since the basis was set according to the definition of the Cobblestone. The hosts which could make use of the more efficient app would get their 'advantage' by being able to run more tasks in a given time interval, rather than deflating the basis value of running the exact same work on a host which doesn't have the features, which is what the 'Plan' does.
As you point out, that can be a lot of extra work not directly related to the science goal of an individual project, and has to be done over any time there is a major algorithmic change, but there isn't really much choice. Given that BOINC was designed right from the start to invite comparisons between projects and 'competition' between participants, it's kind of late in the game to come out and start saying the whole foundation of that is doo-doo! ;-)
Given that BOINC was designed right from the start to invite comparisons between projects and 'competition' between participants, it's kind of late in the game to come out and start saying the whole foundation of that is doo-doo! ;-)
I like the idea of having an in lab cruncher which represents the median rig.
This would crunch in the normal way, and, once settled down, would provide a dynamic basis for server side credits awarded. More, it would also contribute to the science.
The only lip service to cross project parity would be the use of that defined medium crunching rig.
Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!
There would probably just as many complaints about the lab cruncher as there are about credits now. OS/RAM/cache memory/number of cpu's/family of cpu etc. etc.
There would probably just as many complaints about the lab cruncher as there are about credits now. OS/RAM/cache memory/number of cpu's/family of cpu etc. etc.
More than that, the informal survey that was done some time ago showed that "credit granted vs. another project" was a very small reason for people...
I keep asking, why is this so darned important to David Anderson and an equally small subset of the user base as the group that he/they are so against (people who pick projects based on credit)???
RE: I figure about 20000
)
I don't know, but I'm scratching my head! ;-)
I just scanned through what you had showing for R4 with credit granted and picked the one which looked closest to the median in runtime.
Then I drilled in further and did the same for R3, and turned the crank for the ballpark rates I mentioned. :-?
That seemed to agree with what I've been seeing in my local task database for my hosts' rates, when I selected power app tasks from R3 and the stock R4.
Alinator
Interesting discussion, but
)
Interesting discussion, but not relevant to what happens in each project.
I just leave my older rigs crunching, calculating and reporting.
I am off in another project with the newer rigs, and will bring these over when I have reached a target there.
What my output per day/hour/etc is unknown ATM.
Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!
RE: I know I've said it
)
Do points make prizes? - can volunteer projects afford not to award credit/s - how many will leave that project because of credit issues, however you define the issues - how many will join because of the credit value/s?
How may volunteers do E@home need to move the science onwards at a rate they hope for/can cope with?
How much time is devoted to the science, how much to the credits etc
Will E@home take that gamble, I'm sure some SQL query will be able to tell.
just a few ponderings/observations/thoughts
RE: Do points make prizes?
)
There's no database, SQL or otherwise, to interrogate for those questions. We have a summary page of current overall project output here.
I don't know of anyone who is proposing abolition of credit. However adjustments of credit 'worth' between projects is a running topic. In that regard I suppose that's a matter for management judgment. As I've indicated before, personally I think there is a watershed being approached by project managements ( existing and potential ) as regards the cross project credit issue. My guess is that it is hard enough to cope with the variety of issues within a single project framework without having to extends one's assessments beyond financial, legal, technical and administrative borders. One will break away and the rest will follow. Mainly because for some projects that may be the only way up, that or getting out of the DC idiom entirely. For others the rigamarole is getting too cumbersome.
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
I agree with Mike, as the
)
I agree with Mike, as the additional burdens imposed on DC Project managements to develop an equitable cross BOINC projects reference is significant and irksome.
I do think that an internal credit system for a project makes things more interesting from a number of view points -
1. Gives a reference on how the individual is and how things change with time.
2. Can give comparisons for different hardware make up, assuming the crunching mix is not to different.
3. Gives a reference for inter-team and extra-team rivalry to happen. It also gives some measure for "inter team competitions, etc,
Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!
RE: I agree with Mike, as
)
Yep, I agree in principle with everything you and Mike have mentioned here.
As it stands now, the only viable way to even get close to parity at any level I can see would be if the projects took a plain vanilla math version of their to be run publicly application, and then ran it on a host calibrated to the reference computer specifications to set the basis. Then they would have to implement some sort of server side determined work to credit conversion or fixed scoring system depending on the characteristics of the work flow, ie whether it was deterministic or not.
Once that was done, then it doesn't matter how much the app is improved later by optimization, since the basis was set according to the definition of the Cobblestone. The hosts which could make use of the more efficient app would get their 'advantage' by being able to run more tasks in a given time interval, rather than deflating the basis value of running the exact same work on a host which doesn't have the features, which is what the 'Plan' does.
As you point out, that can be a lot of extra work not directly related to the science goal of an individual project, and has to be done over any time there is a major algorithmic change, but there isn't really much choice. Given that BOINC was designed right from the start to invite comparisons between projects and 'competition' between participants, it's kind of late in the game to come out and start saying the whole foundation of that is doo-doo! ;-)
Alinator
RE: Given that BOINC was
)
I like the idea of having an
)
I like the idea of having an in lab cruncher which represents the median rig.
This would crunch in the normal way, and, once settled down, would provide a dynamic basis for server side credits awarded. More, it would also contribute to the science.
The only lip service to cross project parity would be the use of that defined medium crunching rig.
Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!
There would probably just as
)
There would probably just as many complaints about the lab cruncher as there are about credits now. OS/RAM/cache memory/number of cpu's/family of cpu etc. etc.
RE: There would probably
)
More than that, the informal survey that was done some time ago showed that "credit granted vs. another project" was a very small reason for people...
I keep asking, why is this so darned important to David Anderson and an equally small subset of the user base as the group that he/they are so against (people who pick projects based on credit)???