Lucky, I guess. Because the problems with the benchmark, the credit system is, um, acuracy challenged. To make up for that they use a method to throw out the top and bottom, average the middle (if more than one) and assign that.
One of the open issues is that the different versions of the BOINC Client is that it has different code for the benchmarking ... the later windows ones are more "accurate" read lower numbers. one of the reasons people hold on to the older versions. At some point in time, I would expect that the majority f the projects will require the later generation clients and so the early ones will fall out of use.
There have been many discussions on how to "fix" the problem once and for all... but no hint if any of the ideas will be put into code or not ...
Given that some clients/operating systems/processors consistently over claim for certain amounts of work, and others underclaim, the value of a wu is calculated (as Paul says) by an average after throwing out the highest and lowest values.
This means that for any wu you take part in, there will be at least one high claimant (you). On average you will be getting more credit than someone else whose client consistently underclaims.
As important as Credit is to many participants, it does not stop BOINC from doing science or supporting the projects. So, a resolution to the accuracy problems is lower on the totem pole than challenges that impact on the projects. For which, in the 4.7x+ series, hopefully we have solved. An improved estimate for the work buffer is one I am eagerly looking forward to my own self.
I have been slacking off and not reading the dailies on the fixes going in ... the last I looked there were over 260 ... lord only knows what there are there today. I have not been doing well, so, work has suffered ... but I *MAY* be coming out of it and will be back to my normal ornery self... :)
I have been slacking off and not reading the dailies on the fixes going in ... the last I looked there were over 260 ... lord only knows what there are there today. I have not been doing well, so, work has suffered ... but I *MAY* be coming out of it and will be back to my normal ornery self... :)
Paul,
I hope you are feeling better. I am looking forward to having your "normal ornery self" back again. :)
Hey Paul, if your like me, do all your outside business early in the day cause it is to hot in the afternoon here in SACKOFTOMATOES. Take care feel better, Paul rules......;0)
Well, I don't leave the house much ... AC is on too ... though with still 4 computers in the computer room it is still warmer than I would like ...
But my next order will be for two more wireless USB devices and I will move to more machines downstairs and one into the next door bedroom (the new Xeon).
As important as Credit is to many participants, it does not stop BOINC from doing science or supporting the projects. So, a resolution to the accuracy problems is lower on the totem pole than challenges that impact on the projects.
It's not about the credit; it's about improving my efficiency.
I've also been wondering how claimed credit is calculated.
I have two machines, a P4 and a Celeron. The Celeron always claims less than the P4 (same OS and client versions), is this because the P4's benchmark score is lower? Actually the P4 crunches more units than the Celery because it takes two at a time (hyperthreading).
I've also been wondering how claimed credit is calculated.
I have two machines, a P4 and a Celeron. The Celeron always claims less than the P4 (same OS and client versions), is this because the P4's benchmark score is lower? Actually the P4 crunches more units than the Celery because it takes two at a time (hyperthreading).
HT P4s tend to claim too much credit. Celerons can claim odd amounts too, sometimes the lack of cache causes problems sometimes it doesn't.
I'm always requesting the most credit
)
Lucky, I guess. Because the problems with the benchmark, the credit system is, um, acuracy challenged. To make up for that they use a method to throw out the top and bottom, average the middle (if more than one) and assign that.
One of the open issues is that the different versions of the BOINC Client is that it has different code for the benchmarking ... the later windows ones are more "accurate" read lower numbers. one of the reasons people hold on to the older versions. At some point in time, I would expect that the majority f the projects will require the later generation clients and so the early ones will fall out of use.
There have been many discussions on how to "fix" the problem once and for all... but no hint if any of the ideas will be put into code or not ...
It is worth spelling out why
)
It is worth spelling out why you are lucky.
Given that some clients/operating systems/processors consistently over claim for certain amounts of work, and others underclaim, the value of a wu is calculated (as Paul says) by an average after throwing out the highest and lowest values.
This means that for any wu you take part in, there will be at least one high claimant (you). On average you will be getting more credit than someone else whose client consistently underclaims.
~~gravywavy
Yes, and I mispelled accuracy
)
Yes, and I mispelled accuracy too ... :)
As important as Credit is to many participants, it does not stop BOINC from doing science or supporting the projects. So, a resolution to the accuracy problems is lower on the totem pole than challenges that impact on the projects. For which, in the 4.7x+ series, hopefully we have solved. An improved estimate for the work buffer is one I am eagerly looking forward to my own self.
I have been slacking off and not reading the dailies on the fixes going in ... the last I looked there were over 260 ... lord only knows what there are there today. I have not been doing well, so, work has suffered ... but I *MAY* be coming out of it and will be back to my normal ornery self... :)
RE: I have been slacking
)
Paul,
I hope you are feeling better. I am looking forward to having your "normal ornery self" back again. :)
Jim
Jim
Hey Paul, if your like me, do
)
Hey Paul, if your like me, do all your outside business early in the day cause it is to hot in the afternoon here in SACKOFTOMATOES. Take care feel better, Paul rules......;0)
Link to Unofficial Wiki for BOINC, by Paul and Friends
Well, I don't leave the house
)
Well, I don't leave the house much ... AC is on too ... though with still 4 computers in the computer room it is still warmer than I would like ...
But my next order will be for two more wireless USB devices and I will move to more machines downstairs and one into the next door bedroom (the new Xeon).
RE: As important as Credit
)
It's not about the credit; it's about improving my efficiency.
I've also been wondering how
)
I've also been wondering how claimed credit is calculated.
I have two machines, a P4 and a Celeron. The Celeron always claims less than the P4 (same OS and client versions), is this because the P4's benchmark score is lower? Actually the P4 crunches more units than the Celery because it takes two at a time (hyperthreading).
RE: I've also been
)
HT P4s tend to claim too much credit. Celerons can claim odd amounts too, sometimes the lack of cache causes problems sometimes it doesn't.
BOINC WIKI
BOINCing since 2002/12/8