Except that many of the new ones that people are joining and that people are abscribing to the new systems can be explained very easy: college students being recruited as part of college efforts to participate So the increase of participants can be just an accident of the timing (the start of college season) and not a result of the new credits.

I for one, would like to see what type of crunchers substituted the kentfields, conroes , high powered AMDs and overclocked machines that left . And how any 24/7 Rosetta only crunching machines replaced the the ones that left?

Then, we can talk and really compare.

I think I tried to fit too much into a small space in my previous post.

The first point (see previous) agrees with your statement here, only with a small qualification. To say that those conroes, etc. have been replaced doesn't make any sense because it would require that someone else came in for the same reason that those left. Most of the new machines coming in would be coming in anyway (as you point out) so they can't be considered a replacement. The machines that are gone are simply gone (for the most part), a loss to the project and nothing more. They have not been replaced even by less powerful machines.

But some powerful machines might be attracted. (point two, previous) I can't think of any wonderful reasons for them to be attracted specifically by the new credit system, but we must remember that it might well play a factor. We'll have to wait a bit and see.

Hopefully those students coming in will stick around (I did). But I don't think this will be determined by one credit system or another. (point three, previous) We'll have to just wait and see.

---

Benny, and I am not accusing you: perfectly argued my foot. Lying, being hypocrites was what got some people heard:

If I remember correctly, back in May, in one of the fair credit threads, I posted, one of the big pharmers posted, and perhaps some more also requested a work based credit system. We may not have been the first; we definately weren't the last. But it was just a few posts and then either David Kim or David Baker posted in the thread and agreed with the idea; although it would have to wait until after CASP was over. After that point, it doesn't matter how many people begged, pleaded, or chanted mantras asking for what was already agreed on.

As for the the other topic.. I've been on a project that changed how they did things and archived the first credit system because of it. It didn't seem to make much difference. A few quit. But the top 100 looked much the same before and after. Zero-ing out the credit is kinda pointless from what I've seen.

---

I for one, would like to see what type of crunchers substituted the kentfields, conroes , high powered AMDs and overclocked machines that left . And how any 24/7 Rosetta only crunching machines replaced the the ones that left?

Then, we can talk and really compare.

How ironic... we could see this pretty accurately if there hadn't been such a fuss kicked up about calculating accurate credits for each decoy returned back to Feb using the credit based system!

There are some big users that have started crunching in capacity recently such as Agafonov_KY - I remember reading that TSC_Russia wouldn't run BOINC until the credit system was sorted out. Hopefully Joker! and the like will now.

I think zeroing would be a bad idea though! What I would like to see is an xml of all jobs submitted (I know it'd be huge) showing the following:

Computer ID
Result ID
WU ID
No Decoys
Time Taken

From this we could answer all the questions and also do some useful comparisons about optimal configurations such as is conroe faster than venice, is Linux faster than Win, does more RAM/cache/FSB make a difference...

---

Self edit: Double post

---

This and no other is the root from which a Tyrant springs; when he first appears he is a protector.”
Plato

I for one, would like to see what type of crunchers substituted the kentfields, conroes , high powered AMDs and overclocked machines that left . And how any 24/7 Rosetta only crunching machines replaced the the ones that left?

Then, we can talk and really compare.

How ironic... we could see this pretty accurately if there hadn't been such a fuss kicked up about calculating accurate credits for each decoy returned back to Feb using the credit based system!

There are some big users that have started crunching in capacity recently such as Agafonov_KY - I remember reading that TSC_Russia wouldn't run BOINC until the credit system was sorted out. Hopefully Joker! and the like will now.

I think zeroing would be a bad idea though! What I would like to see is an xml of all jobs submitted (I know it'd be huge) showing the following:

Computer ID
Result ID
WU ID
No Decoys
Time Taken

From this we could answer all the questions and also do some useful comparisons about optimal configurations such as is conroe faster than venice, is Linux faster than Win, does more RAM/cache/FSB make a difference...

Well that is something that could solve the issue . Alas, I wonder how much of the archived data is still there. I for one know how big the archive can be as I got a sample of the partial data for one work unit.

But one of the big problems you will start with: BOINC did not for a lot times ID the CPUs correctly, specially the high end ones.

For the record, it was not the credit for work issue that created the fights but the free and reckless use of the adjective cheater and variations of that theme by people we now know had glass ceilings in their team's houses when they were throwing stones at other teams.

---

This and no other is the root from which a Tyrant springs; when he first appears he is a protector.”
Plato

Sorry - just re-read my last post and it reads a bit arsey... It's still early here!

Well that is something that could solve the issue . Alas, I wonder how much of the archived data is still there.

But one of the big problems you will start with: BOINC did not for a lot times ID the CPUs correctly, specially the high end ones.

Jose - I know there's a problem with the CPU reporting - Intels seem to report anything - but it's not info that we need.

For any given decoys within a certain WU type the credits should be the same, whether a kentsfield or a P2 crunch it. Of course the Kentfield will crunch its decoys much quicker than the P2, but for each decoy credits should be the same. We can work out how many credits to assign to each decoy by selecting golden machines - machines of known configuration and use these to determine how many credits to assign to decoys in each WU.

---

Sorry - just re-read my last post and it reads a bit arsey... It's still early here!

Well that is something that could solve the issue . Alas, I wonder how much of the archived data is still there.

But one of the big problems you will start with: BOINC did not for a lot times ID the CPUs correctly, specially the high end ones.

Jose - I know there's a problem with the CPU reporting - Intels seem to report anything - but it's not info that we need.

For any given decoys within a certain WU type the credits should be the same, whether a kentsfield or a P2 crunch it. Of course the Kentfield will crunch its decoys much quicker than the P2, but for each decoy credits should be the same. We can work out how many credits to assign to each decoy by selecting golden machines - machines of known configuration and use these to determine how many credits to assign to decoys in each WU.

Drat to think I had to erase that data base . There was a noticeable difference in the number of decoys.

BTW for what you want to do , you need the correct id of the CPU, the correct id of the Ram, the cache and the mother board, the correct id of the OS and to known if the machine is OC.

Also , you cannot make an assumption based on small samples.

BTW This that you propose now , I was in the process of doing when the flaming against me and others started. That is something else (outside my production, then and now that I have Diablo Legion) that was lost to the project.

---

This and no other is the root from which a Tyrant springs; when he first appears he is a protector.”
Plato

Stop Rosetta. Start Rosetta-2.

maybe it should be called Rose-etc?

but jokes aside, the credit system has always been as-is. It has been improved, and that is a good thing, but to retrospectively change what people already have is not fair, as people who fall backwards in the lists will feel that they are being treated as worth less than nothing for a month or two till they catch up to where they thought they were.

I know what you will say - they would not be treated unfairly, simply be having an unfair advantage removed. But that is not how it would *feel*.

Taxes have been mentioned. OK imagine there is a change of government, and tax policy changes drastically. You feel the new policy is much more fair than the old. [and no, please do NOT give examples of fairer tax systens here - each reader pick your own please]

Do you advocate imposing the new tax codes retrospectivey back to the start of people's lives? No. However fair the new policy is, that would be outweighed by the unfairness of rocking people's boats.

The new policy started from the date it was feasible to implement it. That is how it should be, and how it should stay. Morally, practically, and in terms of minimising user dismay.

Segregation - splitting into Rosetta and Rose-etc2 also causes user dismay - SETI did it because it had to, but lok at the numbers of folk who disliked that. Actually segraegation would suit all those who have a good showing in Rosetta-1: we could resign totally from Rosetta knowing that our position in the lists would be unaffected by influxes of new people. But that is hardly a reason in favour of it, except from a very individualist persepctive.

Segragation on SETI was a one-off, excused (so far as people did excuse it) because of a change in platform. Among other things, people can be sure it will not happen again as there is no risk of SETI moving to another platofrm with another incompatible scoring system.

Segregation on Rosetta would be trust destroying. We would not know if later on another new even fairer system would be introduced, leading us on to Rose-etc-etc-3. I would be less happy to commit to a project that has proved its willingness to undermine the cerdits I have already accumulated with them. From the SETI migration rants I guess I am not the only one who feels like this.

In my not-so-umble opinion, the appropriate response to the improved credit sstem is to say thanks to the project team for working on these user issues (which do not directly affect the science at all), and to commit to staying onboard long enough that the new, fairer, credits swamp the old unfair ones.

Leaving in a huff is *silly*. In a couple of years your unfair credits will be dwarfed by the new fairer credits.

River~~

---

BTW for what you want to do , you need the correct id of the CPU, the correct id of the Ram, the cache and the mother board, the correct id of the OS and to known if the machine is OC.

Also , you cannot make an assumption based on small samples.

You don't need the specs for any computers to calculate the credits (stage 1), just as the computer specs are unknown in the new credit system. We have a bunch of golden machines, where we know the credit/hr given under the new system. Therefore for any given WU we can backtrack that for all WUs, providing we have a few 'golden machines' that returned jobs for each of the WUs crunched. We don't need each of these golden machines to have crunched each type of job - just a selection of them to have run each WU type, and that these machines haven't changed in config since a certain date (anything before that date would have to be ignored for that machine).

This can be done for all Rosetta results as it can be automated fairly easily.

e.g.



The coloured info is what we'd need to do this, and the rest can be calculated based on that.

The numbers used here are made up so the credit/decoy is all over the place, but with the real data this should be even more accurate than the current system as it indirectly relies on averaged benchmark (as it is based on the average credit for each golden machine) rather than the current system where it's only dependant on the benchmarks previously submitted.

Stage 2 would be putting config info against different machines to show the effects of different configs on Rosetta throughput.

---

Hi River

As extra info I think this would prove very useful. We could show the effect of the credit switch in real terms, and more importantly, the differences between hardware and software configs.

Does anyone actually know whether Linux is faster than Windows or vice versa? Or how a conroe stacks up agaisnt a yonah against a venice against a tualatin?

We could actually do this second part (comparing setups) just from the data post credit system switch. We just need an xml with the required data in (WU Name, No Decoys, Time Taken), and then a volunteer to play with the data to add people's computer configs against their computer IDs (or set up a web page where we could add it ourselves).

---

Stop Rosetta. Start Rosetta-2.

maybe it should be called Rose-etc?

but jokes aside, the credit system has always been as-is. It has been improved, and that is a good thing, but to retrospectively change what people already have is not fair, as people who fall backwards in the lists will feel that they are being treated as worth less than nothing for a month or two till they catch up to where they thought they were.

I know what you will say - they would not be treated unfairly, simply be having an unfair advantage removed. But that is not how it would *feel*.

Taxes have been mentioned. OK imagine there is a change of government, and tax policy changes drastically. You feel the new policy is much more fair than the old. [and no, please do NOT give examples of fairer tax systens here - each reader pick your own please]

Do you advocate imposing the new tax codes retrospectivey back to the start of people's lives? No. However fair the new policy is, that would be outweighed by the unfairness of rocking people's boats.

The new policy started from the date it was feasible to implement it. That is how it should be, and how it should stay. Morally, practically, and in terms of minimising user dismay.

Segregation - splitting into Rosetta and Rose-etc2 also causes user dismay - SETI did it because it had to, but lok at the numbers of folk who disliked that. Actually segraegation would suit all those who have a good showing in Rosetta-1: we could resign totally from Rosetta knowing that our position in the lists would be unaffected by influxes of new people. But that is hardly a reason in favour of it, except from a very individualist persepctive.

Segragation on SETI was a one-off, excused (so far as people did excuse it) because of a change in platform. Among other things, people can be sure it will not happen again as there is no risk of SETI moving to another platofrm with another incompatible scoring system.

Segregation on Rosetta would be trust destroying. We would not know if later on another new even fairer system would be introduced, leading us on to Rose-etc-etc-3. I would be less happy to commit to a project that has proved its willingness to undermine the cerdits I have already accumulated with them. From the SETI migration rants I guess I am not the only one who feels like this.

In my not-so-umble opinion, the appropriate response to the improved credit sstem is to say thanks to the project team for working on these user issues (which do not directly affect the science at all), and to commit to staying onboard long enough that the new, fairer, credits swamp the old unfair ones.

Leaving in a huff is *silly*. In a couple of years your unfair credits will be dwarfed by the new fairer credits.

River~~

River I notice you are not takig your own advice.

---

This and no other is the root from which a Tyrant springs; when he first appears he is a protector.”
Plato

It's the usual problem with this topic.

The Project admins or scientists (D. Kim or D. Baker) refuse to come on here and say one way or the other whether backdating is even possible, or if they would even consider zeroing out credits and startin over with Rosetta 2 or whatever you would call it.

Until that happens, this is just all noise.

They have, WEEKS ago - to them it is a dead issue - they have come here on this message board and said NO -

That is why this thread is so stupid, discussing a dead issue.

KevinT find that post and I'll close the thread.

tony

---

SETI.USA

River I notice you are not takig your own advice.

off topic

River I notice you are not taking your own advice.

off topic

May be so but, you made it an issue of ignoring us . I don't want people to about you because of that.

See I for one , don't want your words to : specially since you are rumoured to be a Mod

---

This and no other is the root from which a Tyrant springs; when he first appears he is a protector.”
Plato

Why is all this Cafe talk rehashed and rehashed and rehashed in the number crunching part of the fora?

A decent discussion about credits, it's fairness and different methods of granting belongs here, no doubt, but the constant wild goose chase for mods is OT in this part of the fora.

BTW for what you want to do , you need the correct id of the CPU, the correct id of the Ram, the cache and the mother board, the correct id of the OS and to known if the machine is OC.

Also , you cannot make an assumption based on small samples.

You don't need the specs for any computers to calculate the credits (stage 1), just as the computer specs are unknown in the new credit system. We have a bunch of golden machines, where we know the credit/hr given under the new system. Therefore for any given WU we can backtrack that for all WUs, providing we have a few 'golden machines' that returned jobs for each of the WUs crunched. We don't need each of these golden machines to have crunched each type of job - just a selection of them to have run each WU type, and that these machines haven't changed in config since a certain date (anything before that date would have to be ignored for that machine).

This can be done for all Rosetta results as it can be automated fairly easily.

e.g.



The coloured info is what we'd need to do this, and the rest can be calculated based on that.

The numbers used here are made up so the credit/decoy is all over the place, but with the real data this should be even more accurate than the current system as it indirectly relies on averaged benchmark (as it is based on the average credit for each golden machine) rather than the current system where it's only dependant on the benchmarks previously submitted.

Stage 2 would be putting config info against different machines to show the effects of different configs on Rosetta throughput.

You need to finish Stage 2 for the statistical analysis to have some meaning. So, get all the data from the get go .To think I had that done for a complete work unit and all the machines that ran it. Better still, after a cluster analysis, I had ID the machines that deviated significantly from the clusters of machine/cpu/mobos combos that were studied.( upwards and downwards) .

Good Luck

---

This and no other is the root from which a Tyrant springs; when he first appears he is a protector.”
Plato

Why is all this Cafe talk rehashed and rehashed and rehashed in the number crunching part of the fora?

A decent discussion about credits, it's fairness and different methods of granting belongs here, no doubt, but the constant wild goose chase for mods is OT in this part of the fora.

Sanger:

I was just reminding River~ of the promise he made . What is OT is trying to reopen something that was closed by the developers using any types of subterfuges. OT and with the potential of nastiness.

I hope you noticed my dialogue with dcdc. See what he is proposing now, I did before the days of the turbulence. So , If there is a person here that knows what he is talking about , that could enter a dialogue with dcdc (other than the developers , specially David Kim) is me.

But if nastiness sprouts, you can count that there will be no dialogue. So why dont we make an effort to prevent the sprouting.

---

This and no other is the root from which a Tyrant springs; when he first appears he is a protector.”
Plato

Tony, I apologise I appear to have taken this thread off topic but it is merely to show the type of hypocrisy that is/was going on.Sorry there is more...

Biggles said like an angel...

What about the math based on XS having advocated the use of optimised clients since the 31st of December 2005? This thread has talked about optimised clients and encouraged the use of them since the day after it was posted. I don't believe that it took three and a half months for XS to notice that part of the setup guide. Optimised clients and the subsequent overclaiming of credit was widespread before April.

But failed to disclose to this forum on 5th September..

But I'm not here to tell you what to do. Crunch3r's 5.5.0 client can be found here(link to crunch3rs client). Just be aware that the credit you claim with it will be far higher than you will be awarded and that some people view it's use as cheating. I just don't feel it's relevant anymore as both SETI and Rosetta have moved on and it does nothing apart from inflate benchmarks. It's still on some of my PCs, but will come off in time when I visit them.

http://episteme.arstechnica.com/eve/forums/a/tpc/f/122097561/m/566008780831 3rd post down.

I just want to get rid of the hypocrisy of some and the idea it was 2-5 teams using this client, it is patently untrue and XS took the brunt of it, moved on and declared it illegal where they were concerned which appears more than any other team have done.

Saying I don`t think it`s a good idea then providing a link to it seem`s pretty silly imho.

Me, I think the whole of Boinc should be wiped out and started again, after all it`s only fair and right eh ?

How can you accuse me of being a hypocrite? I explicitly stated in the post you PARTIALLY quoted "With that in mind I no longer recommend that people use Crunch3r's 5.5.0 BOINC client."

It was mentioned only in the interests of completeness and to give a better understanding before the guy came across a reference to it. I even linked back to the R@H boards and my own posts about why it was bad.

---

Tony, I apologise I appear to have taken this thread off topic but it is merely to show the type of hypocrisy that is/was going on.Sorry there is more...

Biggles said like an angel...

What about the math based on XS having advocated the use of optimised clients since the 31st of December 2005? This thread has talked about optimised clients and encouraged the use of them since the day after it was posted. I don't believe that it took three and a half months for XS to notice that part of the setup guide. Optimised clients and the subsequent overclaiming of credit was widespread before April.

But failed to disclose to this forum on 5th September..

But I'm not here to tell you what to do. Crunch3r's 5.5.0 client can be found here(link to crunch3rs client). Just be aware that the credit you claim with it will be far higher than you will be awarded and that some people view it's use as cheating. I just don't feel it's relevant anymore as both SETI and Rosetta have moved on and it does nothing apart from inflate benchmarks. It's still on some of my PCs, but will come off in time when I visit them.

http://episteme.arstechnica.com/eve/forums/a/tpc/f/122097561/m/566008780831 3rd post down.

I just want to get rid of the hypocrisy of some and the idea it was 2-5 teams using this client, it is patently untrue and XS took the brunt of it, moved on and declared it illegal where they were concerned which appears more than any other team have done.

Saying I don`t think it`s a good idea then providing a link to it seem`s pretty silly imho.

Me, I think the whole of Boinc should be wiped out and started again, after all it`s only fair and right eh ?

How can you accuse me of being a hypocrite? I explicitly stated in the post you PARTIALLY quoted "With that in mind I no longer recommend that people use Crunch3r's 5.5.0 BOINC client."

It was mentioned only in the interests of completeness and to give a better understanding before the guy came across a reference to it. I even linked back to the R@H boards and my own posts about why it was bad.

Biggles you dont know when to quit, don't you?

Want to keep the issue going? Do so at your risk.

---

This and no other is the root from which a Tyrant springs; when he first appears he is a protector.”
Plato

Now is probably a good time to remind people of the saying, "Don't sweat the little stuff. <long pause> It's all little stuff."

---

Rosetta Moderator: Mod.Tymbrimi
ROSETTA@home FAQ
Moderator Contact