Message boards : Number crunching : Why my Granted Credit is just the half GC than other who did same work?
Message board moderation

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
Sepionet

Send message
Joined: 3 Sep 05
Posts: 27
Credit: 35,521
RAC: 0
Message 12208 - Posted: 21 Jan 2006, 12:08:05 UTC

I'm using optimized seti clients and boinc clients so my computers ask more credit than they must to, but what about ignoring my Claimed Credit instead of granting me the half of the median of the C.C.

I'm physicist and I know people working at CERN/ LHC. I'm the first in crunching power in Spain i have 12 PC only crunching workunits, it's a good donation i think, so why the hell they penalize me.

I've corrected this with the fair Truxoft Boinc Callibrated Client.

The only perturbations on global median ara around 10% if I claim double the credit I do, so the correrctive factor is 1/1.10 and it isn't 1/2.

Sorry, but I'm very uppset with this because, i have 2500 C.C. pending that will transform in only 800
G.C. It's not fair
ID: 12208 · Report as offensive     Reply Quote
alpina

Send message
Joined: 3 Aug 05
Posts: 49
Credit: 143,072
RAC: 0
Message 12214 - Posted: 21 Jan 2006, 13:52:17 UTC

I don't really see the problem. You got the same credit as the other hosts that completed your workunits, that's fair isn't it? Or what am I missing?


BOINC.BE: The team for Belgians and their friends who love the smell of glowing red cpu's in the morning
ID: 12214 · Report as offensive     Reply Quote
Profile Keck_Komputers

Send message
Joined: 1 Sep 04
Posts: 275
Credit: 2,652,452
RAC: 0
Message 12215 - Posted: 21 Jan 2006, 14:05:27 UTC

There is no reason to ask a question on every board. The other thread had a basically correct answer. LHC grants partial credit for invalid and late work sometimes. So you got a bonus not a penalty.
BOINC WIKI

BOINCing since 2002/12/8
ID: 12215 · Report as offensive     Reply Quote
J D K

Send message
Joined: 27 Aug 05
Posts: 50
Credit: 24,055
RAC: 0
Message 12229 - Posted: 21 Jan 2006, 18:54:26 UTC

First of all you are claiming a lot more than you should by using the optimized CC, and some people consider this cheating or borderline cheating...

The credits are awarded after a quorum is met and the high score and low score are removed and the middle score is awarded to everyone, so I see this as fair.....
____________
ID: 12229 · Report as offensive     Reply Quote
Michael Karlinsky
Avatar

Send message
Joined: 18 Sep 04
Posts: 163
Credit: 1,682,370
RAC: 0
Message 12232 - Posted: 21 Jan 2006, 19:20:11 UTC - in response to Message 12229.  
Last modified: 21 Jan 2006, 19:22:03 UTC

First of all you are claiming a lot more than you should by using the optimized CC, and some people consider this cheating or borderline cheating...


Luckily this is coming to an end with the intruduction of FLOPS counting. Currently tested at SETI BETA, to be introduced with SETI Enhanced, and the upcoming Malaria Control. Bruce from E@H is looking into it too.

Michael
Team Linux Users Everywhere
ID: 12232 · Report as offensive     Reply Quote
Sepionet

Send message
Joined: 3 Sep 05
Posts: 27
Credit: 35,521
RAC: 0
Message 12297 - Posted: 23 Jan 2006, 10:57:40 UTC - in response to Message 12214.  

I don't really see the problem. You got the same credit as the other hosts that completed your workunits, that's fair isn't it? Or what am I missing?

i claim double, median is the half of it because my client claims more credits in other projects. but they granted me only the half of the median.

http://lhcathome.cern.ch/results.php?hostid=100692

new results arrived first or second to database and they granted me half the credit.

if we calculate median of the values my result doesn't affect this value.
don't matters if the max or min value are infinity or 0.

now i claim better the credit with truxoft callibrated client

http://boinc.truxoft.com/core-cal.htm
ID: 12297 · Report as offensive     Reply Quote
Profile Mr.Pernod
Avatar

Send message
Joined: 16 Jul 05
Posts: 65
Credit: 369,728
RAC: 0
Message 12299 - Posted: 23 Jan 2006, 11:10:40 UTC - in response to Message 12297.  

i claim double, median is the half of it because my client claims more credits in other projects. but they granted me only the half of the median.

http://lhcathome.cern.ch/results.php?hostid=100692

new results arrived first or second to database and they granted me half the credit.

if we calculate median of the values my result doesn't affect this value.
don't matters if the max or min value are infinity or 0.

now i claim better the credit with truxoft callibrated client

http://boinc.truxoft.com/core-cal.htm

http://lhcathome.cern.ch/result.php?resultid=5454844 Validate state Invalid
http://lhcathome.cern.ch/result.php?resultid=5454797 Validate state Invalid
http://lhcathome.cern.ch/result.php?resultid=5454770 Validate state Invalid

If I were you I would have a good look at my computer with prime95 and memtest.
you are returning a lot of invalid results, are you overclocking?
ID: 12299 · Report as offensive     Reply Quote
J D K

Send message
Joined: 27 Aug 05
Posts: 50
Credit: 24,055
RAC: 0
Message 12304 - Posted: 23 Jan 2006, 12:16:20 UTC
Last modified: 23 Jan 2006, 12:16:46 UTC

One more time, the way it works three to make a quorum, you take the high score and the low score and delete them and you get what is left, the middle score, simple but effective....
ID: 12304 · Report as offensive     Reply Quote
jsprague78

Send message
Joined: 10 Dec 05
Posts: 12
Credit: 8,932
RAC: 0
Message 12306 - Posted: 23 Jan 2006, 12:48:45 UTC - in response to Message 12304.  

One more time, the way it works three to make a quorum, you take the high score and the low score and delete them and you get what is left, the middle score, simple but effective....



Yep, workunits get sent out to 3 computers, more if needed but on the credit the lowest amount of claimed credit gets tossed and the highest amount of claimed credit gets tossed.

I followed the link you provided to your results list and it appears that the reason you are getting less than the granted credit when you do is because of the CPU times, the credit you're getting is more on par with those times than what you trying to claim. I did see a some workunits where that's not the case and I honestly have no idea why that is.

I wouldn't exactly call the Truxoft Boinc Callibrated Client fair as you did since it lies and inflates the claimed credit in an effort to put you ahead of the others on the rankings list. Can we say Steroids for BOINC.

Just ignore the credit and do it for the science.
ID: 12306 · Report as offensive     Reply Quote
jsprague78

Send message
Joined: 10 Dec 05
Posts: 12
Credit: 8,932
RAC: 0
Message 12307 - Posted: 23 Jan 2006, 12:53:47 UTC - in response to Message 12208.  

Just to say this last thing on it you mentioned in the Cafe LHC thread of yours that you don't care about the credit but obviously you do otherwise you wouldn't have posted about it in every forum and be using the client you are.

As far as the differences in credit goes have you tried just using the normal LHC client to see if the lower credit being given continues? If it does then at least you know it's not the client you're using now but if it doesn't then you'll have your answer.
ID: 12307 · Report as offensive     Reply Quote
Profile Mr.Pernod
Avatar

Send message
Joined: 16 Jul 05
Posts: 65
Credit: 369,728
RAC: 0
Message 12308 - Posted: 23 Jan 2006, 13:02:33 UTC

this is not about the credit mechanism or quorums.

he is returning invalid results and LHC is kind enough to grant partial credit to invalid results instead of chucking them out the window and grant 0 credit for them like most other projects do.

he should check his pc and make sure there are no defective or overclocked-beyond-the-borders parts.
ID: 12308 · Report as offensive     Reply Quote
alpina

Send message
Joined: 3 Aug 05
Posts: 49
Credit: 143,072
RAC: 0
Message 12309 - Posted: 23 Jan 2006, 13:38:25 UTC - in response to Message 12297.  

I don't really see the problem. You got the same credit as the other hosts that completed your workunits, that's fair isn't it? Or what am I missing?

i claim double, median is the half of it because my client claims more credits in other projects. but they granted me only the half of the median.

http://lhcathome.cern.ch/results.php?hostid=100692

new results arrived first or second to database and they granted me half the credit.

if we calculate median of the values my result doesn't affect this value.
don't matters if the max or min value are infinity or 0.

now i claim better the credit with truxoft callibrated client

http://boinc.truxoft.com/core-cal.htm


Sorry, I misunderstood. Like everyone else allready mentioned, you should take a look at your PC because it's returning invalid results. You might want to take a look at the CPU-temperature. Did you overclok the cpu?


BOINC.BE: The team for Belgians and their friends who love the smell of glowing red cpu's in the morning
ID: 12309 · Report as offensive     Reply Quote
J D K

Send message
Joined: 27 Aug 05
Posts: 50
Credit: 24,055
RAC: 0
Message 12312 - Posted: 23 Jan 2006, 15:35:13 UTC
Last modified: 23 Jan 2006, 15:35:42 UTC

Dud, good catch Mr Pernod..

Maybe this will help him, this is from the FAQs;

2.5 How and when do I get credit?
The new validator works like this:

A workunit gets validated when it has 3 identical results.
Users that returned identical results will get same amount of credit, which is an average of claimed credit when lowest and highest value is ignored. In most cases this is the median of the claimed credits. This credit is called "canonical credit".
If work unit has 10 successful results but there are not enough identical results, the results are granted points according to how well they match with other results. The best-matching result will receive median average of claimed credits (calculated in same way as the canonical credit). Other results will get credit proportional to the best credit and result's match points. Although users get credit from this kind of results, the results are not used in physical studies.
Please note that different CPU architectures may yield different floating point results. This is especially true between Pentium and Athlon XP CPUs. Although differing results are not used in physical studies, they may still be useful to us in finding possible problems in our software. Of course, this requires that the CPU used for calculations doesn't produce real errors. This can happen especially on overclocked CPUs. Even if the overclocked system seems to work fine (maybe an unexplained crash once in a month or so) it might skip one bit in one of the 10^14 floating point calculations commonly done in a work unit, producing highly different end result. Such a bad results will actually cause gray hairs to us, so if you have overclocked CPU and get lots of unvalidated results, consider lowering the clock speed or detaching from the project.
ID: 12312 · Report as offensive     Reply Quote
Profile Paul D. Buck

Send message
Joined: 2 Sep 04
Posts: 545
Credit: 148,912
RAC: 0
Message 12318 - Posted: 23 Jan 2006, 17:08:11 UTC

I need to update the Wiki I think, this is another exception rule ...
ID: 12318 · Report as offensive     Reply Quote
Profile Ananas

Send message
Joined: 17 Jul 05
Posts: 102
Credit: 542,016
RAC: 0
Message 12327 - Posted: 23 Jan 2006, 19:39:58 UTC - in response to Message 12312.  
Last modified: 23 Jan 2006, 19:46:12 UTC

... Such a bad results will actually cause gray hairs to us, ....


That's not "grey hair", it's a graphical method to show the experience level. If you have got #808080, you reached the maximum level.

Another valid method would be visual cortex enhancement devices - people, who don't understand them, call them "wrinkles".

p.s.: Ties have never validated OK in this context.
ID: 12327 · Report as offensive     Reply Quote
senatoralex85

Send message
Joined: 17 Sep 05
Posts: 60
Credit: 4,221
RAC: 0
Message 12334 - Posted: 23 Jan 2006, 22:24:09 UTC - in response to Message 12327.  

... Such a bad results will actually cause gray hairs to us, ....


That's not "grey hair", it's a graphical method to show the experience level. If you have got #808080, you reached the maximum level.

Another valid method would be visual cortex enhancement devices - people, who don't understand them, call them "wrinkles".

p.s.: Ties have never validated OK in this context.


----------------------------------------------------------------------

What is a calibrated client? How does it work? Does it improve the speed at which work units are created? Someone mentioned using it is cheating, why?

ID: 12334 · Report as offensive     Reply Quote
jsprague78

Send message
Joined: 10 Dec 05
Posts: 12
Credit: 8,932
RAC: 0
Message 12337 - Posted: 23 Jan 2006, 23:33:00 UTC - in response to Message 12334.  

... Such a bad results will actually cause gray hairs to us, ....


That's not "grey hair", it's a graphical method to show the experience level. If you have got #808080, you reached the maximum level.

Another valid method would be visual cortex enhancement devices - people, who don't understand them, call them "wrinkles".

p.s.: Ties have never validated OK in this context.


----------------------------------------------------------------------

What is a calibrated client? How does it work? Does it improve the speed at which work units are created? Someone mentioned using it is cheating, why?


Well that depends really, there's optimized clients out there that allow the computer to better utilize the specific aspects of your CPU and whatever instruction set it's best at and they're not available for all projects.

Depending on the nature of those clients they either will just speed up the time that your computer completes a workunit thereby giving you less credit per workunit but allowing you to complete more workunits and having roughly the same amount of credit overall. I use one like this for SETI@Home and it shaves an hour off my workunit times so now they only take about 1hr 15min. This to me is the honest type of optimized client as it's not done for the purpose of getting ahead but for the purpose of working the science faster and allowing projects to be completed faster and become of benefit to humanity faster.

The other type of optimized client will do the same thing as the one above but will report a inflated completion time, for instance this result of mine on this project...

Work unit

You'll see that I had the median time of 10,548.50 seconds and the median amount of claimed credit of 33.39 that was given to everyone. The second type of optimized clients would've reported a time greater than that, in some cases about double, but either way it'd be a false amount of time. The only purpose of that is to try to raise in the rankings on the project and has no other benefit.

Not to mention that if I remember correctly you have to be careful with some of those optimized clients as they can return invalid results which is more than likely what the original poster is having happen hence their question about the credit.

I always say that if you're going to participate in a project then do it for the science and not the credit and stick with the default client. Or at least just stick to the first type of optimized client and research it first before you download it and install it to make sure it's actually going to work correctly.
ID: 12337 · Report as offensive     Reply Quote
jsprague78

Send message
Joined: 10 Dec 05
Posts: 12
Credit: 8,932
RAC: 0
Message 12338 - Posted: 23 Jan 2006, 23:33:45 UTC - in response to Message 12335.  

Bah sorry, forgot to mention that a calibrated client is the same thing as an optimized client.
ID: 12338 · Report as offensive     Reply Quote
Profile Paul D. Buck

Send message
Joined: 2 Sep 04
Posts: 545
Credit: 148,912
RAC: 0
Message 12339 - Posted: 23 Jan 2006, 23:57:00 UTC

um ... sorry, I have to disagree with the explanation.

An Optimized BOINC Client is compiled with a specific processor in mind and specified with compiler switches. This makes the software run as fast as possible. The primary effect of this is to increase the benchmark scores. This results in higher credit claims.

An optimized Science Application is a version of the science program compiled with a specific processor in mind. This makes the software run as fast as possible. The primary effect is to complete work faster. As a consequence the credit claim will be lower (shorter time). Some hold that this "cheats" them of credit.

So, they want to run an optimized BOINC Client with the Optimized science applications to bring everything back into "balance". The problem is that there is an implicit assumption that the optimization of the BOINC Client is indeed proportional to the speed up of the science applications. It may or may not be true. Also, different science applications run with better or worse efficiency on specific CPUs. So, optimization to "balance" for SETI@Home may unbalance Einstein@Home

Worse, for Rosetta@Home the optimized BOINC CLient will grossly over-claim. Thus, the argument about this is cheating, or not ...

Trux, to sidestep this made a modified BOINC Cleint that will "adjust" over time the claims to be in line with the supposed standard work unit which is usually worth 32 Cobblestones. In theory, the calibration will only come into play when it is used with an optimized Science Application. It does not change the benchmark scores directly. Rather it fiddles with the reported numbers when the Result is reported to make the claim "accurate".

Though I have no problem with this in a very theoretical sense, the problem is that ALL of these areguments are based on an assumption that the person that is releasing the modified client really knows what the credit claim should be ... and this is where I have the difficulty ... something about hubris ...

The good news is that the newer method of counting FLOPS seems to be more stable and perhaps we can put this nightmare behind us ... though I have a feeling that the cat is out of the bag and the argument is still going to be that the modifiers know more about what the claim and grants of credit should be than the projects. So, long term, I suspect that something along the lines of my proposed calibration mechanism is going to have to be implemented ... then the testing and certification will be done double-blind ...

There is a lot about this in the Wiki, just wander about reading up on credit and validation process ... tons more ...
ID: 12339 · Report as offensive     Reply Quote
jsprague78

Send message
Joined: 10 Dec 05
Posts: 12
Credit: 8,932
RAC: 0
Message 12341 - Posted: 24 Jan 2006, 1:31:06 UTC - in response to Message 12339.  

um ... sorry, I have to disagree with the explanation.

An Optimized BOINC Client is compiled with a specific processor in mind and specified with compiler switches. This makes the software run as fast as possible. The primary effect of this is to increase the benchmark scores. This results in higher credit claims.

An optimized Science Application is a version of the science program compiled with a specific processor in mind. This makes the software run as fast as possible. The primary effect is to complete work faster. As a consequence the credit claim will be lower (shorter time). Some hold that this "cheats" them of credit.

So, they want to run an optimized BOINC Client with the Optimized science applications to bring everything back into "balance". The problem is that there is an implicit assumption that the optimization of the BOINC Client is indeed proportional to the speed up of the science applications. It may or may not be true. Also, different science applications run with better or worse efficiency on specific CPUs. So, optimization to "balance" for SETI@Home may unbalance Einstein@Home

Worse, for Rosetta@Home the optimized BOINC CLient will grossly over-claim. Thus, the argument about this is cheating, or not ...

Trux, to sidestep this made a modified BOINC Cleint that will "adjust" over time the claims to be in line with the supposed standard work unit which is usually worth 32 Cobblestones. In theory, the calibration will only come into play when it is used with an optimized Science Application. It does not change the benchmark scores directly. Rather it fiddles with the reported numbers when the Result is reported to make the claim "accurate".

Though I have no problem with this in a very theoretical sense, the problem is that ALL of these areguments are based on an assumption that the person that is releasing the modified client really knows what the credit claim should be ... and this is where I have the difficulty ... something about hubris ...

The good news is that the newer method of counting FLOPS seems to be more stable and perhaps we can put this nightmare behind us ... though I have a feeling that the cat is out of the bag and the argument is still going to be that the modifiers know more about what the claim and grants of credit should be than the projects. So, long term, I suspect that something along the lines of my proposed calibration mechanism is going to have to be implemented ... then the testing and certification will be done double-blind ...

There is a lot about this in the Wiki, just wander about reading up on credit and validation process ... tons more ...


Ah thanks for the clarification, I was trying to hurry and answer his question before I had to leave and forgot that there is the optimized BOINC client part and then the optimized project specific part. The SETI@Home one I'm using is just a project specific one. If you pull up any of my workunits to see how it affects it I should mention that right before I started running it I also changed out my motherboard/processor from a 2.4ghz P4 to a AMD 3600 with an actual clock speed of 2.0 ghz with HyperTransport.
ID: 12341 · Report as offensive     Reply Quote
1 · 2 · Next

Message boards : Number crunching : Why my Granted Credit is just the half GC than other who did same work?


©2024 CERN